WO2018133874A1 - 一种发送报警消息的方法和装置 - Google Patents

一种发送报警消息的方法和装置 Download PDF

Info

Publication number
WO2018133874A1
WO2018133874A1 PCT/CN2018/073775 CN2018073775W WO2018133874A1 WO 2018133874 A1 WO2018133874 A1 WO 2018133874A1 CN 2018073775 W CN2018073775 W CN 2018073775W WO 2018133874 A1 WO2018133874 A1 WO 2018133874A1
Authority
WO
WIPO (PCT)
Prior art keywords
area
image
head
shoulder
image area
Prior art date
Application number
PCT/CN2018/073775
Other languages
English (en)
French (fr)
Inventor
童昊浩
童俊艳
任烨
Original Assignee
杭州海康威视数字技术股份有限公司
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 杭州海康威视数字技术股份有限公司 filed Critical 杭州海康威视数字技术股份有限公司
Priority to US16/480,126 priority Critical patent/US11386698B2/en
Priority to EP18741996.5A priority patent/EP3572974B1/en
Publication of WO2018133874A1 publication Critical patent/WO2018133874A1/zh

Links

Images

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06VIMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V40/00Recognition of biometric, human-related or animal-related patterns in image or video data
    • G06V40/10Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
    • G06V40/16Human faces, e.g. facial parts, sketches or expressions
    • G06V40/161Detection; Localisation; Normalisation
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00Pattern recognition
    • G06F18/20Analysing
    • G06F18/21Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06VIMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00Scenes; Scene-specific elements
    • G06V20/50Context or environment of the image
    • G06V20/52Surveillance or monitoring of activities, e.g. for recognising suspicious objects
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06VIMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V40/00Recognition of biometric, human-related or animal-related patterns in image or video data
    • G06V40/20Movements or behaviour, e.g. gesture recognition
    • GPHYSICS
    • G08SIGNALLING
    • G08BSIGNALLING OR CALLING SYSTEMS; ORDER TELEGRAPHS; ALARM SYSTEMS
    • G08B21/00Alarms responsive to a single specified undesired or abnormal condition and not otherwise provided for
    • G08B21/18Status alarms
    • GPHYSICS
    • G08SIGNALLING
    • G08BSIGNALLING OR CALLING SYSTEMS; ORDER TELEGRAPHS; ALARM SYSTEMS
    • G08B21/00Alarms responsive to a single specified undesired or abnormal condition and not otherwise provided for
    • G08B21/02Alarms for ensuring the safety of persons
    • G08B21/0297Robbery alarms, e.g. hold-up alarms, bag snatching alarms
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04NPICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N23/00Cameras or camera modules comprising electronic image sensors; Control thereof
    • H04N23/60Control of cameras or camera modules
    • H04N23/61Control of cameras or camera modules based on recognised objects
    • H04N23/611Control of cameras or camera modules based on recognised objects where the recognised objects include parts of the human body

Definitions

  • the present application relates to the field of computer technologies, and in particular, to a method and apparatus for transmitting an alarm message.
  • Telecommunications fraud mainly refers to the use of telephone calls by mobile agents to users of mobile terminals, tricking them into ATM (Automatic Teller Machine) for financial transactions, and using mobile phones to control their operations, thereby seeking benefits.
  • ATM Automatic Teller Machine
  • the bank will be equipped with some security personnel around the ATM.
  • the security personnel will observe the users who handle the business on the ATM. If the security personnel judge that the user may be a user controlled by the fraudster, then The user alerts and intervenes in their transactions as necessary. In this way, the user can be prevented from being deceived to a certain extent, and the user's property is prevented from being lost.
  • the embodiment of the present application provides a method for sending an alarm message.
  • the technical solution is as follows:
  • a method of transmitting an alert message comprising:
  • the determining the target detection area in the detection image includes:
  • the target detection area is determined in the detected image based on the detection result of the face image area and the head and shoulder image area.
  • the determining the target detection area in the detection image according to the detection result of the face image area and the head and shoulder image area includes:
  • the face image area and the head and shoulder image area are detected, determining the target detection area in the detected image according to the positional relationship of the face image area, the head and shoulder image area, and the detection area stored in advance;
  • the detected face image area is enlarged, and the enlarged face image area is used as the target detection area;
  • the detected head and shoulder image area is subjected to reduction processing, and the reduced head and shoulder image area is used as the target detection area.
  • the method further includes:
  • the training samples including image samples, and a face image area and/or a head and shoulder image area in the image samples;
  • the method further includes:
  • the training samples including image samples, and character-to-phone state information corresponding to the image samples;
  • the detecting a face image area and a head and shoulder image area in the detected image based on a preset face and head and shoulder detection algorithm model including:
  • the face is The image pending area is used as a face image area
  • the head and shoulders are The image pending area serves as a head and shoulder image area.
