WO2023032274A1 - 情報処理装置、情報処理方法、及び、プログラム - Google Patents

情報処理装置、情報処理方法、及び、プログラム Download PDF

Info

Publication number
WO2023032274A1
WO2023032274A1 PCT/JP2022/009864 JP2022009864W WO2023032274A1 WO 2023032274 A1 WO2023032274 A1 WO 2023032274A1 JP 2022009864 W JP2022009864 W JP 2022009864W WO 2023032274 A1 WO2023032274 A1 WO 2023032274A1
Authority
WO
WIPO (PCT)
Prior art keywords
image
detection
cis
mode
sensing
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Ceased
Application number
PCT/JP2022/009864
Other languages
English (en)
French (fr)
Japanese (ja)
Inventor
悠史 西牧
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Sony Semiconductor Solutions Corp
Original Assignee
Sony Semiconductor Solutions Corp
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Sony Semiconductor Solutions Corp filed Critical Sony Semiconductor Solutions Corp
Priority to JP2023545031A priority Critical patent/JPWO2023032274A1/ja
Priority to CN202280057200.3A priority patent/CN117859338A/zh
Priority to US18/291,939 priority patent/US20240346844A1/en
Publication of WO2023032274A1 publication Critical patent/WO2023032274A1/ja
Anticipated expiration legal-status Critical
Ceased legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
    • G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
    • G06V40/16—Human faces, e.g. facial parts, sketches or expressions
    • G06V40/172—Classification, e.g. identification
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F21/00—Security arrangements for protecting computers, components thereof, programs or data against unauthorised activity
    • G06F21/30—Authentication, i.e. establishing the identity or authorisation of security principals
    • G06F21/31—User authentication
    • G06F21/32—User authentication using biometric data, e.g. fingerprints, iris scans or voiceprints
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F21/00—Security arrangements for protecting computers, components thereof, programs or data against unauthorised activity
    • G06F21/70—Protecting specific internal or peripheral components, in which the protection of a component leads to protection of the entire computer
    • G06F21/82—Protecting input, output or interconnection devices
    • G06F21/84—Protecting input, output or interconnection devices output devices, e.g. displays or monitors
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06K—GRAPHICAL DATA READING; PRESENTATION OF DATA; RECORD CARRIERS; HANDLING RECORD CARRIERS
    • G06K7/00—Methods or arrangements for sensing record carriers, e.g. for reading patterns
    • G06K7/10—Methods or arrangements for sensing record carriers, e.g. for reading patterns by electromagnetic radiation, e.g. optical sensing; by corpuscular radiation
    • G06K7/10544—Methods or arrangements for sensing record carriers, e.g. for reading patterns by electromagnetic radiation, e.g. optical sensing; by corpuscular radiation by scanning of the records by radiation in the optical part of the electromagnetic spectrum
    • G06K7/10712—Fixed beam scanning
    • G06K7/10722—Photodetector array or CCD scanning
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06K—GRAPHICAL DATA READING; PRESENTATION OF DATA; RECORD CARRIERS; HANDLING RECORD CARRIERS
    • G06K7/00—Methods or arrangements for sensing record carriers, e.g. for reading patterns
    • G06K7/10—Methods or arrangements for sensing record carriers, e.g. for reading patterns by electromagnetic radiation, e.g. optical sensing; by corpuscular radiation
    • G06K7/14—Methods or arrangements for sensing record carriers, e.g. for reading patterns by electromagnetic radiation, e.g. optical sensing; by corpuscular radiation using light without selection of wavelength, e.g. sensing reflected white light
    • G06K7/1404—Methods for optical code recognition
    • G06K7/1408—Methods for optical code recognition the method being specifically adapted for the type of code
    • G06K7/1413—1D bar codes
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06K—GRAPHICAL DATA READING; PRESENTATION OF DATA; RECORD CARRIERS; HANDLING RECORD CARRIERS
    • G06K7/00—Methods or arrangements for sensing record carriers, e.g. for reading patterns
    • G06K7/10—Methods or arrangements for sensing record carriers, e.g. for reading patterns by electromagnetic radiation, e.g. optical sensing; by corpuscular radiation
    • G06K7/14—Methods or arrangements for sensing record carriers, e.g. for reading patterns by electromagnetic radiation, e.g. optical sensing; by corpuscular radiation using light without selection of wavelength, e.g. sensing reflected white light
    • G06K7/1404—Methods for optical code recognition
    • G06K7/1408—Methods for optical code recognition the method being specifically adapted for the type of code
    • G06K7/1417—2D bar codes
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T3/00—Geometric image transformations in the plane of the image
    • G06T3/60—Rotation of whole images or parts thereof
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/70—Determining position or orientation of objects or cameras
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/10—Image acquisition
    • G06V10/12—Details of acquisition arrangements; Constructional details thereof
    • G06V10/14—Optical characteristics of the device performing the acquisition or on the illumination arrangements
    • G06V10/147—Details of sensors, e.g. sensor lenses
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
    • G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
    • G06V40/16—Human faces, e.g. facial parts, sketches or expressions
    • G06V40/161—Detection; Localisation; Normalisation
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N5/00—Details of television systems
    • H04N5/66—Transforming electric information into light information
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30196—Human being; Person
    • G06T2207/30201—Face

Definitions

  • the present technology relates to an information processing device, an information processing method, and a program, and more particularly to an information processing device, an information processing method, and a program that can reduce power consumption that impairs convenience.
  • Patent Document 1 discloses a technique for switching a mobile terminal from a normal mode to a low power consumption power saving mode based on the result of face recognition using an image captured by a camera.
  • the power consumption of the imaging unit and the arithmetic unit in the terminal becomes a problem.
  • An information processing device or a program according to the present technology is an image recognition unit included in an image sensor, the image recognition unit detecting that a predetermined detection target image is included in an image captured by the image sensor. and a processing unit that acquires an output image output from the image sensor and executes predetermined processing on the output image, wherein the image recognition unit detects that the detection target image is included.
  • an information processing device having a processing unit that causes the image sensor to output the output image when the output image is detected, or a program for causing a computer to function as such an information processing device.
  • the image recognition unit of an information processing device having an image sensor, an image recognition unit of the image sensor, and a processing unit detects a predetermined image in a captured image captured by the image sensor. detecting that an image to be detected is included, acquiring an output image output from the image sensor, performing predetermined processing on the output image, and detecting that the image to be detected is included by the image recognition unit;
  • the information processing method causes the image sensor to output the output image when it is detected that the output image is detected.
  • the information processing device In the information processing device, the information processing method, and the program of the present technology, it is detected that a predetermined detection target image is included in an image captured by an image sensor, and an output image is output from the image sensor. is obtained, a predetermined process is performed on the output image, and the output image is output by the image sensor when it is detected that the image to be detected is included.
  • FIG. 1 is a block diagram illustrating a schematic configuration of a terminal device to which the present technology is applied;
  • FIG. 2 is a block diagram illustrating the configuration of a CIS;
  • FIG. 4 is a diagram illustrating a first usage pattern of CIS image recognition and AP image recognition;
  • FIG. 10 is a diagram illustrating a second usage pattern of CIS image recognition and AP image recognition;
  • FIG. 11 is a diagram illustrating a third usage pattern of CIS image recognition and AP image recognition;
  • FIG. 11 is a diagram illustrating a fourth usage pattern of CIS image recognition and AP image recognition;
  • FIG. 10 is a diagram for explaining image recognition performed by the CIS and the AP when the terminal is unlocked;
  • FIG. 2 is a diagram illustrating an example (example 1) of a specific situation of a terminal unlock function and an outline of state transitions of CIS and AP at that time;
  • FIG. 10 is a diagram illustrating an example (example 2) of a specific situation of terminal lock and terminal unlock functions and an overview of state transitions of CIS and AP at that time;
  • FIG. 10 is a flowchart showing state transitions as a terminal device and state transitions of CIS and AP in a terminal unlock function;
  • FIG. 2 is a flowchart showing state transitions as a terminal device and state transitions of a CIS and an AP in a terminal lock function;
  • FIG. 11 is a sequence diagram illustrating a procedure for terminal lock and terminal unlock;
  • FIG. 10 is a flow chart showing a processing procedure performed by CIS and AP in terminal lock
  • FIG. 10 is a flow chart showing a processing procedure performed by CIS and AP in terminal unlock
  • FIG. FIG. 4 is a diagram for explaining image recognition performed by CIS and AP for Smart rotate function
  • FIG. 10 is a flow chart showing a processing procedure performed by CIS and AP in Smart rotate
  • FIG. 10 is a diagram for explaining image recognition performed by the CIS and AP 12 with respect to the moving image reproduction control function
  • FIG. 10 is a sequence diagram illustrating a processing procedure of moving image reproduction control
  • FIG. 19 is a sequence diagram illustrated for comparison with FIG. 18
  • 4 is a flow chart showing a processing procedure performed by CIS and AP in video playback control.
  • FIG. 4 is a diagram for explaining image recognition performed by CIS and AP for the function of peeping prevention;
  • FIG. 11 is a flow chart showing a processing procedure performed by CIS and AP in Peeping prevention;
  • FIG. 3 is a diagram illustrating an example of a specific situation of the voice/video call automatic transition function and an overview of state transitions of the CIS and AP at that time;
  • 4 is a flow chart showing processing procedures performed by CIS and AP in a voice/video call;
  • 10 is a flow chart showing a processing procedure performed by CIS and AP in screen luminance automatic adjustment;
  • FIG. 4 is a diagram illustrating an example of a specific situation of the pet camera function and an outline of state transitions of CIS and AP at that time; 4 is a flow chart showing a processing procedure performed by CIS and AP in the pet camera; FIG. 4 is a diagram illustrating an example of a specific situation of the security doorphone function and an outline of state transitions of the CIS and AP at that time; 4 is a flow chart showing a processing procedure performed by a CIS and an AP in a security doorphone; FIG. 4 is a diagram illustrating an example of a specific situation of a function of application activation by 2D code and an overview of state transitions of CIS and AP at that time; FIG. 10 is a flow chart showing a processing procedure performed by CIS and AP in the function of application activation by 2D code; FIG.
  • FIG. 1 is a block diagram illustrating a schematic configuration of a terminal device to which the present technology is applied.
  • a terminal device 1 is, for example, a mobile terminal such as a smart phone, a mobile phone, a notebook computer, or a tablet.
  • the terminal device 1 is not limited to a mobile terminal and may be any type of information processing device.
  • the terminal device 1 has a CIS (CMOS Image Sensor) 11, an AP (Application Processor) 12, a storage 13, and the like.
  • the CIS 11 is included in the camera (imaging unit) included in the terminal device 1.
  • the CIS 11 is a one-chip imaging device having a light receiving section for performing photoelectric conversion and a signal processing section for performing signal processing. It is assumed that the terminal device 1 is mainly a smart phone.
  • the terminal device 1 includes a display unit (display), and the camera is configured such that the user (especially the head) viewing the display in a normal state enters the angle of view (shooting range) of the camera (CIS 11). Assume that it is installed in the device 1 .
  • the AP 12 is an arithmetic processing unit that the terminal device 1 has.
  • the AP 12 executes not only the imaging unit including the CIS 11, but also the processing related to each unit such as the input unit, display unit, power supply unit, storage unit, communication unit, and the like of the terminal device 1, application processing, and the like according to a program.
  • the AP 12 performs processing related to switching of the operating state of the terminal device 1 due to a notification signal such as a trigger from the CIS 11 .
  • the operating state of the terminal device 1 includes a startup state that is the normal operating state of the terminal device 1, a power saving state that consumes less power than the startup state, and an unlock state that restores the original operating state or a predetermined operating state. It represents a predetermined state such as a locked state that requires a lock operation.
  • the storage 13 is a storage unit provided in the terminal device 1 and stores various data.
  • the CIS 11 has a preprocessing unit 21, a detection processing unit 22, a ROM (Read Only Memory) 23, and a RAM (Random Access Memory) 24.
  • the preprocessing unit 21 includes a light receiving unit in the CIS 11, and an image formed by an optical system (not shown) is input to the light receiving unit (light receiving element).
  • the preprocessing unit 21 performs predetermined signal processing on an image that has been input to the light receiving unit and photoelectrically converted, and supplies the processed image to the detection processing unit 22 as image data.
  • image data for one screen (one frame) obtained by photoelectric conversion by the light receiving unit and whose vertical and horizontal pixel sizes are determined is simply referred to as an image or a captured image.
  • the range of a subject captured in an image is also called an image frame (or angle of view). Acquiring an image through photoelectric conversion by the light receiving unit is also called imaging.
  • the detection processing unit 22 includes a processing unit that performs arithmetic processing (referred to as DNN processing) based on an inference model having a DNN (Deep Neural Network) structure.
  • the detection processing unit 22 performs DNN processing on the image data from the preprocessing unit 21 .
  • DNN processing within the range of the image captured by the light receiving unit (the image input to the preprocessing unit 21), that is, within the image frame, the range of the target area detected as the person (person's image) and the , the degree of certainty that the target is a person (the degree of certainty that the object is a person), etc. are obtained as detection results.
  • the object to be detected by DNN processing is not limited to humans.
  • the detection result of DNN processing also differs depending on the structure of the inference model used for DNN processing.
  • the detection processing unit 22 supplies an activation trigger (detection trigger) and a detection result to the AP 12 based on the detection result of the DNN processing.
  • the ROM 23 stores data that can only be read, such as the data of the operating program of the CIS 11 .
  • Inference model data used for DNN processing in the detection processing unit 22 is also stored in the ROM 23 and read out by the detection processing unit 22 .
  • the RAM 24 temporarily stores data used or generated by the CIS 11 in a writable and readable manner. Inference model data used for DNN processing in the detection processing unit 22 can be written from the AP 12 to the RAM 24 in addition to being written to the ROM 23 at the time of manufacture.
  • one or more types of predetermined detection targets are detected from within the captured image (within the image frame), and the position (position or range of the detection target region) is determined.
  • Detecting is called object detection.
  • the detection target is a person, it is called human detection, and when the detection target is a face, it is called face detection.
  • face recognition When the detection target is the face of a specific person, it is called face recognition.
  • Face authentication refers to comparing a face detected by face detection with a face (face image) of a specific person prepared in advance and determining whether the detected face is the face of a specific person.
  • image recognition is a general term for all processes and techniques for acquiring information about objects contained in images, such as object detection, person detection, face detection, face recognition, and face recognition. Face recognition and face recognition are referred to as face recognition without distinction.
  • the detection processing unit 22 is a processing unit that performs processing including image recognition using DNN processing, and the type of image recognition is not limited to specific types such as human detection, face detection, or face authentication. However, in this embodiment, in order to simplify the explanation, image recognition (DNN processing) in the detection processing unit 22 is basically human detection.
  • the inference model data used for the DNN processing of the detection processing unit 22 includes program data for performing calculations according to the inference model structure (DNN structure) and various parameters (learning coefficients) predetermined by learning. data.
  • Data of the inference model is written in the ROM 23 at the time of manufacture, and the detection processing unit 22 reads the data of the inference model from the ROM 23 and performs DNN processing.
  • the detection processing unit 22 reads data of the inference model from the RAM 24 and performs DNN processing.
  • the changed inference model data is supplied from the AP 12 to the RAM 24 and stored therein.
  • FIG. 1 shows the process of generating an inference model that the detection processing unit 22 uses in DNN processing.
  • the data D1 of the inference model consisting of the program data that determines the structure of the inference model itself and the initial learning coefficients corresponds to the detection target (human, animal, etc.) in the inference model. Learning is performed using the learned data D2.
  • inference model data D3 in which the inference model is learned and the learning coefficient is updated is generated.
  • the data D3 of the learned inference model are evaluated using the evaluation data D4 different from the learning data D2.
  • FIG. 2 is a block diagram illustrating the configuration of the CIS 11. As shown in FIG.
  • the CIS 11 includes a light receiving unit 31, a pixel readout unit 32, a drive timing control unit 33, an image preprocessing unit 34, a CPU (Central Processing Unit) 35, a sensing processing unit 36, a sensing image signal processing unit 37, a viewing It has an image signal processing unit 38, a register control unit 39, an external IO control unit 40, a GPO 41, a power management unit 42, and an image output interface 43.
  • a light receiving unit 31 a pixel readout unit 32, a drive timing control unit 33, an image preprocessing unit 34, a CPU (Central Processing Unit) 35, a sensing processing unit 36, a sensing image signal processing unit 37, a viewing It has an image signal processing unit 38, a register control unit 39, an external IO control unit 40, a GPO 41, a power management unit 42, and an image output interface 43.
  • a CPU Central Processing Unit
  • the light-receiving unit 31 is composed of a large number of light-receiving elements (pixels) that are regularly arranged at high density, and photoelectrically converts an image formed by an optical system (lens) (not shown) pixel by pixel and accumulates it as an electric charge. .
  • the pixel reading unit 32 reads a voltage value (pixel value) corresponding to the amount of charge accumulated in each pixel of the light receiving unit 31 and supplies it to the image preprocessing unit 34 .
  • the drive timing control unit 33 supplies a control signal according to the drive mode (image readout size, shooting mode, etc.) to the pixel readout unit 32 and instructs the pixel readout unit 32 to read out pixel values from the light receiving unit 31 .
  • the image preprocessing unit 34 performs preprocessing such as rearrangement or pixel addition on the pixel values from the pixel readout unit 32 according to subsequent signal processing (for sensing/viewing, etc.).
  • the image preprocessing unit 34 supplies the preprocessed image data to the CPU 35 , the sensing processing unit 36 , the sensing image signal processing unit 37 , or the viewing image signal processing unit 38 .
  • a CPU (Central Processing Unit) 35 operates by firmware, and performs various processes such as processing as an exposure control section 61, a System section 62, and a Sensing post-processing section 63, for example.
  • the exposure control unit 61 calculates the brightness value of the image (captured image) obtained by the light receiving unit 31 by ALS (Ambient Light Sensor) calculation based on the image data from the image preprocessing unit 34 . Based on the result, the exposure control unit 61 controls exposure time, gain, etc. by AE control so that an image with appropriate brightness is captured by the light receiving unit 31 .
  • ALS Ambient Light Sensor
  • the System section 62 controls the overall operation of the CIS 11.
  • the System unit 62 performs mode transition control such as automatic transition of the Sensing function and transition control between the Sensing mode and the Viewing mode.
  • the post-sensing processing unit 63 suppresses erroneous detection by performing stabilization processing and threshold determination processing on the detection results (movement detection results, human detection confidence, etc.) supplied from the sensing processing unit 36. , supplies the stabilized detection result to the register control unit 39 and the external IO control unit 40 .
  • the stabilization process suppresses fluctuations in the detection result from the sensing processing unit 36 by filtering.
  • threshold determination processing threshold determination is performed on the stabilized detection result from the sensing processing unit 36 . Based on this threshold determination, the determination result when it is determined that motion has been detected within the image frame and the determination result when it is determined that a person has been detected within the image frame have high reliability.
  • the Sensing post-processing unit 63 determines that a motion has been detected within the image frame by threshold determination, or determines that a person has been detected within the image frame, the sensing post-processing unit 63 sets the detection flag to the register control unit 39 or external IO control. 40. Note that the Sensing post-processing unit 63 supplies all detection results such as the position in the image frame where the person is detected in addition to the motion detection result and the human detection confidence from the Sensing processing unit 36 to the register control unit 39. .
  • the sensing processing unit 36 has a DNN processing unit 81 and a motion detection unit 82.
  • the DNN processing unit 81 performs DNN processing on the image data from the image preprocessing unit 34 using the DNN processing core based on the above-described inference model to detect a person.
  • the DNN processing unit 81 detects the degree of certainty that the object detected as a person is a person (human detection certainty), the range (position) of the area where the person exists in the image frame, etc. 63.
  • the object to be detected by DNN processing is a person, but the object to be detected is not limited to a person.
  • the detection result of DNN processing also differs depending on the inference model that performs DNN processing.
  • the motion detection unit 82 performs motion detection on the image data from the image preprocessing unit 34, and detects the presence of motion in any part of the image frame, the range of motion (motion range), and the detected motion. Detects the amount of motion that indicates the size of The motion detection unit 82 supplies the detection result (motion detection result) to the sensing post-processing unit 63 . Note that the content of the motion detection result of motion detection by the motion detection unit 82 is not limited to the content described here.
  • the motion detection method of the motion detection unit 82 includes, for example, a method of detecting a change in an image at the same position in the image frame as a motion by comparing the captured image of the previous frame and the captured image of the subsequent frame, or a change in the imaging method. Any method can be used, such as a method of detecting an object that is actually moving in consideration of .
  • the image signal processing unit for sensing 37 performs signal processing for generating an image suitable for sensing system processing (image recognition such as face detection and face authentication) in the AP 12 on the image data from the image preprocessing unit 34. I do.
  • the Sensing image signal processing unit 37 supplies the processed image data to the image output interface 43 .
  • the viewing image signal processing unit 38 performs signal processing for generating a viewing image on the image data from the image preprocessing unit 34 .
  • the viewing image signal processing unit 38 supplies the processed image data to the image output interface 43 .
  • the register control unit 39 holds the parameters necessary for controlling the CIS 11 and the calculation results of each unit such as the detection results of the sensing processing unit 36.
  • the data held in the register control unit 39 can be appropriately referenced from the AP 12 via I2C (Inter-Integrated Circuit), which is a communication interface standard.
  • the external IO control unit 40 transmits the detection results (movement and existence of people) from the Sensing post-processing unit 63 to the AP 12 via the GPO 41, which is a general-purpose output port. For example, when the external IO control unit 40 receives a detection flag indicating that a motion is detected within the image frame or a detection flag indicating that a person is detected within the image frame from the Sensing post-processing unit 63, the detection flag indicates that effect. to the AP 12.
  • the detection trigger may be sent to the AP 12 only in response to the detection flag when a person is detected within the image frame.
  • the power management unit 42 controls the power control of the entire CIS 11 and the ON/OFF of each unit according to the drive mode of the CIS 11 .
  • the image output interface 43 transmits the image data from the Sensing image signal processing unit 37 or the Viewing image signal processing unit 38 to the AP 12 by controlling MIPI, which is an interface standard.
  • the CIS 11 when the CIS 11 captures an image with the light receiving unit 31, it performs single exposure imaging (SDR: Standard Dynamic Range) and multiple exposure imaging (High Dynamic Range) capable of acquiring an image with a higher gradation than single exposure imaging. be able to.
  • SDR Standard Dynamic Range
  • High Dynamic Range High Dynamic Range
  • the image supplied to the viewing image signal processing unit 38 is Since the image is an image for generating an image for viewing, it is a high-resolution image and is also called a high-resolution image for viewing.
  • the image supplied to the image signal processing unit 37 for sensing is an image for generating an image suitable for sensing system processing (face detection, face recognition, etc.) in the AP 12, so it has a lower resolution than the high-resolution image for viewing. It is a high-resolution image, and is also called a low-resolution image for sensing. As the viewing high-resolution image and the sensing low-resolution image, it is possible to select either an image obtained by single-exposure imaging or an image obtained by multiple-exposure imaging.
  • the image supplied to the sensing processing unit 36 (referred to as an image for detection processing) is an image used for human detection where high-speed detection is desired with low processing load and low power consumption in the chip of the CIS 11. .
  • the detection processing image has a lower resolution than at least the viewing high-resolution image.
  • the image for detection processing may have a similar resolution or a lower resolution.
  • the image for detection processing is an image acquired by multiple exposure imaging so as to obtain a highly accurate result in DNN processing.
  • the image for detection processing may be an image obtained by single-exposure imaging.
  • the AP 12 has a CPU 101 and an ISP (Image Signal Processor) 102.
  • the CPU 101 performs various processes according to programs.
  • the CPU 101 receives a detection trigger from the external IO control unit 40 via the GPO 40 when the Sensing processing unit 36 in the CIS 11 detects movement within the image frame or detects a person within the image frame.