  • the accuracy of determining the face image area and the head and shoulder image area can be improved.
  • the method further includes:
  • the method further includes:
  • the step of transmitting the first alert message to the server is performed.
  • the method further includes:
  • the second alarm message is sent to the server.
  • the alarm message can be sent according to the voice information, and multiple levels of alarms can be implemented, and the security personnel can perform different processing according to different alarm messages, such as paying attention or going to the site for viewing, thereby improving the user experience.
  • the pre-set face and head and shoulder detection algorithm model before detecting the face image area and the head and shoulder image area in the detection image, further includes:
  • the preset detection trigger condition is reached, performing a step of detecting a face image area and a head and shoulder image area in the detection image based on a preset face and head and shoulder detection algorithm model;
  • the preset detection trigger condition includes at least:
  • Character activity information is detected in the detected image.
  • the method further includes:
  • the method further includes:
  • the step of transmitting the first alarm message to the server is performed.
  • an apparatus for transmitting an alert message comprising:
  • a first acquiring module configured to acquire a detection image captured by the imaging device
  • a first determining module configured to determine a target detection area in the detection image
  • the first sending module is configured to detect, according to the preset call determination algorithm model, the person-to-phone status information corresponding to the image of the target detection area, and send the message to the server if the person's call status information is a call. An alarm message.
  • the device further includes:
  • a detecting module configured to detect a face image area and a head and shoulder image area in the detected image based on a preset face and head and shoulder detection algorithm model
  • the first determining module is configured to determine a target detection area in the detection image according to a detection result of a face image area and a head and shoulder image area.
  • the first determining module is configured to:
  • the face image area and the head and shoulder image area are detected, determining the target detection area in the detected image according to the positional relationship of the face image area, the head and shoulder image area, and the detection area stored in advance;
  • the detected face image area is enlarged, and the enlarged face image area is used as the target detection area;
  • the detected head and shoulder image area is subjected to reduction processing, and the reduced head and shoulder image area is used as the target detection area.
  • the device further includes:
  • a second acquiring module configured to acquire a plurality of pre-stored training samples, where the training samples include image samples, and a face image area and/or a head and shoulder image area in the image samples;
  • the first training module is configured to train the preset first initial algorithm model based on the plurality of training samples to obtain the face and head and shoulder detection algorithm model.
  • the device further includes:
  • a third acquiring module configured to acquire a plurality of pre-stored training samples, where the training samples include image samples, and the person-to-phone status information corresponding to the image samples;
  • the second training module is configured to train the preset second initial algorithm model based on the plurality of training samples to obtain the call determination algorithm model.
  • the detecting module includes:
  • a first determining submodule configured to determine a face image pending area and a head and shoulder image pending area in the detected image based on a preset face and head and shoulder detection algorithm model, and determine that the face image pending area corresponds to a confidence level corresponding to the head and shoulder image pending area;
  • a second determining submodule configured to determine a weight corresponding to the face image pending area and a weight corresponding to the head and shoulder image pending area according to the correspondence between the previously stored location information and the weight;
  • the second determining sub-module is further configured to: if the confidence level corresponding to the face image pending area is greater than a preset first confidence threshold, and the weight corresponding to the face image pending area is greater than a preset number a weight threshold, the face image pending area is used as a face image area;
  • the second determining submodule is further configured to: if the confidence level corresponding to the head and shoulder image pending area is greater than a preset second confidence threshold, and the weight corresponding to the head and shoulder image pending area is greater than a preset number
  • the second weight threshold is used as the head and shoulder image area.
  • the device further includes:
  • a second determining module configured to determine a currently detected face image area and/or a head and shoulder image area, similar to a face image area and/or a head and shoulder image area in a previous frame detection image that is closest to the current time degree;
  • the first sending module is configured to:
  • the step of transmitting the first alert message to the server is performed.
  • the device further includes:
  • a fourth acquiring module configured to acquire voice information detected by the voice input device
  • the second sending module is configured to send a second alarm message to the server if the voice information includes a preset keyword.
  • the device further includes:
  • a third determining module configured to perform a step of detecting a face image area and a head and shoulder image area in the detected image according to a preset face and head and shoulder detection algorithm model if a preset detection trigger condition is reached;
  • the preset detection trigger condition includes at least:
  • Character activity information is detected in the detected image.
  • the first determining module is further configured to:
  • the first sending module is further configured to:
  • the step of transmitting the first alarm message to the server is performed.
  • a computer readable storage medium having stored therein a computer program that, when executed by a processor, implements the method steps of the first aspect.
  • a terminal where the terminal includes:
  • One or more processors are One or more processors.
  • the memory stores one or more programs, the one or more programs configured to be executed by the one or more processors, the one or more programs comprising performing the method as described in the first aspect Instructions for method steps.