  • the CPU 101 can perform processing triggered by a detection trigger.
  • the CPU 101 can acquire data held in the register control unit 39 through I2C (Inter Integrated Circuits), which is a communication interface standard. Since the detection result detected by the sensing processing unit 36 is held in the register control unit 39, the CPU 101 can acquire the detection result detected by the sensing processing unit 36 via the register control unit 39.
  • the acquired detection result can also be used in arbitrary processing.
  • the ISP 102 mainly performs signal processing such as image adjustment.
  • the ISP 102 acquires the sensing low-resolution image processed by the sensing image signal processing unit 37 of the CIS 11 and the viewing high-resolution image processed by the viewing image signal processing unit 38 via the image output interface 43. .
  • the ISP 102 performs signal processing such as image adjustment on the acquired low-resolution image for sensing and high-resolution image for viewing, and then supplies them to the CPU 101 . Transmission of the low-resolution image for sensing from the image signal processing unit 37 for sensing to the ISP 102 and transmission of the high-resolution image for viewing from the image signal processing unit 38 for viewing to the ISP 102 are performed exclusively. Therefore, ISP 102 does not acquire those images at the same time.
  • the CPU 101 can use the sensing low-resolution image or the viewing high-resolution image from the ISP 102 for any purpose.
  • the CPU 101 may perform image recognition such as face detection by DNN processing, and unlock the terminal.
  • the CPU 101 acquires a low-resolution image for sensing from the ISP 102 and performs DNN processing on the low-resolution image for sensing.
  • the DNN processing in the AP 12 can be any of object detection, human detection, face detection, face recognition, etc., but for the sake of simplification of explanation, face detection is mainly performed.
  • the terminal device 1 can use the image recognition (DNN processing) of the CIS 11 and the image recognition of the AP 12 .
  • the terminal device 1 can use the image recognition of the CIS 11 and the image recognition of the AP 12 in the following first to fourth usage patterns. Note that the motion detection performed by the motion detection unit 82 of the sensing processing unit 36 is included in the image recognition technology of the CIS 11 .
  • FIG. 3 is a diagram illustrating a first usage pattern of image recognition by the CIS 11 and image recognition by the AP 12.
  • the CIS 11 has a preprocessing section 21 , a detection processing section 22 and a postprocessing section 25 .
  • the pre-processing unit 21 and the detection processing unit 22 correspond to the pre-processing unit 21 and the detection processing unit 22 denoted by the same reference numerals in the CIS 11 of FIG. 1, so description thereof will be omitted.
  • the post-processing unit 25 in FIG. 3 is omitted in FIG. 1, and in FIG. That is, the post-processing unit 25 includes a processing unit that processes, stores, and transmits to the AP 12 the detection result of image recognition in the detection processing unit 22 .
  • the AP 12 has an ISP 102, a detection determination processing unit 103, a DRAM (dynamic random access memory) 104, and a storage 105.
  • the ISP 102 corresponds to the ISP 102 denoted by the same reference numerals in the AP 12 of FIG. 2, so the explanation is omitted.
  • the detection determination processing unit 103 performs processing including image recognition among the processing executed by the CPU 101 in the AP 12 of FIG. 2, for example.
  • DRAM 104 temporarily stores images processed by ISP 102 in AP 12 .
  • a storage 105 stores various data in the AR 12 .
  • the CIS 11 is in an active state (operating state) and the AP 12 is in a low power consumption standby state (waiting state). In this state, the detection processing unit 22 of the CIS 11 performs image recognition on the image supplied from the preprocessing unit 21 .
  • the detection determination processing unit 103 of the AP 12 does not perform image recognition. If the detection determination processing unit 103 performs image recognition, the preprocessing unit 21 of the CIS 11 transmits an image for face detection (low-resolution image for sensing) to the ISP 102 of the AP 12, and the detection determination processing unit 103 of the AP 12 acquires an image for face detection via the ISP 102 . Since the detection determination processing unit 103 of the AP 12 does not perform image recognition, the face detection image is not transmitted from the CIS 11 to the AP 12 either.
  • the post-processing unit 25 of the CIS 11 determines that a person has been detected within the image frame based on the detection result of image recognition supplied from the detection processing unit 22, it transmits a detection trigger to the AP 12.
  • the AP 12 receives the detection trigger, it switches from the standby state to the active state and starts predetermined processing.
  • the AP 12 can acquire the detection result of the image recognition of the detection processing unit 22 from the post-processing unit 25 of the CIS 11 (information such as human detection certainty and subject position) as necessary.
  • the AP 12 According to the first usage pattern, it is effective in a situation where the AP 12 can be in the standby state when no person is detected within the image frame.
  • the AP 12 does not waste power, and the power consumption required for transmitting images from the CIS 11 to the AP 12 is also eliminated. Since the power consumption required for image recognition by the CIS 11 is also small, the power consumption is less than when the AP 12 performs image recognition.
  • FIG. 4 is a diagram illustrating a second usage pattern of image recognition by the CIS 11 and image recognition by the AP 12.
  • FIG. 4 parts corresponding to those of the CIS 11 and AP 12 in FIG.
  • both CIS 11 and AP 12 are in active state (operating state), AP 12 .
  • the detection processing unit 22 of the CIS 11 does not perform image recognition.
  • a detection determination processing unit 103 of the AP 12 performs image recognition.
  • the preprocessing unit 21 of the CIS 11 transmits the image for face detection (low-resolution image for sensing) to the ISP 102 of the AP 12, and the detection determination processing unit 103 of the AP 12 acquires the image for face detection via the ISP 102.
  • the second usage pattern is effective when highly accurate image recognition is required.
  • the face detection image transmitted from the CIS 11 to the AP 12 can be a Sensing low-resolution image having a lower resolution than the Viewer high-resolution image used for viewing. Therefore, the amount of data required for transmitting the face detection image from the CIS 11 to the AP 12 is small, and power consumption can be reduced.
  • the second usage pattern can be applied in combination with the first usage pattern. In that case, power consumption can be reduced, which is more effective. Specifically, when no person is detected within the image frame, the CIS 11 performs person detection by image recognition as in the first usage pattern. When the CIS 11 detects a person by image recognition, a detection trigger is sent to the AP 12 .
  • the AP 12 When the AP 12 receives the detection trigger, the AP 12 performs image recognition as in the second usage pattern. In this case, the AP 12 does not need to transmit a face detection image from the CIS 11 to the AP 12 at least while no person is detected in the image frame, and the AP 12 does not need to perform image recognition. Therefore, the power consumption required for image transmission and the power consumption required for image recognition by the AP 12 are reduced.
  • FIG. 5 is a diagram for explaining a third usage pattern of the image recognition of the CIS 11 and the image recognition of the AP 12.
  • FIG. 5 parts corresponding to those of the CIS 11 and AP 12 in FIG.
  • a third usage pattern represents a case where the terminal device 1 has two CISs 11 and 121 .
  • the CIS 121 does not have the function of image recognition.
  • the CIS 121 has a pre-processing section 141 , an ISP 142 and a post-processing section 143 .
  • the pre-processing unit 141 corresponds to the pre-processing unit 21 of the CIS 11, so description thereof will be omitted.
  • the ISP 142 acquires, for example, an image for face detection (low-resolution image for sensing) from the preprocessing unit 141 and generates an image suitable for sensing processing (face detection, face authentication, etc.) in the AP 12 by signal processing.
  • the post-processing unit 143 supplies the image for face detection to the ISP 102 of the AP 12 .
  • the AP 12 can perform image recognition on both the face detection image acquired from the CIS 11 and the face detection image acquired from the CIS 121 .
  • CIS11, CIS121, and AP12 are all active.
  • the detection processing unit 22 of the CIS 11 performs image recognition on the face detection image from the preprocessing unit 21
  • the detection determination processing unit 103 of the AP 12 performs image recognition on the face detection image from the CIS 121 .
  • image recognition can be distributed according to the detection target.
  • the CIS 11 can perform human detection by image recognition
  • the AP 12 can perform face detection and the like by image recognition.
  • FIG. 6 is a diagram illustrating a fourth usage pattern of image recognition by the CIS 11 and image recognition by the AP 12. In FIG. In the figure, parts corresponding to those of the CIS 11 and AP 12 in FIG.
  • the CIS 11 has a preprocessing unit 21, a detection processing unit 22, a postprocessing unit 25, and an encoder 26.
  • the encoder 26 encodes the face detection image from the preprocessing unit 21 to reduce the amount of data.
  • the encoder 26 transmits the encoded image for face detection to the AP 12 .
  • the encoder 26 transmits the face detection image to the AP 21 using, for example, MIPI, I2C, or 13C (Improved InterIntegrated Circuits), which are communication interface standards.
  • the AP 12 has an ISP 102, a detection determination processing unit 103, a DRAM 104, a storage 105, and a decoder 106.
  • the decoder 106 receives and decodes the face detection image transmitted from the encoder 26 of the CIS 11 .
  • the decoder 106 supplies the decoded image for face detection to the ISP 102 .
  • the detection determination processing unit 103 acquires from the ISP 102 the image for face detection supplied from the decoder 106 to the IPS 102 and performs image recognition.
  • CIS 11 and AP 12 are active.
  • the preprocessing unit 21 of the CIS 11 supplies the face detection image to the encoder 26 .
  • the encoder 26 encodes the face detection image from the preprocessing unit 21 and transmits the encoded image to the decoder 106 of the AP 12 .
  • the decoder 106 decodes the face detection image from the encoder 26 and supplies it to the ISP 102 .
  • a detection determination processing unit 103 acquires an image for face detection via the ISP 102 and performs image recognition.
  • both CIS 11 and AP 12 are active (operating state), AP 12 .
  • the detection processing unit 22 of the CIS 11 does not perform image recognition.
  • a detection determination processing unit 103 of the AP 12 performs DNN processing.
  • the preprocessing unit 21 of the CIS 11 transmits the low-resolution image for sensing (image for face detection) to the ISP 102 of the AP 12, and the detection determination processing unit 103 of the AP 12 acquires it via the ISP 102.
  • the fourth usage pattern even when the AP 12 performs image recognition, the image data transmitted from the CIS 11 to the AP 12 is encoded, so the amount of data transmitted is reduced. Therefore, the power consumption required for transmitting images from the CIS 11 to the AP 12 is reduced.
  • image recognition (also called sensing) of the CIS 11 includes both DNN processing by the DNN processing unit 81 and motion detection by the motion detection unit 82 in the sensing processing unit 36 of the CIS 11 in FIG.
  • Application example 1 is an application example in which the image recognition of the CIS 11 is applied to the terminal unlock (Face unlock) and terminal lock (Face lock) functions of the terminal device 1 .
  • the terminal unlock of the terminal device 1 is a function (also called face unlock) of switching the locked state of the terminal device 1 to the unlocked state when the user's face is authenticated when the terminal device 1 is in the locked state.
  • the locked state of the terminal device 1 includes a general state in which operations of the terminal device 1 are prohibited except for specific operations, and an unlock state for returning the terminal device 1 to its original operating state or a predetermined operating state. Contains all states that require processing.
  • a typical operation state included in the lock state of the terminal device 1 is a standby state (a state in which processing is stopped and a state activated by a trigger associated with unlock processing), which is also one of the power saving states. ), and a power saving state in which the display is turned off.
  • the terminal device 1 is, for example, a smart phone, and the camera including the CIS 11 takes an image of a range facing the display of the smart phone.
  • FIG. 7 is a diagram for explaining the image recognition performed by the CIS 11 and the AP 12 in contrast to the image recognition performed when the terminal device 1 is unlocked.
  • a captured image 161 is an example of a captured image captured by the CIS 11, and represents the subject within the image frame.
  • a user 181 represents the owner of the terminal device 1, for example.
  • Moving object 182 represents any moving object.
  • a motion detection frame 201 represents a range in which motion is detected within the image frame by motion detection when it is assumed that the user 181 and the moving body 182 move simultaneously. In the motion detection, it is not necessary to detect the range of the motion detection frame 201 within the image frame, and it is a case where motion (the presence of motion) is detected within any range within the image frame. may This motion detection is performed by image recognition of the CIS 11 .
  • a human detection frame 202 represents a range in which a person is detected within the image frame by human detection.
  • human detection for example, when a plurality of people exist within the image frame, the respective ranges of the plurality of people (range of the human detection frame 202) are detected.
  • range of the user 181 is detected.
  • whether or not the detected person is the user 181 is not specified. This person detection is performed by image recognition of the CIS 11 .
  • a face detection frame 203 represents the range in which a face is detected within the image frame by face detection.
  • face detection for example, when a plurality of faces exist within an image frame, the respective ranges of the plurality of faces (ranges of the face detection frame 203) are detected.
  • the range of the face of the user 181 is detected. However, whether or not the detected face is the user 181 is not specified.
  • This face detection is performed by image recognition of the AP 12 .
  • Image recognition by the AP 12 is performed by DNN processing in the detection determination processing unit 103 in FIG. 3, for example.
  • terminal unlocking if the terminal device 1 is unlocked only when the face of the user 181 is detected instead of an arbitrary face, the face detected by the face detection is prepared in advance. Verification (face authentication) with information on the face of the user 181 is performed.
  • motion detection, human detection, and face detection are performed in cooperation with each other.
  • the process transitions from human detection to face detection.
  • the person detection transitions to the motion detection, and the motion detection is restarted.
  • face authentication is performed. If face detection does not detect a face within a certain period of time, the face detection transitions to motion detection, and the motion detection is restarted.
  • the terminal device 1 transitions from the locked state to the unlocked state, and image recognition ends.
  • the face detection transitions to motion detection, and the motion detection is restarted.
  • the processing when face authentication fails is not limited to this.
  • Coordination of motion detection, human detection, and face detection is not limited to temporal coordination.
  • the human detection may be performed by limiting the range of the motion detection frame 201 .
  • face detection may be performed within the range of the human detection frame 202 .
  • the AP 12 sets the terminal device 1 in the locked state or the unlocked state.
  • FIG. 8 shows an example (example 1) of a specific situation (scene development) until the terminal device 1 switches from the locked state to the unlocked state by the terminal unlock function of the terminal device 1, and the states of the CIS 11 and AP 12 at that time. It is a figure explaining the outline
  • scene C1 is a scene where the user puts the terminal device 1 (for example, a smartphone) in the standby state in his pocket and walks.
  • the terminal device 1 for example, a smartphone
  • the display is turned off and only specific operations such as the power button are accepted.
  • Scene C2 is a scene in which the user takes out the terminal device 1 in the standby state from the pocket in order to use the terminal device 1 .
  • Scene C3 is a scene in which the user faces the terminal device 1 (camera) in order to unlock the terminal device 1 using the terminal unlock function. Note that the camera of the terminal device 1 may be installed in a direction in which the user's face is captured when the terminal device 1 is used. Scene C3 in this case includes a scene in which the camera is directed at the face of the user as a result of the user intending to use the terminal device 1 without being conscious of unlocking the terminal device 1 .
  • Scene C4 is a scene in which the user's face authentication in the terminal unlock function is successful, and the terminal device 1 switches from the standby state (locked state) to the activated state (unlocked state) that allows normal use.
  • the state of AP12 is "Waiting low power event". "Waiting low power event” indicates that the AP 12 is in sleep mode. In sleep mode, the AP 12 does not perform processing and waits for a detection trigger (human detection notification signal) from the CIS 11 . In sleep mode, AP 12 consumes very little power.
  • the state of the CIS 11 is "AO MD” or "AO HE” for scenes C1 and C2. Both "AO MD” and “AO HE” indicate that the CIS 11 is in sensing mode. In the sensing mode, the CIS 11 repeats imaging by at least the light receiving section. There are a plurality of types of sensing modes, and in the “AO MD” and “AO HE” sensing modes, the CIS 11 captures low-resolution images for detection processing.
  • AO MD MO: Motion Detection
  • AO HE represents a human detection mode that performs human detection by DNN processing using images for detection processing.
  • the CIS 11 detects motion within the image frame in the motion detection mode ("AO MD”) and detects the person. Automatically transition to detection mode (“AOHE"). In the scenes C1 and C2, no person is detected within the image frame (assuming that), so the CIS 11 automatically switches to the motion detection mode ("AO HE") after a certain period of time has passed in the human detection mode (“AO HE”). MD”). In scenes C1 and C2, the CIS 11 repeats such state transitions.
  • the state of CIS 11 transitions from the human detection mode ("AO HE") to "Sensing (VGA)".
  • the CIS 11 transmits "ON” to the AP 12.
  • "ON” is a detection trigger (human detection notification signal) transmitted from the CIS 11 to the AP 12 when the CIS 11 detects a person in the image frame in the human detection mode ("AO HE").
  • the detection trigger is transmitted to the AP 12 by the Sensing post-processing unit 63 in FIG.
  • “Sensing (VGA)” is a kind of sensing mode, and is a mode for transmitting a low-resolution image for image recognition to the AP 12 .
  • the CIS 11 In the sensing mode (“Sensing (VGA)”), the CIS 11 continuously captures (acquires) low-resolution images for Sensing and transmits the acquired low-resolution images for Sensing to the AP 12 .
  • the Sensing low-resolution image is transmitted from the Sensing image signal processing unit 37 in FIG. 2 to the AP 12 .
  • the sensing mode (“Sensing (VGA)”) is a mode intended to output low-resolution Sensing images (low-resolution images for Sensing) used for image recognition outside CIS11. , Sensing mode for Sensing image output, or Sensing image output mode. In the sensing mode for Sensing image output, the CIS 11 can also perform motion detection and human detection at the same time as outputting the Sensing image. shall not be performed.
  • the CIS 11 detects a person within the image frame in the human detection mode ("AO HE"). As a result, the CIS 11 transmits a detection trigger (“ON”) to the AP 12 to notify the fact to the AP 12, and switches from the human detection mode (“AO HE”) to the sensing mode for Sensing image output (“Sensing (VGA)”). ).
  • AO HE human detection mode
  • VGA Sensing image output
  • the state of AP 12 transitions from sleep mode ("Waiting low power event") to "Face recog.” when a detection trigger ("ON") is received from CIS 11.
  • "Face recog.” represents a face recognition mode in which the AP 12 performs face recognition.
  • the AP 12 performs face detection by image recognition (DNN processing, etc.) prior to authentication.
  • the AP 12 acquires a low-resolution image for sensing from the CIS 11 and performs DNN processing and the like on the low-resolution image for sensing.
  • the state of AP 12 transitions from the face authentication mode ("Face recog.") to "Phone unlocked”. During this transition, AP 12 transmits “Comp” to CIS 11 .
  • “Comp” is a notification signal instructing the CIS 11 to stop the sensing mode (“Sensing (VGA)”) for Sensing image output.
  • “Phone unlocked” represents a state in which the AP 12 performs processing for setting the terminal device 1 to an unlocked state.
  • the AP 12 terminates face authentication when transitioning to "Phone unlocked”.
  • the user's face is included in the image frame, so the AP 12 detects the face within the image frame by face detection in the face authentication mode (“Face recog.”).
  • the AP 12 When the AP 12 detects a face, the AP 12 compares the detected face with information on the user's face (user's facial image or facial features) obtained in advance. Since the face in the image frame is the user's face, the AP 12 determines that the face recognition has succeeded as a result of matching, and transitions from the face recognition mode ("Face recog.") to "Phone unlocked.” During this transition, AP 12 transmits a communication signal (“Comp”) to CIS 11 .
  • Face recog. face recognition mode
  • the state of the CIS 11 transitions from the sensing mode ("Sensing (VGA)") to "OFF ⁇ Controlled by AP” when the notification signal (“Comp") from the AP 12 is received.
  • "OFF ⁇ Controlled by AP” is a state in which the processing of the CIS 11 is stopped, and represents a standby state in which the AP 12 waits to receive a notification signal regarding control or the like.
  • the CIS 11 receives the notification signal (“Comp”) from the AP 12 .
  • the CIS 11 stops the sensing mode (“Sensing (VGA)”) for sensing image output, stops capturing the low-resolution image for sensing, and stops transmitting the low-resolution image for sensing to the CIS 11 .
  • the sensing mode (VGA))
  • the CIS 11 transitions to the standby mode (“OFF ⁇ Controlled by AP”) and switches to a state of waiting for a notification signal from the AP 12 .
  • FIG. 9 shows a specific situation (scene development) in which the terminal device 1 is switched from the unlocked state to the locked state by the terminal lock and terminal unlock (face unlock) functions of the terminal device 1, and then switched from the locked state to the unlocked state. ) and an overview of state transitions of the CIS 11 and the AP 12 at that time (example 2).
  • scene C1 is a scene in which the user is cooking while referring to a recipe displayed on the display of terminal device 1 in the unlocked state. It is assumed that the user continues to perform no operation on the terminal device 1 .
  • Scene C2 is a scene in which the terminal device 1 is locked by the terminal lock function after a specified period of time has elapsed since the user did not perform any operation on the terminal device 1 .
  • the locked state of the terminal device 1 is, for example, a power saving state (a state in which power consumption is reduced compared to the normal activation state) and a state in which the display is turned off.
  • Scene C3 is a scene in which the user faces the display of the terminal device 1 to see the recipe. At this time, the user's face can be confirmed within the image frame.
  • Scene C4 is a scene in which the user's face authentication in the terminal unlock function is successful, and the terminal device 1 returns from the locked state to the unlocked state in which the display before the locked state is turned on.
  • the state of AP12 is "Phone unlocked”. "Phone unlocked” indicates that the AP 12 has set the terminal device 1 to an unlocked state.
  • the state of CIS 11 is standby mode (“OFF ⁇ Controlled by AP”).
  • the state of AP 12 transitions from "Phone unlocked” to sleep mode ("Waiting low power event"). During this transition, the AP 12 transmits to the CIS 11 a notification signal requesting activation of the sensing mode (sensing mode based on motion detection and human detection).
  • the sensing mode sensing mode based on motion detection and human detection.
  • the terminal device 1 is locked as a terminal lock function. At this time, the AP 12 transitions from the "Phone unlocked” state to the sleep mode ("Waiting low power event").
  • the state of the CIS 11 changes from the standby mode (“OFF ⁇ Controlled by AP”) to the sensing mode, motion detection mode (“AO MD”).
  • the CIS 11 captures a low-resolution image for detection processing and performs motion detection using the image for detection processing.
  • the CIS 11 does not detect motion within the picture frame and continues the motion detection mode ("AO MD").
  • the state of the CIS 11 transitions from the motion detection mode ("AO MD”) to the human detection mode (“AO HE").
  • the user has moved his/her face toward the terminal device 1, so the CIS 11 detects the movement within the image frame.
  • the CIS 11 automatically transitions from the motion detection mode (“AO MD”) to the human detection mode (“AO HE”) as the sensing mode.
  • the human detection mode (“AO HE”) the CIS 11 captures a low-resolution image for detection processing, and uses the image for detection processing to perform human detection.
  • the state of AP 12 continues in sleep mode ("Waiting low power event").
  • the state of the CIS 11 transitions from the human detection mode ("AO HE") to the sensing mode (“Sensing (VGA)”). During this transition, the CIS 11 transmits a detection trigger (“ON”) (human detection notification signal) to the AP 12 . After that, when the CIS 11 receives a notification signal (“Comp”) from the AP 12, it transitions from the sensing mode (“Sensing (VGA)”) to the standby mode (“OFF ⁇ Controlled by AP”).
  • AO HE human detection mode
  • VGA sensing
  • OFF Controlled by AP
  • AP 12 For scene C4, when the detection trigger ("ON") from CIS 11 is received, the state of AP 12 is activated from sleep mode ("Waiting low power event") (transition to active mode), face authentication mode ( "Face recog.”). After that, the state of the AP 12 transitions from the face authentication mode (“Face recog.”) to “Phone unlocked” when the face authentication succeeds (assuming that it succeeds). During this transition, AP 12 transmits a notification signal (“Comp”) to CIS 11 .
  • a notification signal (“Comp”)
  • the CIS 11 detects a person within the image frame in the human detection mode ("AO HE"). As a result, the CIS 11 transmits a detection trigger (“ON”) notifying that a person has been detected to the AP 12, and transitions to the sensing mode (“Sensing (VGA)”). In the sensing mode (“Sensing (VGA)”), the CIS 11 does not perform image recognition, continuously captures low-resolution Sensing low-resolution images, and transmits the acquired Sensing low-resolution images to the AP 12 . On the other hand, when the AP 12 receives the detection trigger (“ON”) from the CIS 11, it transitions from the sleep mode (“Waiting low power event”) to the face recognition mode (“Face recog.”).
  • the AP 12 performs face detection by image recognition (DNN processing) prior to authentication.
  • the AP 12 acquires a sensing low-resolution image from the CIS 11 and performs image recognition on the sensing low-resolution image.
  • the AP 12 detects the face within the image frame by face detection in the face authentication mode (“Face recog.”).
  • the AP 12 compares the detected face with information on the user's face (user's facial image or facial features) obtained in advance.
  • the AP 12 determines that the face recognition has succeeded as a result of matching, and transitions from the face recognition mode ("Face recog.") to "Phone unlocked.” During this transition, AP 12 transmits a notification signal (“Comp”) to CIS 11 .
  • the CIS 11 receives the notification signal (“Comp") from the AP 12, it stops the sensing mode ("Sensing (VGA)"), captures a low-resolution image for sensing, and transmits the low-resolution image for sensing to the CIS 11. Stop sending. After stopping the sensing mode (“Sensing (VGA)”), the CIS 11 transitions to the standby mode (“OFF ⁇ Controlled by AP”) and switches to a state of waiting for a notification signal regarding control or the like from the AP 12 .
  • FIG. 10 is a flowchart showing state transitions of the terminal device 1 and state transitions of the CIS 11 and AP 12 when image recognition of the CIS 11 and AP 12 is used for the terminal unlock function.
  • states C1 to C7 represent state transitions of the terminal device 1.
  • the terminal device 1 is executing an application.
  • the state of the AP 12 is the state in which the startup setting of the CIS 11 is being performed, and the state of the CIS 11 is the standby mode ("HW Standby").
  • the standby mode (“HW Standby”) is a state in which power is not supplied to the CIS 11 and the CIS 11 stops processing.
  • the terminal device 1 transitions from state C1 to state C2.
  • state C2 the terminal device 1 is in a standby state (locked state).
  • the transition from state C1 to state C2 is not limited to specific conditions. For example, a human face may not be detected within the image frame for a predetermined period of time, or the terminal device 1 may not be operated for a predetermined period of time.
  • AP 11 is in sleep mode ("AP Sleep”). In the sleep mode (“AP Sleep”), power is not supplied to the AP 12, and the AP 12 is in a standby state without doing anything.
  • the AP 12 transitions to the sleep mode (“AP Sleep”), the AP 12 transmits to the CIS 11 a notification signal (“Turn sensing mode”) requesting activation of the sensing mode based on motion detection and human detection.
  • CIS 11 transitions from standby mode (“HW Standby”) in which power is not supplied to standby mode (“SW Standby”) in which power is supplied by the notification signal (“Turn Sensing mode”) from AP 11, and AP 11 Predetermined startup settings are performed.
  • the terminal device 1 transitions from state C2 to state C3.
  • the terminal device 1 performs motion detection.
  • the CIS 11 transitions from the standby mode (“SW Standby”) in state C2 to the motion detection mode (“Sensing Mode (Motion Detection)”) in the sensing mode based on motion detection and human detection.
  • the motion detection mode (“Sensing Mode (Motion Detection)”) indicates the same mode as the motion detection mode (“AO MD”) described in FIG.
  • the terminal device 1 enters state C3 in which motion detection is performed by image recognition by the CIS 11 .
  • the AP 12 is in sleep mode ("AP Sleep”).
  • the terminal device 1 transitions from state C3 to state C4.
  • state C4 the terminal device 1 performs human detection.
  • the CIS 11 switches from the motion detection mode ("Sensing Mode (Motion Detection)”) to the human detection mode ( "Sensing Mode (Human Existence)”) automatically.
  • the human detection mode (“Sensing Mode (Human Existence)”) indicates the same mode as the human detection mode (“AO HE”) described in FIG.
  • the terminal device 1 enters the state C4 in which the CIS 11 performs image recognition to detect a person.
  • the AP 12 continues in sleep mode ("AP Sleep”). The terminal device 1 transitions from state C4 to state C5.
  • the terminal device 1 performs face detection.
  • the human detection mode (“Sensing Mode (Human Existence)” in state C4
  • the CIS 11 triggers detection (“Wake up AP”) to the AP 12 (human detection notification signal ).
  • the detection trigger (“Wake up AP”) is the same signal as the detection trigger (“ON”) described in FIG.
  • the AP 12 wakes up from sleep mode (“AP Sleep”) (transitions to active mode) and transitions to face detection mode (“Face detection”).
  • the AP 12 When the AP 12 transitions to the face detection mode (“Face detection”), the AP 12 transmits a notification signal (“(Request sensing image”) requesting transmission of a low-resolution Sensing image to the CIS 11.
  • the CIS 11 receives from the AP 12
  • a notification signal (“(Request sensing image”)
  • the human detection mode (“Sensing Mode (Human Existence)” transitions to the sensing mode for Sensing image output (“Sensing Mode (Sensing image streaming)”.
  • the sensing mode (“Sensing Mode (Sensing image streaming)” indicates the same mode as the sensing mode (“Sensing (VGA)”) described in FIG. 8.
  • the CIS 11 transmits a low-resolution low-resolution image for sensing to the AP 12.
  • the AP 12 receives the low-resolution image for sensing from the CIS 11, performs image recognition on the low-resolution image for sensing, and recognizes the face.
  • the terminal device 1 enters state C5 in which face detection is performed by image recognition of the AP 12.
  • the terminal device 1 transitions from state C5 to state C6.
  • the terminal device 1 performs face authentication.
  • the AP 12 detects a human face in the image frame in the face detection mode (“Face detection”) in state C5, it transitions to the face authentication mode (“Face identification”).
  • the AP 12 performs face authentication by matching the face detected by face detection with the user's face information obtained in advance.
  • the terminal device 1 enters state C6 in which face authentication is performed by image recognition of the AP 12 .
  • the terminal device 1 transitions from state C6 to state C7.
  • state C7 the terminal device 1 is in an unlocked state.
  • the AP 12 transitions to "Unlock” when the face authentication is successful in the face authentication mode (“Face identification”) in state C6.
  • "Unlock” is the same state as "Phone unlocked” described with reference to FIG.
  • the AP 12 sends a notification signal (“Turn HW Standby”) requesting the CIS 11 to switch to the standby mode (“HW Standby”). to CIS11.
  • the CIS 11 When the CIS 11 receives the notification signal (“Turn HW Standby”) from the AP 12, it transitions from the sensing mode (“Sensing Mode (Sensing image streaming)” to the standby mode (“HW Standby”). becomes state C7 in which the locked state is released by AP12.
  • FIG. 11 is a flowchart showing state transitions of the terminal device 1 and state transitions of the CIS 11 and AP 12 when image recognition of the CIS 11 and AP 12 is used for the terminal lock function.
  • the sensing mode of the CIS 11 includes a sensing mode for detecting a person in the human detection mode when motion is detected in the motion detection mode as in states C3 and C4 in FIG.
  • the former is referred to as a sensing mode based on motion detection and human detection, and the latter is referred to as a sensing mode based only on human detection.
  • Sensing using motion detection and human detection as in the former is called sensing by motion detection and human detection, and sensing using only human detection as in the latter is called sensing by human detection only.
  • the detection result whether or not a person is detected is continuously output to the AP 12 etc. can do.
  • Image recognition (sensing) by the CIS 11 in this case is called constant sensing, and a mode in which the CIS 11 always performs sensing is also called a constant sensing mode.
  • the detection results that are continuously output in constant sensing not only indicate whether or not a person has been detected within the image frame, but also whether or not movement has been detected within the image frame. Varies accordingly.
  • states C1 to C3 represent state transitions of the terminal device 1 .
  • the terminal device 1 is executing an application.
  • the state of the AP 12 is the state in which the startup setting of the CIS 11 is being performed (“CIS startup setting”), and the state of the CIS 11 is the standby mode (“HW Standby”).
  • the terminal device 1 transitions from state C1 to state C2.
  • state C2 the terminal device 1 performs human detection in constant sensing.
  • a transition from the state C1 to the state C2 occurs, for example, when the user does not perform any operation on the terminal device 1 for a predetermined period of time.
  • the AP 12 has transitioned to idle mode ("AP Idle”).
  • AP Idle In the idle mode (“AP Idle”), power is supplied to the AP 12, but the AP 12 is in a standby state without doing anything.
  • the AP 12 transmits to the CIS 11 a notification signal (“Turn sensing mode”) requesting activation of the constant sensing mode based only on human detection.
  • the CIS 11 transitions from a standby mode (“HW Standby”) in which power is not supplied to a standby mode (“SW Standby”) in which power is supplied by a notification signal (“Turn sensing mode”) from the AP 12.
  • the CIS 11 After transitioning to the standby mode (“SW Standby”), the CIS 11 transitions to the human detection mode (“Sensing Mode (Human Existence)”).
  • the human detection mode at this time is a constant sensing mode in which only human detection is performed. As a result, the CIS 11 performs continuous human detection through image recognition.
  • the CIS 11 In the human detection mode (“Sensing Mode (Human Existence)”), when a person is detected in the image frame, the CIS 11 sends a human detection notification signal ("Human Detected") to the AP 12. send to.
  • the human detection mode In the human detection mode (“Sensing Mode (Human Existence)", when the CIS 11 does not detect a person in the image frame, the human undetected notification signal (“Human Not Detected”) to the AP 12.
  • the CIS 11 transmits a human undetected notification signal (“Human Not Detected”) to the AP 12, it transitions to a standby mode (“SW Standby”).
  • the AP 12 When the AP 12 receives the human undetected notification signal (“Human Not Detected”) from the CIS 11, it performs processing to lock the terminal device 1, and enters the “CIS startup setting” state for setting up the startup of the CIS 11. . In this state, the AP 12 transmits to the CIS 11 a notification signal (“Turn sensing mode”) requesting activation of the sensing mode based on motion detection and human detection. The AP 12 then switches to an unpowered sleep mode ("AP Sleep"). As a result, the terminal device 1 transitions to state C3, which is a standby state. In the standby state, for example, the display of the terminal device 1 is turned off.
  • AP Sleep unpowered sleep mode
  • the CIS 11 When the CIS 11 receives a notification signal ("Turn sensing mode") requesting activation of the sensing mode by motion detection and human detection from the AP 12, it switches from the standby mode ("SW Standby”) to sensing by motion detection and human detection. mode. The reason why the CIS 11 shifts to the sensing mode by motion detection and human detection is to unlock the terminal.
  • a notification signal (“Turn sensing mode”) requesting activation of the sensing mode by motion detection and human detection from the AP 12
  • SW Standby standby mode
  • the reason why the CIS 11 shifts to the sensing mode by motion detection and human detection is to unlock the terminal.
  • FIG. 12 is a sequence diagram illustrating the terminal lock and terminal unlock processing procedures.
  • FIG. 12 shows the flow of processing relating to the camera (CIS 11), AP 12, user 181, other object (moving object 182), and user 181.
  • steps S1 to S7 relate to terminal locking
  • steps S9 to S21 relate to terminal unlocking.
  • the AP 12 requests the CIS 11 to enable constant sensing only by human detection (step S1).
  • the CIS 11 transitions to the human detection mode (corresponding to the human detection mode (“Sensing Mode (Human Existence)”) in FIG. 11) in response to the request in step S1 (step S2).
  • the CIS 11 notifies the AP 12 that no person has been detected (that no person has been detected) (step S4).
  • the AP 12 transitions to terminal lock processing upon notification of undetected person in step S4 (step S5).
  • the AP 12 shifts to terminal lock processing it requests the CIS 11 to enable sensing (constant sensing) based on motion detection and human detection (step S6).
  • the CIS 11 transitions to the motion detection mode in response to the request of step S6 (step S7).
  • the AP 12 sets the terminal device 1 to the lock state and transitions to the sleep mode. As a result, the terminal device 1 is locked.
  • the CIS 11 transitions to the sensing mode for motion detection and human detection, and when transitioning to the human detection mode in step S7, it is assumed that a moving object 182 other than a human is captured within the image frame (step S9).
  • the CIS 11 detects motion within the image frame, it transitions from the motion detection mode to the human detection mode (step S10), and enters the human detection mode (step S11). Since the CIS 11 does not detect a person in the image frame in the person detection mode of step S11, the CIS 11 transitions to the motion detection mode (step S12) and returns to the motion detection mode (step S13).
  • step S13 when the user 181 is captured in the image frame (step S14), the CIS 11 transitions to the human detection mode (step S15) and enters the human detection mode (step S16). ).
  • step S16 the CIS 11 detects the user 181 as a person and notifies the AP 12 of human detection (that a person has been detected) (step S17).
  • step S18 the AP 12 requests the CIS 11 for an image for face recognition (sensing image) (step S18).
  • the CIS 11 transitions to the sensing image output mode (sensing mode for sensing image output) in response to the request of step S18 (step S19).
  • the CIS 11 After transitioning to the Sensing image output mode, the CIS 11 transmits the Sensing image (low-resolution image for Sensing) to the AP 12 (step S20).
  • the AP 12 performs face authentication including face detection using the Sensing image in step S20, and sets the terminal device 1 to an unlocked state when the face authentication is successful (step S21).
  • FIG. 13 is a flow chart showing a processing procedure performed by the CIS 11 and AP 12 in terminal lock.
  • the AP 12 sets the CIS 11 to the constant sensing mode only for human detection. That is, the AP 12 requests the CIS 11 to activate the constant sensing mode only by human detection, and causes the CIS 11 to transition to that mode.
  • the CIS 11 performs human detection.
  • the CIS 11 determines whether or not there is a person within the image frame. If the result in step S43 is affirmative, the process returns to step S42, and the process repeats from step S42.
  • step S44 the CIS 11 notifies the AP 12 of non-detection of a person (issues a non-detection of a person flag).
  • step S45 the AP 12 sets the CIS 11 to a sensing mode based on motion detection and human detection. That is, the AP 12 requests the CIS 11 to activate the sensing mode based on motion detection and human detection, and causes the CIS 11 to transition to that mode.
  • step S46 the AP 12 sets the terminal device 1 to the locked state.
  • step S47 the AP 12 transitions to sleep mode.
  • the terminal device 1 is set to the locked state when no person is detected within the image frame.
  • FIG. 14 is a flow chart showing a processing procedure performed by the CIS 11 and AP 12 in terminal unlock.
  • the CIS 11 performs motion detection and human detection by image recognition (sensing processing) in the sensing mode using motion detection and human detection. It is assumed that the AP 12 is in sleep mode.
  • the CIS 11 performs motion detection and determines whether motion has been detected within the image frame. If the result in step S62 is NO, the process returns to step S61. If the determination in step S62 is affirmative, the process proceeds to step S63.
  • step S63 the CIS 11 performs human detection and determines whether or not a human has been detected within the image frame. If the determination in step S63 is NO, the process returns to step S61 and repeats from step S61. If the determination in step S63 is affirmative, the process proceeds to step S64.
  • step S64 the CIS 11 notifies the AP 12 of human detection (that a person has been detected) (issues a human detection flag).
  • step S65 the AP 12 wakes up from sleep mode (transitions to active mode) in response to the notification in step S64.
  • step S66 the AP 12 sets the CIS 11 to sensing mode for sensing image output.
  • the AP 12 acquires the Sensing image from the CIS 11.
  • step S68 the AP 12 uses the Sensing image from the CIS 11 to perform face authentication to verify that the face within the image frame is the user's face.
  • step S69 the AP 12 determines whether or not face authentication has succeeded.
  • step S69 the process proceeds to step S70.
  • step S70 the AP 12 sets the CIS 11 to a sensing mode based on motion detection and human detection.
  • step S71 the AP 12 transitions to sleep mode. The process returns from step S71 to step S61 and repeats from step S61.
  • step S69 If the result in step S69 is affirmative, the process proceeds to step S72.
  • step S72 the AP 12 sets the terminal device 1 to an unlocked state.
  • human detection can be realized with lower processing power than face recognition, human detection can be performed by image recognition of the CIS 11 whose processing power is lower than that of the AP 12, and power consumption can also be reduced. Human detection is performed by image recognition of CIS 11 not only for terminal unlocking but also for terminal locking. , or concentrate on processing other than human detection. Therefore, when human detection is performed by the CIS 11, power consumption required for transmission of image data and power consumed by the AP 12 are reduced compared to when human detection is performed by the AP 12.
  • Human detection by image recognition (DNN processing) of CIS11 can handle occlusions such as hair, masks, or eyeglasses, reflections at the edge of the image frame, and clipping due to scaling of the image frame.
  • the sensitivity of detection (degree of over-detection or omission of detection) may be adjustable.
  • Human detection by image recognition (DNN processing) of the CIS 11 may be restricted to a state in which face recognition is easy to perform, and detection may be made when a person exists within the image frame.
  • a state in which face recognition is easy to perform is, for example, a state in which the size of a person with respect to the image frame is within a range of a predetermined ratio, or a state in which the person faces the front (toward the camera).
  • Such restrictions on human detection may be handled by adjusting the conditions for notifying the AP 12 of the detection of a human (detection trigger) from the CIS 11 .
  • Adjustment of the conditions may be, for example, adjustment of a threshold value (lower limit value) at which it is determined that a person has been detected with respect to the degree of certainty of human detection, which is the detection result of DNN processing.
  • Human detection by image recognition of CIS 11 is performed when motion is detected within the image frame by motion detection. Since motion detection can be performed with a lower processing power than human detection, power consumption is further reduced as compared to when only human detection is performed by the CIS 11 . Only one of the motion detection and the human detection in the CIS 11 may be performed. When only motion detection is performed by the CIS 11, the notification to the AP 12 of the detection of motion within the image frame may be performed in the same manner as when a person is detected.