  • the terminal may determine the target detection area in the detection image, and detect the call status information of the person corresponding to the image of the target detection area according to the preset call determination algorithm model. If the character's phone status information is a call, the first alarm message is sent to the server, so that the user can be identified in time to make a call, and the alarm message can be sent in time to detect the call, so that the security is provided. The personnel are informed in time that there may be users who are deceived and processed to avoid loss of the user's property.
  • FIG. 1a is a system frame diagram provided by an embodiment of the present application.
  • FIG. 1b is a flowchart of a method for sending an alarm message according to an embodiment of the present application
  • FIG. 2 is a flowchart of a method for sending an alarm message according to an embodiment of the present application
  • 3a, 3b, and 3c are schematic diagrams of a detection result provided by an embodiment of the present application.
  • FIGS. 4a, 4b, and 4c are schematic diagrams of target detection areas provided by embodiments of the present application.
  • FIG. 5 is a schematic diagram of an interface display according to an embodiment of the present application.
  • FIG. 6 is a schematic structural diagram of an apparatus for sending an alarm message according to an embodiment of the present application.
  • FIG. 7 is a schematic structural diagram of an apparatus for sending an alarm message according to an embodiment of the present application.
  • FIG. 8 is a schematic structural diagram of an apparatus for sending an alarm message according to an embodiment of the present application.
  • FIG. 9 is a schematic structural diagram of an apparatus for sending an alarm message according to an embodiment of the present application.
  • FIG. 10 is a schematic structural diagram of an apparatus for sending an alarm message according to an embodiment of the present application.
  • FIG. 11 is a schematic structural diagram of an apparatus for sending an alarm message according to an embodiment of the present application.
  • FIG. 12 is a schematic structural diagram of an apparatus for sending an alarm message according to an embodiment of the present application.
  • FIG. 13 is a schematic structural diagram of an apparatus for sending an alarm message according to an embodiment of the present application.
  • FIG. 14 is a schematic structural diagram of a terminal according to an embodiment of the present application.
  • the embodiment of the present application provides a method for sending an alarm message, and the execution body of the method may be a terminal or a server.
  • the execution subject is taken as an example for description, and other cases are similar.
  • the terminal can be a terminal with data processing functions, such as a computer.
  • the terminal can be respectively connected to the camera device and the background server of an alarm platform.
  • the camera device can be installed on a device (such as an ATM) for handling funds, or can be installed around a device for handling funds.
  • the image capturing device can perform image capturing, and can transmit the captured image (ie, the detected image) to the terminal in real time. After receiving the detected image, the terminal can analyze the detected image to identify whether the person in the detected image is making a call.
  • the system framework diagram provided in this embodiment includes an imaging device, a terminal, and a server.
  • the terminal can include a transceiver, a processor, and a memory.
  • the transceiver may be configured to receive a detection image captured by the imaging device;
  • the processor may be a CPU (Central Processing Unit), etc., and may detect a face in the detected image based on a preset face and head and shoulder detection algorithm model.
  • the image area and the head and shoulder image area are then determined according to the detection result of the face image area and the head and shoulder image area, and the target detection area is determined in the detected image, and then the image corresponding to the target detection area is detected according to the preset call determination algorithm model.
  • the character communicates with the phone status information.
  • the transceiver can be controlled to send a first alarm message to the server;
  • the memory can be RAM (Random Access Memory), Flash (Flash), etc. It can be used to store received data, data required for processing, data generated during processing, and the like, such as detection images, face and head and shoulder detection algorithm models, and call determination algorithm models.
  • the terminal may also include components such as Bluetooth and power.
  • the device handling the capital service is taken as an example of the ATM, and the processing flow shown in FIG. 1b is described in detail, and the content is as follows:
  • Step 101 Acquire a detection image captured by the imaging device.
  • the image capturing device around the ATM can perform image capturing processing in the power-on state, and can transmit the captured image (ie, the detected image) to the terminal in real time, and the terminal can receive the detected image sent by the imaging device, and can The detected image is stored.
  • the imaging device may also send the identifier of the corresponding ATM to the terminal, and the terminal may store the detected image corresponding to the identifier of the ATM.
  • Step 102 determining a target detection area in the detected image.
  • the target detection region for determining the person's call state information may be determined in the acquired detection image.
  • Step 103 Detect a call-to-phone status information corresponding to the image of the target detection area according to the preset call determination algorithm model, and send a first alarm message to the server if the person's call status information is a call.
  • the terminal may store a preset call determination algorithm model, and the call determination algorithm model may be obtained through training. After determining the target detection area, the terminal may detect the call status information of the person corresponding to the image of the target detection area according to the preset call determination algorithm model. Among them, the character telephone status information may include a call and a no call.
  • the terminal may count the number of frames of the detected image of the call by the character in the preset duration, and if the number of frames is greater than the preset threshold, the terminal may send an alarm message (which may be referred to as a first alarm message), that is, If the person's call status information is a call, and the number of frames of the detected call state detected by the person within the preset time period is greater than a preset threshold, the alarm message may be sent to the server.