  • Application example 2 is an application example when the image recognition of CIS 11 is applied to the Smart rotate function of the terminal device 1 .
  • the smart rotate of the terminal device 1 is based on the vertical direction of the image displayed on the display of the terminal device 1 (the vertical direction of the screen of the image), the posture of the terminal device 1, or the face of the user who is watching the display of the terminal device 1. It is a function that rotates according to the direction of (vertical direction).
  • the camera (CIS 11) of the terminal device 1 depending on whether the camera (CIS 11) of the terminal device 1 can detect the direction of the user's face or not, the screen is rotated based on the attitude of the terminal device 1, or the direction of the face is detected.
  • the image recognition of CIS 11 is used to determine whether to rotate the screen. Assume that the terminal device 1 is, for example, a smart phone, and the camera including the CIS 11 takes an image of a range facing the display of the smart phone.
  • FIG. 15 is a diagram for explaining image recognition performed by the CIS 11 and the AP 12 with respect to the Smart rotate function of the terminal device 1.
  • a captured image 221 is an example of a captured image captured by the CIS 11, and represents the state of the subject within the image frame.
  • the user 241 represents, for example, the owner of the terminal device 1 , but does not have to be the owner of the terminal device 1 and may be a user who is viewing the display of the terminal device 1 .
  • motion detection indicated by a motion detection frame 261, human detection indicated by a human detection frame 262, and face detection indicated by a face detection frame 263 are performed as image recognition.
  • a motion detection frame 261, a person detection frame 262, and a face detection frame 263 correspond to the motion detection frame 201, the person detection frame 202, and the face detection frame 203 described with reference to FIG. 7, respectively, so detailed description thereof is omitted. do.
  • the motion detection represented by the motion detection frame 261 and the human detection represented by the human detection frame 262 are both performed by image recognition of the CIS 11.
  • Face detection represented by the face detection frame 263 is performed by image recognition of the AP 12 .
  • the face detection in this application example 2 not only the range of the face within the image frame but also the orientation of the face (head) is detected. Smart rotate rotates the image so that the orientation of the detected face and the orientation of the image displayed on the display are appropriate.
  • FIG. 16 is a flow chart showing a processing procedure performed by the CIS 11 and AP 12 in Smart rotate.
  • the AP 12 activates a video application (application) according to the user's operation.
  • a moving image application is an application that reproduces moving images.
  • the AP 12 reproduces the moving image according to the user's operation.
  • the AP 12 sets the CIS 11 to the constant sensing mode only for human detection. That is, the AP 12 requests the CIS 11 to activate the constant sensing mode only by human detection, and causes the CIS 11 to transition to that mode.
  • the AP 12 measures the tilt of the terminal device 1 (display) using the gyro sensor. For example, the AP 12 detects, as the detection direction, the direction along the display surface with respect to the horizontal plane, which direction is more inclined with respect to the horizontal plane.
  • any sensor may be used to detect the tilt (orientation) of the terminal device 1, and the method of representing the detected tilt is not limited to a particular method.
  • step S95 the AP 12 determines whether or not the tilt of the terminal device 1 is detected based on the measurement result in step S94. If the result in step S95 is NO, the process returns to step S94. If the determination in step S95 is affirmative, the process proceeds to step S96.
  • step S96 the CIS 11 performs human detection by image recognition (sensing processing) in the sensing mode for human detection only.
  • step S97 the CIS 11 determines whether or not a person has been detected within the image frame. If the result in step S97 is NO, the process proceeds to step S98.
  • step S98 the AP 12 rotates the moving image displayed on the display based on the direction detected by the gyro sensor in step S94. For example, the moving image screen is rotated so that the detection direction and the vertical direction of the moving image screen are closest to parallel.
  • step S103 If the result in step S97 is affirmative, the process proceeds to step S99.
  • step S99 the AP 12 sets the CIS 11 to sensing mode for sensing image output.
  • the AP 12 acquires a Sensing image from the CIS 11.
  • FIG. the AP 12 detects the face rotation angle based on the Sensing image from the CIS 11.
  • the face rotation angle is, for example, the rotation angle of the face in the vertical direction with respect to the reference direction (for example, the vertical direction) of the image frame.
  • step S102 the AP 12 rotates the screen of the moving image to be displayed on the display based on the face rotation angle (face direction with respect to the image frame) detected in step S101. For example, the moving image screen is rotated so that the vertical direction of the face and the vertical direction of the moving image screen are most parallel.
  • step S103 the AP 12 determines whether the video application has ended. If the determination in step S103 is NO, the process returns to step S93. If the determination in step S103 is affirmative, the processing of this flowchart ends.
  • Human detection can be realized with lower processing power than face detection, so human detection can be performed by image recognition of CIS11, which has lower processing power than AP12, and power consumption can be reduced. Since human detection is performed by image recognition of CIS 11, compared to the case where human detection is performed by AP 12, there is no need to transmit a large amount of image data from CIS 11 to AP 12 during human detection. The power consumption required for transmitting the image data of the AP 12 and the power consumption consumed by the AP 12 are reduced. The AP 12 can focus on processing other than human detection.
  • Human detection by image recognition (DNN processing) of CIS11 can handle occlusions such as hair, masks, or eyeglasses, reflections at the edge of the image frame, and clipping due to scaling of the image frame.
  • the sensitivity of detection (degree of over-detection or omission of detection) may be adjustable.
  • Human detection by image recognition (DNN processing) of the CIS 11 may be restricted to a state in which the person is gazing at the screen so that the presence of the person within the image frame is detected.
  • a state in which a person is gazing at the screen is, for example, a state in which the size of the person with respect to the image frame is within a predetermined ratio range, or a state in which the person faces the front (direction of the camera).
  • Such restrictions on human detection may be handled by adjusting the conditions for notifying the AP 12 of the detection of a human (detection trigger) from the CIS 11 .
  • Adjustment of the conditions may be, for example, adjustment of a threshold value (lower limit value) at which it is determined that a person has been detected with respect to the degree of certainty of human detection, which is the detection result of DNN processing.
  • Human detection by image recognition of the CIS 11 may be performed when motion is detected within the image frame by motion detection.
  • motion detection can be performed with lower processing power than human detection, so power consumption is further reduced compared to when only human detection is performed by the CIS 11 .
  • Only one of the motion detection and the human detection in the CIS 11 may be performed.
  • the notification to the AP 12 of the detection of motion within the image frame may be performed in the same manner as when a person is detected.
  • Application example 3 is an application example when the image recognition of the CIS 11 is applied to the moving image reproduction control function of the terminal device 1 .
  • the video playback control function is such that, when the terminal device 1 plays a video, the video is played when the user is viewing the video on the display, and the video is automatically played back when the user is not viewing the video. It is a function to stop.
  • the image recognition of the CIS 11 is used to determine whether or not the user is browsing moving images. Assume that the terminal device 1 is, for example, a smart phone, and the camera including the CIS 11 takes an image of a range facing the display of the smart phone.
  • FIG. 17 is a diagram for explaining image recognition performed by the CIS 11 and the AP 12 with respect to the video playback control function of the terminal device 1.
  • a captured image 281 is an example of a captured image captured by the CIS 11, and represents the state of the subject within the image frame.
  • a user 301 represents, for example, the owner of the terminal device 1 , but may not be the owner of the terminal device 1 and may be a user viewing the display of the terminal device 1 .
  • motion detection indicated by a motion detection frame 321 and human detection indicated by a human detection frame 322 are performed as image recognition.
  • a motion detection frame 321 and a human detection frame 322 correspond to the motion detection frame 201 and the human detection frame 202 described with reference to FIG. 7, respectively, so detailed description thereof will be omitted.
  • Motion detection represented by the motion detection frame 321 and human detection represented by the human detection frame 322 are both performed by image recognition of the CIS 11 .
  • the image recognition (sensing) of the CIS 11 is not limited to motion detection and human detection.
  • the image recognition of the CIS 11 may include any one or more of motion detection, human detection, and line-of-sight detection.
  • the line-of-sight detection may be performed not only to identify the direction of the line of sight, but also to detect whether the line of sight is in a specific direction (the direction of the display of the terminal device 1). Any method, such as a method based on detection of the positional relationship between the inner corner of the eye and the iris, may be used to detect the line of sight.
  • the sensing mode of the CIS 11 is the constant sensing mode.
  • FIG. 18 is a sequence diagram exemplifying a processing procedure of moving image reproduction control.
  • FIG. 18 shows the flow of processing relating to the user 301, camera (CIS 11), AP 12, and video playback service 341.
  • the moving image reproduction service 341 corresponds to a program for performing moving image reproduction control.
  • the video playback service 341 when the user 301 activates the video application (step S121), the video playback service 341 is activated.
  • the video playback service 341 requests the CIS 11 to activate the constant sensing mode (step S122).
  • the constant sensing mode is a sensing mode in which at least one of motion detection, human detection, and line-of-sight detection is performed.
  • human detection includes line-of-sight detection. It is assumed that detection results such as whether or not the subject is watching the subject are included.
  • the CIS 11 transitions to the constant sensing mode in response to the request of step S122 (step S123).
  • the CIS 11 detects a person within the image frame because the user 301 is reflected in the image frame. As a result, the CIS 11 notifies the motion picture reproduction service 341 of human detection (that a person has been detected) (step S125).
  • the moving picture reproduction service 341 reproduces the moving picture in response to the instruction of step S124.
  • the CIS 11 notifies the moving image reproduction service 341 of the detection result in the constant sensing mode, such as that no person is detected in the image frame or that the line of sight is not gazing at the display (step S127). ).
  • the video playback service 341 stops playing the video (step S128).
  • the CIS 11 notifies the motion picture reproduction service 341 of the detection result of detection in the constant sensing mode, such as that a person is detected in the image frame or that the line of sight is gazing at the display (step S129).
  • the video playback service 341 resumes playback of the video (step S130).
  • the moving image reproduction service 341 causes the CIS 11 to terminate the continuous sensing mode based only on human detection (step S132).
  • FIG. 19 is a sequence diagram illustrating, for comparison with FIG. 18, the processing procedure of moving image reproduction control when the function equivalent to that of FIG. 18 is realized by image recognition of the AP 12 without using image recognition of the CIS 11.
  • FIG. 19 shows the flow of processing relating to the user 301, camera (CIS 11-1), AP 12-1, user 301, and video playback service 341.
  • the camera (CIS11-1) can be regarded as the CIS11 in FIG. may be AP12-1 corresponds to AP12 in FIG. 18 and has a function of performing image recognition.
  • the user 301 and video playback service 341 correspond to the user 301 and video playback service 341 in FIG. In FIG.
  • step S151 when the user 301 activates the video application (step S151), the video playback service 341 is activated.
  • the video playback service 341 requests the AP 12 to enable constant sensing (step S152). Note that the constant sensing is the same as the image recognition in the constant sensing mode as in the case of FIG. 18, so description thereof will be omitted.
  • AP 12-1 activates CIS 11-1 (camera) and requests AP 12-1 to transmit the captured image (step S153).
  • the user 301 instructs the moving image reproduction service 41 to reproduce the moving image (step S154)
  • the user 301 appears in the image frame.
  • the AP 12-1 acquires the image (captured image) transferred from the CIS 11-1 (step S155), it performs image recognition (sensing) on the image.
  • Image recognition is image recognition (sensing) similar to that in FIG. 18, and includes line-of-sight detection. As a result, the captured image that detects AP 12-1 is transferred to AP 12-1 (step S155).
  • the AP 12-1 detects a person in the image frame, notifies the moving image reproduction service 341 of the detection result (step S156), and stops imaging (stops the camera) to the CIS 11-1. to direct.
  • the video reproduction service 341 reproduces the video in response to the notification in step S156.
  • the CIS 11-1 stops imaging according to the instruction in step S156.
  • the AP 12-1 activates the CIS 11-1 (camera) (step S159) and acquires an image captured by the CIS 11-1 (step S160).
  • the AP 12-1 notifies the motion picture reproduction service 341 of the detection result, such as that no person is detected in the image frame or that the line of sight is not gazing at the display, as a result of sensing (step S161). ), and instructs the CIS 11-1 to stop imaging (stop the camera) (step S162).
  • the video playback service 341 stops playing the video (step S163).
  • the CIS 11-1 stops imaging according to the instruction in step S162.
  • the AP 12-1 activates the CIS 11-1 (camera) for periodic sensing (step S164) and acquires an image captured by the CIS 11-1 (step S165).
  • the AP 12-1 notifies the moving picture reproduction service 341 of the detection result such as that a person is detected in the image frame or that the line of sight is gazing at the display as a detection result of sensing (step S166), and instructs the CIS 11-1 to stop imaging (stop the camera) (step S167).
  • the video playback service 341 resumes playback of the video (step S168).
  • the CIS 11-1 stops imaging according to the instruction in step S167.
  • the video reproduction service 341 causes the AP 12-1 to terminate constant sensing (step S170).
  • FIG. 20 is a flow chart showing a processing procedure performed by the CIS 11 and AP 12 in moving image reproduction control.
  • the video playback service 341 is activated.
  • the moving picture reproduction service 341 sets the CIS 11 to constant sensing mode.
  • the constant sensing mode is the constant sensing mode described with reference to FIG. 18, but here, it is the constant sensing mode that only detects people.
  • step S193 the video playback service 341 plays back the video.
  • step S194 the CIS 11 performs image recognition (sensing) on the captured image (image for detection processing), and notifies the video reproduction service 341 of the detection result.
  • step S195 the moving image reproduction service 341 determines whether or not a person is detected in the image frame (is there a person in the image frame) based on the detection result notified in step S194.
  • step S195 If the result in step S195 is affirmative, the process returns to step S194. If the result in step S195 is NO, the process proceeds to step S196.
  • step S196 the video playback service 341 stops playing the video.
  • step S197 the CIS 11 performs image recognition (sensing) on the captured image (image for detection processing), and notifies the video reproduction service 341 of the detection result.
  • step S198 the video playback service 341 determines whether or not a person is detected within the image frame (is there a person within the image frame) based on the detection result notified in step S197. If the determination in step S198 is NO, the process returns to step S197. If the determination in step S198 is affirmative, the process proceeds to step S199.
  • step S199 the video playback service 341 resumes playback of the video.
  • step S200 the video playback service 341 determines whether the video application has ended. If the determination in step S200 is NO, the process returns to step S194. If the determination in step S200 is affirmative, the processing of this flowchart ends.
  • Human detection by image recognition (DNN processing) of CIS11 can handle occlusions such as hair, masks, or eyeglasses, reflections at the edge of the image frame, and clipping due to scaling of the image frame.
  • the sensitivity of detection may be adjustable.
  • Human detection by image recognition (DNN processing) of the CIS 11 may be restricted to a state in which the person is gazing at the screen so that the presence of the person within the image frame is detected.
  • the state in which the person is gazing at the screen is, for example, a state in which the line of sight is directed toward the screen, a state in which the size of the person with respect to the image frame is within a predetermined ratio, or a state in which the person is in front of the screen. It is facing the direction of the camera.
  • line-of-sight detection as described above, such restrictions on human detection can be handled by adjusting the conditions for notifying (detection trigger) from the CIS 11 to the AP 12 that a person has been detected. good.
  • Adjustment of the conditions may be, for example, adjustment of a threshold value (lower limit value) at which it is determined that a person has been detected with respect to the degree of certainty of human detection, which is the detection result of DNN processing.
  • the line-of-sight detection may be performed by the AP 12 that receives notification from the CIS 11 that a person has been detected within the image frame.
  • Human detection by image recognition of the CIS 11 may be performed when motion is detected within the image frame by motion detection.
  • motion detection can be performed with lower processing power than human detection, so power consumption is further reduced compared to when only human detection is performed by the CIS 11 .
  • Only one of the motion detection and the human detection in the CIS 11 may be performed.
  • the notification to the AP 12 of the detection of motion within the image frame may be performed in the same manner as when a person is detected.
  • Application example 4 is an application example in which the image recognition of the CIS 11 is applied to the peeping prevention function of the terminal device 1 .
  • Peeping prevention of the terminal device 1 is a function of performing predetermined warning processing such as displaying a warning message when a person other than the user is watching the display of the terminal device 1 .
  • the image recognition of the CIS 11 is used to determine whether or not there is a person other than the user at a position where the display of the terminal device 1 can be viewed.
  • the terminal device 1 is, for example, a smart phone, and the camera including the CIS 11 takes an image of a range facing the display of the smart phone.
  • FIG. 21 is a diagram for explaining image recognition performed by the CIS 11 and the AP 12 for the Peeping prevention function of the terminal device 1.
  • a captured image 351 is an example of a captured image captured by the CIS 11, and represents the subject within the image frame.
  • a user 361 represents the owner of the terminal device 1, for example.
  • a person 362 represents a person other than the user 361 .
  • Moving object 363 represents any moving object.
  • the motion detection frame 381 and the human detection frames 382-1 and 382-2 correspond to the motion detection frame 201 and the human detection frame 202 described with reference to FIG. 7, respectively, so detailed description thereof will be omitted.
  • Human detection frames 383-1 and 383-2 represent ranges in which humans are detected within the image frame by human detection. For the human detection frames 383-1 and 383-2, it is determined whether or not the user is gazing at the display (peeping determination). Motion detection represented by motion detection frame 381 and human detection represented by human detection frames 382-1 and 382-2 are both performed by image recognition of CIS 11. FIG. Human detection and peeping determination represented by human detection frames 383-1 and 383-2 are performed by image recognition of AP12. Note that the peeping determination may be performed by detecting whether or not the line of sight is in the direction of the display of the terminal device 1 in the same manner as the line of sight detection described in Application Example 3.
  • FIG. 22 is a flowchart showing a processing procedure performed by the CIS 11 and AP 12 in Peeping prevention.
  • the AP 12 activates a predetermined application (mail application, etc.).
  • the AP 12 displays a display image (message or the like) corresponding to the type of application on the display.
  • the AP 12 sets the CIS 11 to constant sensing mode.
  • the constant sensing mode is a constant sensing mode based only on human detection, but may be a constant sensing mode based on motion detection and human detection.
  • step S224 the CIS 11 performs human detection by image recognition (sensing processing).
  • step S225 the CIS 11 determines whether or not multiple people have been detected within the image frame. If the result in step S225 is NO, the process returns to step S224. If the determination in step S225 is affirmative, the process proceeds to step S226. In step S226, the AP 12 sets the CIS 11 to sensing mode for sensing image output.
  • step S227 the AP 12 acquires a low-resolution Sensing image (low-resolution image for Sensing) transmitted from the CIS 11.
  • step S228, the AP 12 performs image recognition on the Sensing image acquired in step S227, and performs peep determination. For example, the AP 12 detects a plurality of people within the image frame, and detects the number of people who are gazing at the display based on the line of sight of each of the detected plurality of people.
  • step S229 the AP 12 determines whether or not multiple people are gazing at the display. If the result in step S229 is NO, the process returns to step S223. If the determination in step S229 is affirmative, the process proceeds to step S230.
  • step S230 the AP 12 performs warning processing.
  • the warning process can be any process that prevents peeping, such as displaying a message to the effect that peeping is taking place, or simplifying the displayed image on the display. good.