  • the terminal may also count the number of frames of the detected image of the call by the character in the preset duration, if the number of frames in the detected image acquired within the preset duration is greater than the total number of frames in the detected image.
  • the preset ratio threshold may be used to send a first alarm message to the server, that is, if the person's call status information is a call, and the number of frames detected by the person detected by the preset duration is the number of frames of the call. If the proportion of the total number of frames in the detected image acquired within the duration is greater than a preset ratio threshold, the first alarm message may be sent to the server.
  • the identifier of the ATM may be carried in the first alarm message.
  • the server may send a first alarm notification message carrying the identifier of the ATM to the terminal of the security personnel, and is used to remind the security personnel that the target device is being used, wherein the target device is the device corresponding to the identifier of the ATM. ) users pay attention.
  • the terminal may determine the target detection area in the detection image, and detect the call status information of the person corresponding to the image of the target detection area according to the preset call determination algorithm model. If the character's phone status information is a call, the first alarm message is sent to the server, so that the user can be identified in time to make a call, and the alarm message can be sent in time to detect the call, so that the security is provided. The personnel are informed in time that there may be users who are deceived and processed to avoid loss of the user's property.
  • the device handling the capital service is taken as an example of the ATM, and the processing flow shown in FIG. 2 is described in detail.
  • the content can be as follows:
  • Step 201 Acquire a detection image captured by the imaging device.
  • the image capturing device around the ATM can perform image capturing processing in the power-on state, and can transmit the captured image (ie, the detected image) to the terminal in real time, and the terminal can receive the detected image sent by the imaging device, and can The detected image is stored.
  • the imaging device may also send the identifier of the corresponding ATM to the terminal, and the terminal may store the detected image corresponding to the identifier of the ATM.
  • Step 202 Detect a face image area and a head and shoulder image area in the detected image based on the preset face and head and shoulder detection algorithm models.
  • a preset face and head and shoulder detection algorithm model may be stored in the terminal, and the face and head and shoulder detection algorithm model may be obtained by training in advance.
  • the terminal acquires the detection image
  • the face image area and the head and shoulder image area may be detected in the acquired detection image.
  • the detected image may be a face image.
  • the area (which may be referred to as a face image pending area), and may be an area of the head and shoulders image (which may be referred to as a head and shoulder image pending area), and may determine a confidence level corresponding to the face image pending area, and a head and shoulder image pending area Corresponding confidence.
  • the confidence can be used to reflect the probability that the detected image is a face image (or a head and shoulder image).
  • the terminal may also store a first confidence threshold corresponding to the face image area and a second confidence threshold corresponding to the head and shoulder image area.
  • the first confidence threshold may be the same as the second confidence threshold or may be different from the second confidence threshold.
  • the terminal may separately set the confidence level of the face image pending area and the first Comparing a confidence threshold for comparing the confidence level corresponding to the head and shoulder image pending area with the second confidence threshold, if the terminal determines that the confidence level of the detected face image pending area is greater than or equal to the first confidence threshold, Then, the face image to-be-determined area can be used as the face image area; otherwise, it is determined that the face image pending area is not the face image area.
  • the terminal determines that the confidence level of the detected head and shoulder image pending area is greater than or equal to the second confidence threshold, the head and shoulder image pending area may be used as the head and shoulder image area. Otherwise, the head and shoulder image pending area is determined not to be the head. Shoulder image area.
  • the face and head and shoulder detection algorithm model may have errors. Therefore, the detection result may be filtered according to the detected position information of the face image area and/or the head and shoulder image area, and the corresponding processing may be performed. As follows: based on the preset face and head and shoulder detection algorithm model, determining the face image pending area and the head and shoulder image pending area in the detected image, and determining the confidence corresponding to the face image pending area and the head and shoulder image pending area Confidence; according to the correspondence between the pre-stored position information and the weight, the weight corresponding to the face image pending area and the weight corresponding to the head and shoulder image pending area; if the face image pending area corresponds to a greater confidence than the pre- If the first confidence threshold is set, and the weight corresponding to the face image pending area is greater than the preset first weight threshold, the face image pending area is used as the face image area; if the head and shoulder image pending area corresponds to the confidence The degree is greater than the preset second confidence threshold, and
  • the terminal may determine a face image pending area and a head and shoulder image pending area in the detected image based on the preset face and head and shoulder detection algorithm model, and determine a confidence and a head and shoulder corresponding to the face image pending area.
  • the confidence level corresponding to the image pending area is similar to the above, and will not be described again.
  • the terminal may determine the position information of the face image pending area and the position information of the head and shoulder image pending area based on the face and head and shoulder detection algorithm model.