  • step S231 the AP 12 determines whether or not the application is terminated according to a user's instruction or the like. If the determination in step S231 is NO, the process returns to step S223. If the determination in step S231 is affirmative, the processing of this flowchart ends.
  • Human detection by image recognition (DNN processing) of CIS11 can handle occlusions such as hair, masks, or eyeglasses, reflections at the edge of the image frame, and clipping due to scaling of the image frame.
  • the sensitivity of detection may be adjustable.
  • Human detection by image recognition (DNN processing) of the CIS 11 may be restricted to a state in which the person is gazing at the screen so that the presence of the person within the image frame is detected.
  • the state in which the person is gazing at the screen is, for example, a state in which the line of sight is directed toward the screen, a state in which the size of the person with respect to the image frame is within a predetermined ratio, or a state in which the person is in front of the screen. It is facing the direction of the camera.
  • Such restrictions on human detection may be achieved by performing line-of-sight detection in the CIS 11 and reflecting the result in the result of human detection, or by notifying the AP 12 of the detection of a person from the CIS 11 (detection trigger).
  • Adjustment of the conditions may be, for example, adjustment of a threshold value (lower limit value) at which it is determined that a person has been detected with respect to the degree of certainty of human detection, which is the detection result of DNN processing.
  • the line-of-sight detection may be performed by the AP 12 that receives notification from the CIS 11 that a person has been detected within the image frame.
  • Human detection by image recognition of the CIS 11 may be performed when motion is detected within the image frame by motion detection.
  • motion detection can be performed with lower processing power than human detection, so power consumption is further reduced compared to when only human detection is performed by the CIS 11 .
  • Only one of the motion detection and the human detection in the CIS 11 may be performed.
  • the notification to the AP 12 of the detection of motion within the image frame may be performed in the same manner as when a person is detected.
  • Application example 5 is an application example in which the image recognition of the CIS 11 is applied to the function of automatic transition between voice call and video call (automatic voice/video call transition) of the terminal device 1 .
  • the automatic transition of the voice/video call of the terminal device 1 is performed automatically when the user (person) is detected in the image frame while the terminal device 1 is holding the terminal device 1 to the ear and making a voice call.
  • the image recognition of the CIS 11 is used to detect whether or not a person exists within the image frame.
  • the terminal device 1 is, for example, a smart phone, and the camera including the CIS 11 takes an image of a range facing the display of the smart phone.
  • FIG. 23 shows an example of a specific situation (scene development) until the terminal device 1 switches from a voice call to a video call by the voice/video call automatic transition function of the terminal device 1, and state transitions of the CIS 11 and AP 12 at that time. It is a figure explaining an outline.
  • a scene C1 is a scene in which the user holds the terminal device 1 in the state of voice communication to his/her ear and makes a call.
  • Scene C2 is a scene in which the user moves to a state of viewing the display of the terminal device 1 (a state in which the user is captured by the camera of the terminal device 1) in order to switch from the voice call to the video call.
  • a scene C3 is a scene in which the user is having a conversation on the terminal device 1 in a state where the voice call has been switched to the video call.
  • the state of AP12 is "Voice call”.
  • “Voice call” is a state in which the AP 12 is processing a voice call.
  • the AP 12 transmits to the CIS 11 a notification signal (“Turn sensing mode”) requesting activation of the sensing mode based on motion detection and human detection during “Voice call”.
  • CIS 11 transitions from standby mode (“HW Standby”) in which power is not supplied by a notification signal (“Turn sensing mode”) from AP 11 to standby mode (“SW Standby”) in which power is supplied. Transition to the motion detection mode (“Sensing Mode (Motion Detection)”).
  • the terminal device 1 moves from the position where it is placed against the user's ear to the front side of the user. Detect motion within the frame. As a result, the CIS 11 automatically transitions from the motion detection mode (“Sensing Mode (Motion Detection)”) to the human detection mode (“Sensing Mode (Human Existence)”). In the human detection mode (“Sensing Mode (Human Existence)”), the CIS 11 detects a person (user) within the image frame by image recognition human detection. When the CIS 11 detects a person in the image frame, the CIS 11 transmits a detection trigger (“Human Detected”) (human detection notification signal) to the AP 12 and transitions to the standby mode (“SW Standby”). As a result, the AP 12 transitions from voice call processing (“Voice call”) to “Switch from voice call to video call”. “Switch from voice call to video call” is a state in which switching from voice call to video call is being processed.
  • a detection trigger (“Human Det
  • AP 11 transitions from "Switch from voice call to video call” to "Video call”.
  • "Video call” is a state in which a video call is being processed.
  • AP11 transitions from “Switch from voice call to video call” to “Video call”
  • it sends a notification signal ("(Request viewing image") to CIS11 requesting transmission of a high-resolution viewing image.
  • the CIS 11 receives the notification signal (“(Request viewing image”) from the AP 12, the CIS 11 transitions from the standby mode (“SW Standby”) to the “Viewing Mode (Streaming)”.
  • SW Standby standby mode
  • Viewing Mode Streaming
  • CIS 11 When the CIS 11 transitions to the viewing mode ("Viewing Mode (Streaming)"), it transmits a high-resolution image for viewing to the P 12. Images are continuously transmitted: AP 12 receives high-resolution images for viewing from CIS 11 and transmits them to the terminal device of the other party of the video call.
  • Viewing Mode Streaming
  • FIG. 24 is a flow chart showing the processing procedure performed by the CIS 11 and AP 12 in the voice/video call automatic transition.
  • the AP 12 activates a call application according to a user's instruction or the like.
  • the AP 12 starts voice call processing.
  • the AP 12 sets the CIS 11 to the constant sensing mode.
  • the constant sensing mode is a constant sensing mode based only on human detection, but may be a constant sensing mode based on motion detection and human detection.
  • the CIS 11 performs human detection by image recognition (sensing processing).
  • the CIS 11 determines whether or not a person has been detected within the image frame.
  • step S255 the process returns to step S254. If the determination in step S255 is affirmative, the process proceeds to step S256.
  • step S256 the AP 12 sets the CIS 11 to viewing mode ("Viewing Mode (Streaming)").
  • the viewing mode (“Viewing Mode (Streaming)”) is a mode in which high-resolution images for viewing are continuously transmitted to the AP 12 . Note that the resolution of the image transmitted to the AP 12 in the viewing mode may be equal to or lower than that of the low-resolution image for sensing.
  • step S257 the AP 12 switches to video call processing.
  • the AP 12 transmits the high-resolution image for viewing received from the CIS 11 in step S256 to the terminal device of the other party.
  • step S258 the AP 12 performs face detection by image recognition on the viewing high-resolution image.
  • step S259 the AP 12 determines whether or not a face has been detected within the image frame as a result of step S258.
  • step S259 determines whether or not the video call has ended by user's operation or the like.
  • step S260 If the answer is NO in step S260, the process proceeds to step S261.
  • step S261 the AP 12 switches from video call processing to voice call processing. Processing returns from step S261 to step S253. If the determination in step S260 is affirmative, the processing of this flowchart ends.
  • the AP 12 can focus on processing other than human detection.
  • Human detection by image recognition (DNN processing) of CIS11 can handle occlusions such as hair, masks, or eyeglasses, reflections at the edge of the image frame, and clipping due to scaling of the image frame.
  • the sensitivity of detection (degree of over-detection or omission of detection) may be adjustable.
  • Human detection by image recognition (DNN processing) of the CIS 11 may be restricted to a state in which face detection is easy to perform, and detection may be made when a person exists within the image frame.
  • the state in which face detection is performed is, for example, a state in which the size of the person with respect to the image frame is within a predetermined ratio range, or a state in which the person faces the front (toward the camera).
  • Such restrictions on human detection may be handled by adjusting the conditions for notifying the AP 12 of the detection of a human (detection trigger) from the CIS 11 .
  • Adjustment of the conditions may be, for example, adjustment of a threshold value (lower limit value) at which it is determined that a person has been detected with respect to the degree of certainty of human detection, which is the detection result of DNN processing.
  • Human detection by image recognition of the CIS 11 may be performed when motion is detected within the image frame by motion detection.
  • motion detection can be performed with lower processing power than human detection, so power consumption is further reduced compared to when only human detection is performed by the CIS 11 .
  • Only one of the motion detection and the human detection in the CIS 11 may be performed.
  • the notification to the AP 12 of the detection of motion within the image frame may be performed in the same manner as when a person is detected.
  • Application example 6 is an application example when the image recognition of the CIS 11 is applied to the screen brightness automatic adjustment function of the terminal device 1 .
  • the screen brightness automatic adjustment of the terminal device 1 is a function of automatically adjusting the brightness of the screen of the display according to the brightness of the environment around the terminal device 1 (display). For example, the terminal device 1 increases the brightness of the screen as the surrounding environment of the display becomes brighter to make the display image easier to see. At that time, even if the surrounding environment is bright, the user may not be viewing the screen. In this case, power consumption is wasted, so the brightness of the screen is reduced.
  • the image recognition of the CIS 11 is used to detect whether or not the user is browsing the screen. Assume that the terminal device 1 is, for example, a smart phone, and the camera including the CIS 11 takes an image of a range facing the display of the smart phone.
  • FIG. 25 is a flow chart showing a processing procedure performed by the CIS 11 and AP 12 in screen brightness automatic adjustment.
  • the AP 12 sets the CIS 11 to constant sensing mode.
  • the constant sensing mode is a constant sensing mode that only detects people, but it may be a constant sensing mode that uses motion detection and human detection.
  • AP12 acquires the ALS value from CIS11. The ALS value is calculated by the exposure controller 61 of the CIS 11 in FIG.
  • the CIS 11 performs human detection by image recognition (sensing).
  • the CIS 11 determines whether or not a person has been detected within the image frame as a result of image recognition. Note that step S284 may be a determination process of the AP 12 that has acquired the detection result from the CIS 11 .
  • step S285 the AP 11 determines that the user is not viewing the screen of the display, and sets or changes the brightness of the screen to the minimum value.
  • step S285 the process of this flowchart ends.
  • step S286 the AP 11 determines that the user is viewing the screen of the display, and sets the brightness of the screen to the optimum brightness value based on the ALS value obtained in step S282.
  • step S286 ends, the process of this flowchart ends.
  • the screen brightness (luminance value) is adjusted based on the measured value (ALS value). For example, if the environment is bright, the brightness of the screen is also increased to make the screen easier to see. When no person is detected within the image frame, the brightness of the screen is set to the minimum value.
  • the minimum screen brightness value is a predetermined brightness value, for example, the lower limit of the brightness range that can be adjusted in the automatic screen brightness adjustment.
  • the ALS value may be acquired from a sensor other than the CIS 11, but by using the ALS value detected by the CIS 11, the terminal device 1 does not need to be equipped with a sensor such as an illuminance sensor other than the CIS 11 for automatic screen brightness adjustment. and the need to activate such sensors. Therefore, power consumption is reduced compared to the case of using a sensor. Since human detection can be realized with low processing power, human detection can be performed by image recognition of the CIS 11 whose processing power is lower than that of the AP 12 .
  • the AP 12 can focus on processing other than human detection.
  • Human detection by image recognition (DNN processing) of CIS11 can handle occlusions such as hair, masks, or eyeglasses, reflections at the edge of the image frame, and clipping due to scaling of the image frame.
  • the sensitivity of detection may be adjustable.
  • Human detection by image recognition (DNN processing) of the CIS 11 may be restricted to a state in which the person is gazing at the screen so that the presence of the person within the image frame is detected.
  • the state in which the person is gazing at the screen is, for example, a state in which the line of sight is directed toward the screen, a state in which the size of the person with respect to the image frame is within a predetermined ratio, or a state in which the person is in front of the screen. It is facing the direction of the camera.
  • Such restrictions on human detection may be achieved by performing line-of-sight detection in the CIS 11 and reflecting the result in the result of human detection, or by notifying the AP 12 of the detection of a person from the CIS 11 (detection trigger).
  • Adjustment of the conditions may be, for example, adjustment of a threshold value (lower limit value) at which it is determined that a person has been detected with respect to the degree of certainty of human detection, which is the detection result of DNN processing.
  • the line-of-sight detection may be performed by the AP 12 that receives notification from the CIS 11 that a person has been detected within the image frame.
  • Human detection by image recognition of the CIS 11 may be performed when motion is detected within the image frame by motion detection.
  • motion detection can be performed with lower processing power than human detection, so power consumption is further reduced compared to when only human detection is performed by the CIS 11 .
  • Only one of the motion detection and the human detection in the CIS 11 may be performed.
  • the notification to the AP 12 of the detection of motion within the image frame may be performed in the same manner as when a person is detected.
  • Application example 7 is an application example when the image recognition of the CIS 11 is applied to the pet camera function of the terminal device 1 .
  • the pet camera of the terminal device 1 is, for example, a function of taking a picture of a pet with a camera of the terminal device 1 installed indoors or the like.
  • image recognition of the CIS 11 is used to detect whether or not the pet is present within the image frame when recording is performed only when the pet is reflected within the image frame.
  • the terminal device 1 may be mainly used as a camera (imaging device), or may be a device such as a smartphone whose main purpose is other than a camera.
  • the detection target of object detection in image recognition (DNN processing) of the CIS 11 is not a person but a pet to be photographed, and the detection target differs depending on the type of pet.
  • the type of pet is a dog, and dog detection is performed instead of human detection by image recognition of the CIS 11 .
  • FIG. 26 shows an example of a specific situation (scene development) until the terminal device 1 transitions from the non-recording state to the recording state, and from the recording state to the non-recording state by the pet camera function of the terminal device 1, and the situation at that time.
  • FIG. 2 is a diagram illustrating an outline of state transitions of CIS 11 and AP 12;
  • Scenes C1 to C3 in FIG. 26 are scenes in which the camera of the terminal device 1 captures a part of the room within the image frame with the terminal device 1 fixed to the floor of the room or the like.
  • the camera of the terminal device 1 starts photographing, and at that point the pet dog is out of the image frame.
  • the dog is inside the frame, and in scene C3, the dog is out of the frame again.
  • the state of AP 12 is the state in which CIS 11 startup settings are being made (“CIS startup settings”), and the state of CIS 11 is in standby mode (“HW Standby”).
  • the AP 12 transitions to the sleep mode (“AP Sleep”), and at that time, transmits a notification signal (“Turn sensing mode”) requesting activation of the sensing mode by motion detection and dog detection to the CIS 11.
  • AP Sleep the sleep mode
  • Turn sensing mode requesting activation of the sensing mode by motion detection and dog detection
  • the CIS 11 transitions from the standby mode (“HW Standby") to the standby mode (“SW Standby") by the notification signal (“Turn sensing mode") from the AP 12, motion detection in the sensing mode by motion detection and dog detection Transition to mode ("Sensing Mode (Motion Detection)").
  • the CIS 11 detects motion within the image frame when the dog enters the image frame in the motion detection mode (“Sensing Mode (Motion Detection)”). "Sensing Mode (Motion Detection)”) to dog detection mode ("Sensing Mode (Dog Existence)”) automatically.
  • a dog detection mode (“Sensing Mode (Dog Existence)”) indicates a mode that detects a dog instead of detecting a person in the human detection mode.
  • the CIS 11 transmits a detection trigger (“Dog Detected”) (a dog detection notification signal) to the AP 12 .
  • the AP 12 wakes up from sleep mode ("AP Sleep”) (transitions to active mode) and transitions to recording mode (“Recording”).
  • AP Sleep transitions to active mode
  • Recording the AP 12 transmits a notification signal requesting activation of the viewing mode (“Viewing Mode”) to the CIS 11 .
  • the CIS 11 receives a notification signal (“(Request viewing image”) requesting activation of the viewing mode from the AP 12, it transitions to the standby mode (“SW Standby”), and then the viewing mode (“Viewing Mode ( Viewing Mode (Streaming)”.
  • the viewing mode "Viewing Mode (Streaming)” is a mode in which high-resolution high-resolution images for viewing are continuously transmitted to the AP 12.
  • the CIS 11 changes to the viewing mode ("Viewing Mode (Streaming)"). Streaming)”), it continuously transmits high-resolution images for viewing to the AP 12.
  • the AP 12 receives the high-resolution images for viewing from the CIS 11, and transmits the received high-resolution images for viewing.
  • the AP 12 performs dog detection on the high-resolution image for viewing by image recognition (DNN processing), etc., and detects the dog in the image frame. exists or not.
  • the AP 12 does not detect a dog in image recognition (dog detection) for the viewing high-resolution image, so it stops recording in the recording mode ("Recording") and transitions to the "CIS startup setting” state.
  • the AP 12 makes a setting to start up the CIS 11 by "CIS start setting”, and then transitions to a sleep mode ("AP Sleep”).
  • AP Sleep a sleep mode
  • the AP 12 transmits to the CIS 11 a notification signal (“Turn sensing mode”) requesting activation of the sensing mode based on motion detection and dog detection.
  • FIG. 27 is a flow chart showing a processing procedure performed by the CIS 11 and AP 12 in the pet camera.
  • the AP 12 sets the CIS 11 to a sensing mode based on motion detection and animal detection.
  • the pet is a dog in FIG. 26, but an animal in this description.
  • the AP 12 transitions to sleep mode.
  • the CIS 11 performs motion detection by image recognition in sensing mode by motion detection and animal detection.
  • the CIS 11 determines whether motion has been detected within the image frame.
  • step S304 If the answer in step S304 is NO, the process returns to step S303. If YES in step S304, the process proceeds to step S305.
  • step S305 the CIS 11 performs animal detection by image recognition in sensing mode using motion detection and animal detection, and determines whether or not an animal has been detected within the image frame. If the result in step S305 is NO, the process returns to step S303. If the result in step S305 is affirmative, the process proceeds to step S306.
  • the CIS 11 transmits a detection flag to the AP 12.
  • the AP 12 transitions from sleep mode to active mode and starts up.
  • the AP 12 sets the CIS 11 to viewing mode ("Viewing Mode (Streaming)".
  • the AP 12 starts recording a high-resolution image for viewing from the CIS 11.
  • the AP 12 detects an animal by image recognition of the high-resolution image for viewing from the CIS 11.
  • the AP 12 determines whether or not an animal has been detected in the image frame as a result of the animal detection in step S310.
  • step S311 If the determination in step S311 is affirmative, the process returns to step S310, and if the determination in step S311 is negative, the process proceeds to step S312, in which the AP 12 stops recording. Then, it is determined whether or not to end the pet camera according to the user's instruction, etc. If the result in step S313 is NO, the process returns to step S301, and if the result in step S314 is YES, this flowchart is executed. processing ends.
  • animal detection is performed to determine whether or not a pet (animal) is detected within the image frame. If no animal is detected within the image frame, no recording is performed. Since animal detection can be realized with low processing power, human detection can be performed by image recognition of CIS 11, which has lower processing power than AP 12, and power consumption can also be reduced. Since animal detection is performed by image recognition of CIS 11 except during recording, there is no need to transmit a large amount of image data from CIS 11 to AP 12 during animal detection, and AP 12 is on standby in a power saving state such as sleep mode. Therefore, the power consumption required for transmitting image data during animal detection and the power consumption consumed by the AP 12 are reduced compared to the case where the AP 12 performs animal detection. Note that the object detected by animal detection may be a person.
  • Animal detection by image recognition (DNN processing) of CIS11 may be able to adjust the sensitivity of animal detection (degree of over-detection or detection omission) so that it can cope with occlusion, cut-out, etc.
  • Animal detection by image recognition (DNN processing) of the CIS 11 may be restricted to a certain state so that it is detected when a person exists within the image frame.
  • the constant state is, for example, a state in which the size of the animal is within a predetermined proportion of the image frame, or a state in which the animal faces the front (toward the camera).
  • Such limitations on animal detection may be accommodated by adjusting the conditions under which the CIS 11 notifies the AP 12 that an animal has been detected (detection trigger). Adjustment of the conditions may be, for example, adjustment of a threshold value (lower limit) for determining that an animal has been detected with respect to the animal detection certainty factor, which is the detection result of DNN processing.
  • Animal detection by image recognition of CIS11 is performed when motion is detected within the image frame by motion detection. Since motion detection can be performed with a lower processing power than animal detection, power consumption is further reduced compared to when only animal detection is performed in the CIS 11 . Only one of motion detection and animal detection in the CIS 11 may be performed. If only motion detection is performed by the CIS 11, the notification to the AP 12 of the detection of motion within the image frame may be made in the same way as when an animal is detected.
  • Application example 8 is an application example when the image recognition of the CIS 11 is applied to the security doorphone function of the terminal device 1 .
  • the security door phone of the terminal device 1 takes a picture of a person outside the door with a camera installed around the door. In the case of a person who has not completed the check-in, the function notifies the person in the room by an alarm without unlocking the door.
  • the image recognition of the CIS 11 is used to determine whether or not there is a person within the image frame. It is assumed that the terminal device 1 is a device that constitutes a security door phone.
  • FIG. 28 is a diagram for explaining an example of a specific situation (scene development) in which the door is unlocked or not unlocked by the security doorphone function of the terminal device 1 and an overview of the state transitions of the CIS 11 and AP 12 at that time.
  • Scene C1 in FIG. 28 is a case where nothing changes within the image frame of the camera installed around the door.
  • Scenes C2, C3, and C4 are scenes that occur selectively with respect to scene C1
  • scene C2 is a scene in which an authenticated person is detected in the image frame.
  • An authenticated person is a person whose face has been previously registered for face authentication and who is permitted to unlock the door.
  • Scene C3 is a scene in which an unauthenticated person is detected within the image frame.
  • An unauthenticated person represents a person other than an authenticated person.
  • Scene C4 is a scene in which only a moving object other than a person is detected in the image frame after scene C2 or scene C3.
  • the state of AP 12 is the state in which CIS 11 startup settings are being made (“CIS startup settings”), and the state of CIS 11 is in standby mode (“HW Standby”).
  • the AP 12 transitions to the sleep mode (“AP Sleep”), and at that time, transmits a notification signal (“Turn sensing mode”) requesting activation of the sensing mode based on motion detection and human detection to the CIS 11.
  • AP Sleep the sleep mode
  • Turn sensing mode requesting activation of the sensing mode based on motion detection and human detection
  • the CIS 11 transitions from the standby mode (“HW Standby") to the standby mode (“SW Standby") by the notification signal (“Turn sensing mode") from the AP 12, motion detection in the sensing mode by motion detection and human detection Transition to the mode (“Sensing Mode (Motion Detection)").
  • the CIS 11 detects motion within the image frame when an authenticated person enters the image frame in the motion detection mode ("Sensing Mode (Motion Detection)"). As a result, the CIS 11 automatically transitions from the motion detection mode (“Sensing Mode (Motion Detection)”) to the human detection mode (“Sensing Mode (Human Existence)”). When the CIS 11 detects a person in the image frame in the human detection mode (“Sensing Mode (Human Existence)”), the CIS 11 transmits a detection trigger (“Human Detected”) (human detection notification signal) to the AP 12 . As a result, the AP 12 wakes up from sleep mode (“AP Sleep”) (transitions to active mode) and transitions to face recognition mode (“Face Detection, Face Identification”).
  • AP Sleep sleep mode
  • Face Detection, Face Identification transitions to face recognition mode
  • the AP 12 When the AP 12 transitions to the face authentication mode ("Face Detection, Face Identification”), the AP 12 transmits to the CIS 11 a notification signal requesting activation of the sensing mode for sensing image output (“Sensing image streaming”). As a result, the CIS 11 transitions to the standby mode (“SW Standby”) and then to the sensing mode for sensing image output (“Sensing image streaming”). When the CIS 11 transitions to the sensing mode for sensing image output (“Sensing image streaming”), it continuously transmits low-resolution sensing images (low-resolution images for sensing) to the AP 12 .
  • the sensing image streaming When the CIS 11 transitions to the sensing mode for sensing image output (“Sensing image streaming”), it continuously transmits low-resolution sensing images (low-resolution images for sensing) to the AP 12 .
  • the AP 12 performs face detection on the Sensing image from the CIS 11 prior to face authentication, and compares the detected face with the authenticated face. As a result, since the person detected in the image frame has been authenticated, the AP 12 succeeds in face authentication and unlocks the door.
  • the person detected in the image frame is unauthenticated, so the AP 12 fails face authentication, does not unlock the door (or locks the door), and is visited by an unauthenticated person. Notify people in the house of this by an alarm or the like.
  • the face authentication mode is changed to the "CIS activation setting" state.
  • the AP 12 makes a setting to start up the CIS 11 by "CIS start setting”, and then transitions to a sleep mode ("AP Sleep").
  • AP Sleep a sleep mode
  • the AP 12 transmits to the CIS 11 a notification signal (“Turn sensing mode”) requesting activation of the sensing mode based on motion detection and human detection.
  • CIS 11 changes from the sensing mode for Sensing image output ("Sensing image streaming") to the standby mode ("SW Standby") by a notification signal (“Turn Sensing mode”) from AP 12, and then detects motion and human. Transition to the motion detection mode (“Sensing Mode (Motion Detection)”) in the sensing mode. In scene C4, since no person is detected within the image frame, the motion detection mode (“Sensing Mode (Motion Detection)”) and the human detection mode (“Sensing Mode (Human Existence)”) in the sensing mode are repeated.
  • FIG. 29 is a flow chart showing a processing procedure performed by the CIS 11 and AP 12 in the security doorphone.
  • the AP 12 sets the CIS 11 to a sensing mode based on motion detection and human detection.
  • the AP 12 enters sleep mode.
  • the CIS 11 performs motion detection and human detection in sensing mode by motion detection and human detection.
  • the CIS 11 determines whether motion has been detected within the image frame. If the result in step S324 is NO, the process returns to step S323. If the result in step S324 is affirmative, the process proceeds to step S325.
  • step S325 the CIS 11 determines whether or not a person has been detected within the image frame. If the result in step 325 is NO, the process proceeds to step S324. If the determination in step S325 is affirmative, the process proceeds to step S326.
  • step S326 the CIS 11 sets a detection flag indicating that a person has been detected to the AP 12 (transmits a detection trigger).
  • the AP 12 is activated by the detection flag at step S326.
  • step S328 the AP 12 sets the CIS 11 to sensing mode for Sensing image output.
  • the AP 12 acquires the Sensing image transmitted from the CIS 11.
  • FIG. the AP 12 performs face authentication based on the sensing image acquired at step S329.
  • step S331 the AP 12 determines whether or not there is an authenticated person within the image frame based on the result of face authentication in step S330. If the result in step S331 is NO, the process proceeds to step S332.
  • step S332 the AP 12 performs alarm processing.
  • the alarm processing includes processing for locking the door, processing for sounding an alarm, processing for reporting to a security company, and the like. Processing proceeds to step S334.
  • step S333 the AP 12 unlocks the door. Processing proceeds to step S334.
  • step S334 the AP 12 determines whether or not to end the process according to a user's instruction or the like. If the result in step S334 is NO, the process returns to step S321. If the determination in step S334 is affirmative, the processing of this flowchart ends.
  • the object to be locked is not limited to the door of a house, and may be any opening/closing part that can be opened and closed, such as a car door, a building entrance door, a locker door, and a window of a house.
  • Human detection by image recognition (DNN processing) of CIS11 can handle occlusions such as hair, masks, or eyeglasses, reflections at the edge of the image frame, and clipping due to scaling of the image frame.
  • the sensitivity of detection (degree of over-detection or omission of detection) may be adjustable.
  • Human detection by image recognition (DNN processing) of the CIS 11 may be restricted to a state in which face recognition is easy to perform, and detection may be made when a person exists within the image frame.
  • a state in which face recognition is easy to perform is, for example, a state in which the size of the person with respect to the image frame is within a range of a predetermined ratio, or a state in which the person faces the front (toward the camera).
  • Such restrictions on human detection may be handled by adjusting the conditions for notifying the AP 12 of the detection of a human (detection trigger) from the CIS 11 .
  • Adjustment of the conditions may be, for example, adjustment of a threshold value (lower limit value) at which it is determined that a person has been detected with respect to the degree of certainty of human detection, which is the detection result of DNN processing.
  • Human detection by image recognition of CIS 11 is performed when motion is detected within the image frame by motion detection. Since motion detection can be performed with a lower processing power than human detection, power consumption is further reduced as compared to when only human detection is performed by the CIS 11 . Only one of the motion detection and the human detection in the CIS 11 may be performed. When only motion detection is performed by the CIS 11, the notification to the AP 12 of the detection of motion within the image frame may be performed in the same manner as when a person is detected.
  • Application Example 9 is an application example in which the image recognition of the CIS 11 is applied to the application activation function of the terminal device 1 using a 1D code (one-dimensional code) or a 2D code (two-dimensional code).
  • a case will be described in which the image recognition of the CIS 11 is applied to the function of activating an application using a 2D code.
  • Image recognition of CIS11 can be applied in the same way as in this description, even if the application is launched using a 1D code such as a bar code instead of a 2D code, or if any code is used between the 1D code and the 2D code.
  • Application activation by the 2D code of the terminal device 1 is a function of activating an application associated with the 2D code when the 2D code is detected within the image frame of the captured image of the camera (CIS 11) of the terminal device 1.
  • image recognition of CIS 11 is used to determine whether a 2D code is detected within the image frame.
  • the detection target for object detection in image recognition (DNN processing) of CIS11 is not a person, but a 2D code.
  • 2D code detection is performed by image recognition of the CIS 11 instead of human detection.
  • a smartphone is assumed as the terminal device 1, for example.
  • the camera including the CIS 11 may be a camera that captures the back side of the display instead of capturing the range facing the display of the smartphone.
  • FIG. 30 shows an example of a specific situation (scene development) until an application associated with a 2D code is activated by the application activation function by the 2D code of the terminal device 1, and an overview of the state transitions of the CIS 11 and AP 12 at that time. It is a figure explaining.
  • a scene C1 is a scene in which the user carries the terminal device 1 in a standby state or an activated state.
  • scene C2 when the user purchases drinking water from a vending machine, the 2D code displayed on the vending machine is imaged by the camera of the terminal device 1, and an application (payment application, etc.) associated with the 2D code is displayed. This is the scene that started it.
  • the state of AP 12 is the state in which CIS 11 startup settings are being made (“CIS startup settings”), and the state of CIS 11 is in standby mode ("HW Standby”).
  • AP 12 transitions to sleep mode ("AP Sleep”) or idle mode ("AP Idle”), and at that time, a notification signal ( "Turn sensing mode”).
  • CIS 11 transitions from standby mode ("HW Standby”) to standby mode ("SW Standby") by a notification signal (“Turn sensing mode”) from AP 12, and then detects movement in sensing mode by motion detection and 2D code detection. Transition to detection mode ("Sensing Mode (Motion Detection)").
  • the CIS 11 moves the terminal device 1 near the position of the 2D code of the vending machine in the motion detection mode ("Sensing Mode (Motion Detection)”), thereby detecting motion within the image frame. To detect. As a result, the CIS 11 automatically transitions from the motion detection mode (“Sensing Mode (Motion Detection)”) to the 2D code detection mode (“Sensing Mode (Code Existence)”). When the CIS11 detects a 2D code within the image frame in the 2D code human detection mode (“Sensing Mode (Human Existence)”), it sends a detection trigger ("Code Detected") (detection notification signal) to the AP12. do. As a result, the AP 12 wakes up (transitions to active mode) from sleep mode (“AP Sleep”) or idle mode (“AP Idle”), and transitions to 2D code authentication mode (“Code recognition”).
  • AP Sleep sleep mode
  • AP Idle idle mode
  • Code recognition transitions to 2D code authentication mode
  • the AP 12 When the AP 12 transitions to the 2D code authentication mode ("Code recognition”), it transmits to the CIS 11 a notification signal requesting activation of the sensing mode for Sensing image output (“Sensing image streaming”). As a result, the CIS 11 transitions to the standby mode (“SW Standby”) and then to the sensing mode for sensing image output (“Sensing image streaming”). When the CIS 11 transitions to the sensing mode for sensing image output (“Sensing image streaming”), it continuously transmits low-resolution sensing images (low-resolution images for sensing) to the AP 12 . The AP 12 performs 2D code analysis on the Sensing image from the CIS 11 prior to authentication in the 2D code authentication mode.
  • the AP 12 launches the app associated with the 2D code in the image frame after transitioning to "CIS setting" ("App Execute”).
  • the AP 12 sends a notification signal ("Turn HW Standby") requesting the CIS 11 to transition to the standby mode (“HW Standby”) when activating the application. ).
  • the CIS 11 transitions to the standby mode (“HW Standby”) in response to the notification signal (“Turn HW Standby”) from the AP 12 .
  • FIG. 31 is a flow chart showing a processing procedure performed by the CIS 11 and AP 12 when starting an application using a 2D code.
  • the AP 12 sets the CIS 11 to a sensing mode based on motion detection and code detection.
  • the AP 12 enters sleep mode.
  • the CIS 11 performs motion detection and 2D code detection in sensing mode by motion detection and code detection.
  • the CIS 11 determines whether motion has been detected within the image frame. If the determination in step S354 is NO, the process returns to step S353. If the result in step S354 is affirmative, the process proceeds to step S355.
  • the CIS 11 determines whether or not the 2D code is detected within the image frame by the 2D code detection.
  • step S355 If the result in step S355 is NO, the process returns to step S353. If the determination in step S355 is affirmative, the process proceeds to step S356.
  • step S356 the CIS 11 transmits to the AP 12 a detection trigger indicating that the 2D code has been detected.
  • the AP 12 is activated by the detection flag at step S356.
  • step S358 the AP 12 sets the CIS 11 to a sensing mode for Sensing image output.
  • step S359 the AP 12 acquires the Sensing image transmitted from the CIS 11.
  • step S360 the AP 12 performs 2D code recognition processing based on the Sensing image acquired in step S359.
  • step S361 the AP 12 determines whether the 2D code recognized at step S360 is a correct 2D code. If the determination in step S361 is NO, the process returns to step S351. If the determination in step S361 is affirmative, the process proceeds to step S362.
  • step S362 the AP 12 analyzes the 2D code.
  • step S363 the AP 12 executes processing corresponding to the 2D code. For example, the AP 12 launches an app (such as a payment app) associated with the 2D code.
  • an app such as a payment app
  • code detection is performed to determine whether or not a 2D code is detected within an image frame when an application is launched using a 2D code (or a 1D code, hereinafter omitted), and a 2D code is detected within the image frame.
  • Chord recognition analysis is performed only when The content of the 2D code is read by code recognition, and if an application (process) corresponding to the 2D code exists, the application (or process) is executed. If the 2D code is not detected by code detection, if the contents of the 2D code cannot be read by 2D code recognition, or if there is no application (process) that supports the 2D code, the application (process) will not be executed.
  • code detection can be realized with lower processing power than code recognition, human detection can be performed by image recognition of CIS 11, which has lower processing power than AP 12, and power consumption can also be reduced. Since code detection is performed by image recognition of the CIS 11, there is no need to transmit a large amount of image data from the CIS 11 to the AP 12 during code detection, and the AP 12 can be on standby in a power saving state such as sleep mode. Therefore, compared with the case where the code detection is performed by the AP 12, the power consumption required for transmitting the image data at the time of code detection and the power consumption consumed by the AP 12 are reduced.
  • Code detection by image recognition of CIS 11 is performed when motion is detected within the image frame by motion detection. Since motion detection can be performed with a lower processing power than code detection, power consumption is further reduced compared to when only code detection is performed in the CIS 11 . Only one of motion detection and code detection in the CIS 11 may be performed. When the CIS 11 performs only motion detection, the AP 12 may be notified that motion has been detected in the image frame in the same way as when a 2D code is detected.
  • a series of processes in the CIS 11 and AP 12 described above can be executed by hardware or by software.
  • a program that constitutes the software is installed in the computer.
  • the computer includes, for example, a computer built into dedicated hardware and a general-purpose personal computer capable of executing various functions by installing various programs.
  • the present technology can also take the following configurations.
  • an image recognition unit included in an image sensor the image recognition unit detecting that an image captured by the image sensor includes an image to be detected;
  • a processing unit that acquires an output image output from the image sensor and executes a predetermined process on the output image, wherein the image recognition unit detects that the detection target image is included and a processing unit that causes the image sensor to output the output image.
  • the information processing apparatus according to (1) wherein the image recognition unit uses a person as the detection target.
  • the processing unit is in a power saving state when the image recognition unit does not detect that the detection target image is included.
  • the information processing apparatus (13) The information processing apparatus according to (12), wherein the processing unit performs the process of preventing the peeping only when the plurality of people are gazing at the image on the display unit. (14) The information processing apparatus according to (2), wherein the processing unit transmits the output image to a communication partner of a video call. (15) The information processing apparatus according to (14), wherein the processing unit performs voice call processing when it is detected that the output image does not include an image of a person's face. (16) The image recognition unit sets the detection target to be an animal, The information processing apparatus according to (1), wherein the processing unit stores the output image in a storage unit.
  • an image recognition unit included in an image sensor the image recognition unit detecting whether or not an image captured by the image sensor includes a predetermined detection target image; and a processing unit that executes a predetermined process based on a result output from the image sensor and detected by the image recognition unit.
  • the processing unit stops reproducing the moving image when the image recognition unit does not detect that the image of the person is included.
  • the processing unit restarts reproduction of the moving image when the image recognition unit detects that the image of the person is included.
  • the processing unit adjusts the brightness of the screen of the display unit based on the brightness of the environment, and the image recognition unit detects the image of the person.
  • the image recognition unit includes image recognition processing using an inference model having a neural network structure.
  • the image recognition unit performs image recognition processing using an image having a resolution lower than that of the output image.
  • the information processing apparatus according to any one of (1) to (20), wherein the image sensor outputs the output image having a resolution lower than that of the viewing image to the processing unit.
  • the image recognition unit detects whether or not the image to be detected is included in the captured image when motion is detected in the captured image. information processing equipment.
  • the image recognition unit of an information processing apparatus comprising a processing unit detects that a predetermined detection target image is included in an image captured by the image sensor, obtaining an output image output from the image sensor, executing a predetermined process on the output image, and detecting that the image to be detected is included by the image recognition unit, An information processing method for outputting the output image.
  • an image recognition unit including a computer and an image sensor, the image recognition unit detecting that an image captured by the image sensor includes a predetermined image to be detected;
  • a processing unit that acquires an output image output from the image sensor and executes a predetermined process on the output image, wherein the image recognition unit detects that the detection target image is included a processing unit that causes the image sensor to output the output image.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • General Health & Medical Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Multimedia (AREA)
  • Electromagnetism (AREA)
  • Artificial Intelligence (AREA)
  • Toxicology (AREA)
  • Computer Security & Cryptography (AREA)
  • Software Systems (AREA)
  • Computer Hardware Design (AREA)
  • Evolutionary Computation (AREA)
  • Human Computer Interaction (AREA)
  • Oral & Maxillofacial Surgery (AREA)
  • General Engineering & Computer Science (AREA)
  • Computing Systems (AREA)
  • Databases & Information Systems (AREA)
  • Medical Informatics (AREA)
  • Vascular Medicine (AREA)
  • Signal Processing (AREA)
  • Studio Devices (AREA)
PCT/JP2022/009864 2021-08-31 2022-03-08 情報処理装置、情報処理方法、及び、プログラム Ceased WO2023032274A1 (ja)