  • the correspondence between the location information and the weight may be pre-stored in the terminal, and the correspondence may include the correspondence between the location information of the face image pending area and the weight, and the correspondence between the location information of the head and shoulder image pending area and the weight. .
  • the terminal may obtain the weight corresponding to the detected position information of the face image pending area from the above correspondence, and the detected The weight corresponding to the position information of the head and shoulder image pending area.
  • the weight may be used to reflect the importance of the detected face image pending area (or head and shoulder image pending area).
  • the terminal may also store a first weight threshold corresponding to the face image area and a second weight threshold corresponding to the head and shoulder image area.
  • the first weight threshold may be the same as the second weight threshold or may be different from the second weight threshold.
  • the terminal may compare the confidence level corresponding to the face image pending area with the first confidence threshold, if the terminal determines If the confidence level of the detected face image pending area is less than the first confidence threshold, it may be determined that the face image pending area is not a face image area, and if the terminal determines that the detected face image pending area corresponds to a confidence greater than Or equal to the first confidence threshold, the terminal may further compare the weight corresponding to the face image pending area with the first weight threshold, if the weight corresponding to the face image pending area is greater than or equal to the first weight threshold
  • the face image to-be-determined area may be used as the face image area; otherwise, the face image pending area may be determined not to be the face image area.
  • the terminal may compare the confidence level corresponding to the head-shoulder image pending area with the second confidence threshold, if the terminal determines that the detection is If the confidence level of the head and shoulder image pending area is less than the second confidence threshold, it may be determined that the head and shoulder image pending area is not the head and shoulder image area, and if the terminal determines that the detected head and shoulder image pending area corresponds to a confidence greater than or equal to The second confidence threshold, the terminal may further compare the weight corresponding to the head and shoulder image pending area with the second weight threshold.
  • the head and shoulder image pending area may be used as the head and shoulder image area, otherwise, the head and shoulder image pending area is determined not to be the head and shoulder image area. In this way, the situation of false detection can be effectively avoided, and the accuracy of detecting the face image area and the head and shoulder image area can be improved.
  • the terminal may first detect whether there is a person around the ATM.
  • the terminal may perform the processing of step 202 when a person is detected, and the corresponding processing may be as follows: if the preset detection is reached.
  • the trigger condition is performed by detecting a face image area and a head and shoulder image area in the detected image based on a preset face and head and shoulder detection algorithm model.
  • the preset detection trigger conditions can be varied. This embodiment provides several feasible ways of processing:
  • the terminal may acquire a detection image captured by an imaging device of an ATM, and then may establish a corresponding background model according to the detected image, and the terminal may also periodically update the background model according to the detected image.
  • the terminal may compare the detection image with the background model to determine the foreground image, and may binarize the foreground image to generate a binary foreground image, and determine the detection image according to the binary foreground image. Whether there is a character in it.
  • the terminal determines whether there is a person around the ATM by determining the size of the motion area in the detected image, in addition to determining whether there is a person in the detected image according to the foreground image.
  • the terminal may acquire the previous frame detection image of the currently acquired detection image, and further, may calculate the adjacent two frames according to the gray value of each pixel in the adjacent two frames.
  • the difference in the gray value of the pixel at the same position in the image may be called the degree of difference.
  • the terminal may determine, in each pixel, a pixel point whose degree of difference is greater than a preset threshold, to obtain a motion area of the detected image. If the foreground image is detected in the detected image within the preset duration, and the motion area in the detected image is greater than a preset threshold, it may be determined that there is a person around the ATM.
  • the terminal may perform the process of step 202 when receiving the operation notification message sent by the target device.
  • the target device may be a device for handling a fund service, such as an ATM.
  • the target device when the user operates in the target device, the target device may detect the instruction input by the user, and then may send an operation notification message to the terminal, where the operation notification message may carry the device identifier of the target device, and correspondingly, the terminal may Receive an operation notification message sent by the target device. After obtaining the detection image, the terminal may determine whether an operation notification message sent by the target device is received. If the operation notification message is received, the terminal may perform the process of step 202.
  • the terminal may perform the processing of step 202 when receiving the object detection notification sent by the sensing device.
  • a sensor such as an infrared sensor
  • the sensor can detect the corresponding detection signal, and then can send an object detection notification to the terminal, and the object detection notification can be The device identifier carrying the target device is received, and correspondingly, the terminal can receive the object detection notification sent by the sensing device. After obtaining the detection image, the terminal may determine whether the object detection notification sent by the sensing device is received. If the terminal receives the object detection notification, the terminal may perform the processing of step 202.
  • the face and head and shoulder detection algorithm model may be pre-trained.
  • the process of training the face and head and shoulder detection algorithm model may be as follows: acquiring a plurality of pre-stored training samples, and the training samples include images. The sample, and the face image area and/or the head and shoulder image area in the image sample; the preset first initial algorithm model is trained based on the plurality of training samples to obtain a face and head and shoulder detection algorithm model.
  • the terminal can train face and head and shoulder detectors using a high performance deep learning network architecture (such as faster-rcnn) and an efficient ZF (Matthew D. Zeiler and Rob Fergus) improved network model.
  • Training samples can include training positive samples and training negative samples.
  • the training positive sample may include an image sample, and a face image area and/or a head and shoulder image area in the image sample, and the image sample may be a face image and a head and shoulder image at different angles; a face image area and/or The head and shoulder image area may be represented by coordinate information of the face image area and/or the head and shoulder image area.
  • the training negative samples may include image samples, as well as areas of the image samples that are not human faces and head and shoulders.
  • the terminal may train the preset first initial algorithm model based on a plurality of training samples and a preset training algorithm to obtain a face and head and shoulder detection algorithm model.
  • the target detection rate and the detection rate of the face and head and shoulder detection algorithm models trained based on deep learning techniques have been greatly improved.
  • Step 203 Determine a target detection area in the detected image according to the detection result of the face image area and the head and shoulder image area.
  • the terminal may detect the face image area and the head and shoulder image area, as shown in FIG. 3a, or may only detect the face image. Area, or only the head and shoulder image area is detected, as shown in Figures 3b and 3c. After detecting the face image area and the head and shoulder image area by the terminal, the terminal may determine the target detection area in the detected image based on the obtained detection result.
  • the manner in which the terminal determines the target detection area in the detection image is also different.
  • the processing of step 203 may be as follows: if the face image area and the head and shoulder image area are detected, according to Pre-stored positional relationship between the face image area, the head and shoulder image area, and the detection area, the target detection area is determined in the detection image; if the face image area is detected, the head and shoulder image area is not detected, the detected face is detected The image area is enlarged, and the enlarged face image area is used as the target detection area; if the head and shoulder image area is detected and the face image area is not detected, the detected head and shoulder image area is reduced and reduced. The processed head and shoulder image area is used as the target detection area.
  • the terminal detects the face image area and the head and shoulder image area
  • the positional relationship between the face image area, the head and shoulder image area, and the detection area stored in advance, and the determined face image area and head may be determined.
  • the shoulder image area determines the target detection area in the detected image. For example, the position of the human eye in the face image area and the neck position in the head and shoulder image area can be determined, and then the target detection area can be determined in the area below the human eye position and above the neck position, as shown in FIG. 4a.
  • the target detection area may be a rectangular area.
  • the face image area may be enlarged in the detected image according to the preset magnification factor, and the enlarged face image area is targeted.
  • the detection area wherein the magnification factor corresponding to the width of the face image area may be larger than the magnification factor corresponding to the length of the face image area, as shown in FIG. 4b. That is, the target detection area is an area including the face image area in the detected image.
  • the detected head and shoulder image area may be reduced in the detected image according to the preset reduction factor, and the processed head and shoulder image will be reduced.
  • the area is used as the target detection area, wherein the reduction coefficient corresponding to the width of the head and shoulder image area may be smaller than the reduction coefficient corresponding to the length of the face image area, as shown in FIG. 4c.
  • the terminal may not perform processing. That is, the target detection area is an area in the head and shoulder image area.
  • Step 204 Detect the person-to-phone state information corresponding to the image of the target detection area according to the preset call determination algorithm model, and send the first alarm message to the server if the person's call status information is a call.
  • the terminal may store a preset call determination algorithm model, and the call determination algorithm model may be obtained through training. After determining the target detection area, the terminal may detect the call status information of the person corresponding to the image of the target detection area according to the preset call determination algorithm model. Among them, the character telephone status information may include a call and a no call.
  • the terminal may count the number of frames of the detected image of the call by the character in the preset duration, and if the number of frames is greater than the preset threshold, the terminal may send an alarm message (which may be referred to as a first alarm message), that is, If the person's call status information is a call, and the number of frames of the detected call state detected by the person within the preset time period is greater than a preset threshold, the alarm message may be sent to the server.
  • the terminal may also count the number of frames of the detected image of the call by the character in the preset duration, if the number of frames in the detected image acquired within the preset duration is greater than the total number of frames in the detected image.
  • the preset ratio threshold may be used to send a first alarm message to the server, that is, if the person's call status information is a call, and the number of frames detected by the person detected by the preset duration is the number of frames of the call. If the proportion of the total number of frames in the detected image acquired within the duration is greater than a preset ratio threshold, the first alarm message may be sent to the server.
  • the identifier of the ATM may be carried in the first alarm message.
  • the server may send a first alarm notification message carrying the identifier of the ATM to the terminal of the security personnel, and is used to remind the security personnel that the target device is being used, wherein the target device is the device corresponding to the identifier of the ATM. ) users pay attention.
  • the terminal may further add a marker image to the detected image according to the detected face image area or the head and shoulder image area for delineating the person in the detected image.