Priority Applications (3)

Application Number Priority Date Filing Date Title
JP2023545031A JPWO2023032274A1 (https=) 2021-08-31 2022-03-08
CN202280057200.3A CN117859338A (zh) 2021-08-31 2022-03-08 信息处理装置、信息处理方法和程序
US18/291,939 US20240346844A1 (en) 2021-08-31 2022-03-08 Information processing apparatus, information processing method, and program

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
JP2021141251 2021-08-31
JP2021-141251 2021-08-31

Publications (1)

Publication Number Publication Date
WO2023032274A1 true WO2023032274A1 (ja) 2023-03-09

Family

ID=85412392

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/JP2022/009864 Ceased WO2023032274A1 (ja) 2021-08-31 2022-03-08 情報処理装置、情報処理方法、及び、プログラム

Country Status (4)

Country Link
US (1) US20240346844A1 (https=)
JP (1) JPWO2023032274A1 (https=)
CN (1) CN117859338A (https=)
WO (1) WO2023032274A1 (https=)

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117499569B (zh) * 2022-07-26 2025-10-03 中兴通讯股份有限公司 视频通话方法、电子设备和存储介质
JP7474888B1 (ja) * 2023-02-21 2024-04-25 レノボ・シンガポール・プライベート・リミテッド 電子機器、及び制御方法

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2006014291A (ja) * 2004-05-25 2006-01-12 Fuji Photo Film Co Ltd 撮影システムおよび撮影方法
JP2007325144A (ja) * 2006-06-05 2007-12-13 Fujifilm Corp 画像表示装置および方法並びにプログラム
JP2012129701A (ja) * 2010-12-14 2012-07-05 Nec Casio Mobile Communications Ltd 携帯機器、情報表示装置、プライバシー保護方法およびプライバシー保護プログラム
JP2012239156A (ja) * 2011-04-26 2012-12-06 Ricoh Co Ltd 撮像装置、撮像方法および撮像プログラム
JP2018160799A (ja) * 2017-03-23 2018-10-11 ソニーセミコンダクタソリューションズ株式会社 制御装置、制御方法、プログラム、及び、電子機器システム
JP2019202089A (ja) * 2018-05-25 2019-11-28 加賀デバイス株式会社 プレイ提供装置、プレイ提供システム、プレイ提供装置の制御方法、及びプレイ提供装置の制御プログラム

Family Cites Families (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US8824747B2 (en) * 2010-06-29 2014-09-02 Apple Inc. Skin-tone filtering
JP2017228975A (ja) * 2016-06-23 2017-12-28 ソニーセミコンダクタソリューションズ株式会社 画像処理装置、画像処理方法、イメージセンサ、情報処理装置、並びにプログラム
US11727718B2 (en) * 2018-09-26 2023-08-15 Apple Inc. Light recognition module for determining a user of a computing device
FR3100074B1 (fr) * 2019-08-21 2022-01-28 Idemia Identity & Security France Procédé d'analyse d'une caractéristique faciale d'un visage
KR20250003957A (ko) * 2022-05-26 2025-01-07 애플 인크. 사용자 인증 및 디바이스 관리를 위한 디바이스들, 방법들 및 그래픽 사용자 인터페이스들

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2006014291A (ja) * 2004-05-25 2006-01-12 Fuji Photo Film Co Ltd 撮影システムおよび撮影方法
JP2007325144A (ja) * 2006-06-05 2007-12-13 Fujifilm Corp 画像表示装置および方法並びにプログラム
JP2012129701A (ja) * 2010-12-14 2012-07-05 Nec Casio Mobile Communications Ltd 携帯機器、情報表示装置、プライバシー保護方法およびプライバシー保護プログラム
JP2012239156A (ja) * 2011-04-26 2012-12-06 Ricoh Co Ltd 撮像装置、撮像方法および撮像プログラム
JP2018160799A (ja) * 2017-03-23 2018-10-11 ソニーセミコンダクタソリューションズ株式会社 制御装置、制御方法、プログラム、及び、電子機器システム
JP2019202089A (ja) * 2018-05-25 2019-11-28 加賀デバイス株式会社 プレイ提供装置、プレイ提供システム、プレイ提供装置の制御方法、及びプレイ提供装置の制御プログラム

Also Published As

Publication number Publication date
US20240346844A1 (en) 2024-10-17
JPWO2023032274A1 (https=) 2023-03-09
CN117859338A (zh) 2024-04-09

Similar Documents

Publication Publication Date Title
US9836639B2 (en) Systems and methods of light modulation in eye tracking devices
US9330322B2 (en) Controlled access to functionality of a wireless device
CN115482568B (zh) 电子设备以及控制方法
US7742625B2 (en) Autonomous camera having exchangable behaviours
WO2019080797A1 (zh) 一种活体检测方法、终端和存储介质
WO2023032274A1 (ja) 情報処理装置、情報処理方法、及び、プログラム
WO2020103526A1 (zh) 拍照方法、装置、存储介质及终端设备
US11843760B2 (en) Timing mechanism to derive non-contaminated video stream using RGB-IR sensor with structured light
WO2021208986A1 (zh) 门开关状态监测方法、装置、设备及计算机可读存储介质
TWI735858B (zh) 適應性紅外線投影控制的方法以及裝置
US20220046194A1 (en) Driver mechanism for rolling shutter sensor to acquire structured light pattern
JP2023084910A (ja) 電子機器、及び制御方法
CN115346254B (zh) 电子设备以及控制方法
CN113225431B (zh) 显示屏的控制方法、装置及存储介质
CN111447497A (zh) 智能播放设备及其节能控制方法
JP2010273276A (ja) テレビ制御装置
CN114697611B (zh) 来访对讲控制方法、对讲控制装置、系统、电子设备及存储介质
US12506955B2 (en) Electronic apparatus for obtaining biometric information of companion animal, and operation method thereof
CN113591658B (zh) 一种基于距离感应的护眼系统
JP2019142125A (ja) 画像形成装置、その制御方法、およびプログラム
CN118626183A (zh) 一种屏幕状态控制方法和电子设备
CN120689952A (zh) 门锁控制方法、装置、门锁、存储介质及程序产品
JP2024033723A (ja) 撮像制御装置、プログラム及び撮像制御方法
KR20250015503A (ko) 반려 동물의 생체 인식을 위한 영상을 취득하기 위한 방법 및 전자 장치
CN106851403B (zh) 防止盗录播放画面的显示装置及内容安全播放方法

Legal Events

Date Code Title Description
WWE Wipo information: entry into national phase

Ref document number: 2023545031

Country of ref document: JP

WWE Wipo information: entry into national phase

Ref document number: 18291939

Country of ref document: US

WWE Wipo information: entry into national phase

Ref document number: 202280057200.3

Country of ref document: CN

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 22863864

Country of ref document: EP

Kind code of ref document: A1