  • the terminal may send the detected image with the tag image to the server, and the server may send the detected image with the tag image to the security personnel's terminal, so that the security personnel can find the corresponding
  • the user as shown in FIG. 5, that is, the first alarm message may also carry the detection image with the marker image added.
  • the first alarm notification message may also carry the detection image with the marker image added.
  • the terminal can also output the preset anti-fraud voice information through the voice broadcast device while sending the first alarm message, and promptly remind the user to prevent being deceived.
  • the processing procedure of the terminal training call determination algorithm model may be as follows: acquiring a plurality of pre-stored training samples, where the training samples include image samples, and the person-to-phone state information corresponding to the image samples; based on the plurality of training samples, The preset second initial algorithm model is trained to obtain a call-by-call determination algorithm model.
  • the terminal may employ a deep learning network improved model (such as googlenet-loss1) to train the call-by-call decision algorithm model.
  • Training samples can include training positive samples and training negative samples.
  • the image sample in the training positive sample may be an image of a person including a call, and the image sample may be an image of a different gesture, such as a left-handed call, a right-handed call, a backhand call, a face and a shoulder clip, etc.
  • the different postures, that is, the person-to-phone status information corresponding to the training positive sample may be a call.
  • the training negative sample may be an image of a person who does not include the telephone call, that is, the person-to-phone status information corresponding to the training negative sample may be a missed call.
  • a second initial algorithm model may be preset in the terminal, where the second initial algorithm model may include a parameter to be determined, and the terminal may perform training on the preset second initial algorithm model according to the plurality of pre-stored training samples.
  • the training value of the parameter to be determined that is, the call determination algorithm model can be obtained.
  • the detection accuracy of the call-based decision algorithm model based on the training of deep learning technology is greatly improved compared with the traditional method (such as SVM), and can eliminate the false detection caused by interference of hair and eyes.
  • the voice recognition technology may be combined to identify a user that may be spoofed.
  • the corresponding processing may be as follows: acquiring voice information detected by the voice input device; if the voice information includes a preset keyword, sending a second message to the server. Alarm message.
  • the voice input device can be installed on the ATM or can be installed around the ATM.
  • the voice input device can detect the corresponding voice information, and can send the detected voice information to the terminal. Accordingly, the terminal can receive the voice input device. voice message.
  • the terminal may determine whether the voice information includes a preset keyword, such as a card number, a transfer, or the like, by using a voice recognition algorithm stored in advance.
  • the speech recognition algorithm may be trained by the terminal using a speech recognition algorithm model in the related art.
  • the terminal can model sensitive vocabulary (ie, keywords) frequently found in telecommunication fraud according to the MFCC voiceprint feature and the HMM hidden Markov model in the related art, and can extract the voiceprint feature of the sound file in the database, and then The HMM model corresponding to each keyword can be established, and then the MFCC voiceprint feature can be used to train each HMM model to obtain a speech recognition algorithm.
  • sensitive vocabulary ie, keywords
  • HMM hidden Markov model in the related art
  • the terminal may send an alarm message (which may be referred to as a second alarm message).
  • the server may send a second alarm notification message carrying the target device identifier to the terminal of the security personnel, for alerting the security personnel to perform transaction intervention on the user who is using the target device, and preventing the user from simultaneously Fraud transfer.
  • the terminal may continuously track the detected characters to avoid repeatedly sending the alarm message, and the corresponding processing may be as follows: determining the currently detected face image area and/or the head and shoulder image area, and the current time.
  • the closest previous frame detects the similarity of the face image area and/or the head and shoulder image area in the image; if the similarity does not satisfy the preset similarity condition, the step of transmitting the first alarm message to the server is performed.
  • the terminal may determine the currently detected face image area and/or the head and shoulder image area, and the last frame detection image that is closest to the current time. The similarity between the face image area and/or the head and shoulder image area. Specifically, after detecting the face image area and/or the head and shoulder image area, the terminal may generate a target frame corresponding to the face image area and/or the head and shoulder image area. Taking the target frame corresponding to the face image area as an example, after determining the target frame, the terminal may determine the attribute information of the target frame, wherein the attribute information may include the gray value of the pixel in the target frame and the size information of the target frame.
  • the location information of the target frame, and further, the attribute information of the target frame in the currently received detection image may be compared with the attribute information of the target frame in the previous frame detection image closest to the current time to determine the two Similarity, the similarity can be used as the similarity between the currently detected face image region and the face image region in the previous frame detection image closest to the current time.
  • the average value of the difference of the pixel points in the target frame can be calculated to obtain the gray level difference degree; the size ratio value can be calculated to obtain the size difference degree; and the position difference degree can also be calculated according to the coordinate information. If the gray level difference is less than the preset first difference degree threshold, and the size difference degree is less than the preset second difference degree threshold, and the position difference degree is less than the preset third difference degree threshold, the similarity between the two may be determined. The degree is high, otherwise, it can be determined that the similarity between the two is low.
  • the terminal After determining the similarity, if the terminal determines that the similarity is low, it indicates that the user in the current detected image may be different from the user in the previous frame detection image, and the terminal may perform the step of sending the first alarm message to the server, that is, After detecting the person-to-phone state information corresponding to the image of the target detection area, if the character-to-phone state information is a call, and the determined similarity does not satisfy the preset similarity condition, then the first alarm message is sent to the server. A step of. If the terminal determines that the similarity is high, it indicates that the user in the current detected image may be the same as the user in the previous frame image. Based on the above processing, the terminal has sent an alarm message corresponding to the user to the server, so there is no need to send the server to the server again. Alarm messages, thus avoiding repeated alarms for the same user.
  • the target frame corresponding to the face image area can be preferentially determined. If the face image area is not detected, the target frame corresponding to the head and shoulder image area can be used for judgment.
  • the specific processing is similar to the above. Narration.
  • the above marked image may be a target frame image generated by the terminal, and if the terminal detects the face image region and/or the head and shoulder image region in the previous frame detection image, but not in the current detected image.
  • the face image area and/or the head and shoulder image area are detected, and the target frame image corresponding to the previous frame detection image may be added to the current detection image for display by the security personnel terminal, if the terminal is in a continuous preset If the face image area and/or the head and shoulder image area are not detected in the number frame detection image, the addition of the target frame image to the detected image may be stopped.
  • there are detection errors in the face and head and shoulder detection algorithm models which may result in the detection results not being continuous.
  • the display of the target frame image is not continuous, so that the target frame in the detection image seen by the security personnel is
  • the image may sometimes be absent or even flickering, and based on the processing, the continuity of the image of the display target frame can be improved, and the user experience can be effectively improved.
  • the corresponding processing may be as follows: determining the currently detected face image area and/or the head and shoulder image area, and acquiring the preset time length. Each frame detects each similarity of the face image area and/or the head and shoulder image area in the image; if each of the obtained similarities does not satisfy the preset similarity condition, the step of transmitting the first alarm message to the server is performed.
  • the terminal may acquire the face image area in each frame detection image acquired within the preset duration and/or The head and shoulder image area, and further, the face image area and/or the head and shoulder image area in the currently acquired detection image may be determined according to the above manner, and the face image in each frame detection image acquired within the preset time length
  • the terminal may perform the step of sending the first alarm message to the server, that is, after detecting the person's call status information corresponding to the image of the target detection area, if the person's call status information is a call, and the determination is made.
  • the step of transmitting the first alarm message to the server is performed. If there is a similarity in the obtained similarity conditions that satisfies the preset similarity condition (in this case, the user in the current detected image may be the same user as another user in the other frame detection images acquired within the preset duration) Then, based on the above processing, the terminal has sent an alarm message corresponding to the user to the server, and therefore, it is not necessary to send an alarm message to the server again, thereby avoiding the situation of repeated alarms for the same user.
  • the preset similarity condition in this case, the user in the current detected image may be the same user as another user in the other frame detection images acquired within the preset duration
  • the terminal may determine the target detection area in the detection image, and detect the call status information of the person corresponding to the image of the target detection area according to the preset call determination algorithm model. If the character's phone status information is a call, the first alarm message is sent to the server, so that the user can be identified in time to make a call, and the alarm message can be sent in time to detect the call, so that the security is provided. The personnel are informed in time that there may be users who are deceived and processed to avoid loss of the user's property.
  • the embodiment of the present application further provides an apparatus for sending an alarm message.
  • the apparatus includes:
  • the first obtaining module 610 is configured to acquire a detection image captured by the imaging device
  • a first determining module 620 configured to determine a target detection area in the detection image
  • the first sending module 630 is configured to detect, according to the preset call determination algorithm model, the person-to-phone status information corresponding to the image of the target detection area, and send the status information to the server if the person's call status information is a call.
  • the first alarm message is configured to notify the user's call status information a call.
  • the apparatus further includes:
  • the detecting module 640 is configured to detect a face image area and a head and shoulder image area in the detected image based on a preset face and head and shoulder detection algorithm model;
  • the first determining module 620 is configured to determine a target detection area in the detection image according to the detection result of the face image area and the head and shoulder image area.
  • the first determining module 620 is configured to:
  • the face image area and the head and shoulder image area are detected, determining the target detection area in the detected image according to the positional relationship of the face image area, the head and shoulder image area, and the detection area stored in advance;
  • the detected face image area is enlarged, and the enlarged face image area is used as the target detection area;
  • the detected head and shoulder image area is subjected to reduction processing, and the reduced head and shoulder image area is used as the target detection area.
  • the apparatus further includes:
  • a second acquiring module 650 configured to acquire a plurality of training samples stored in advance, where the training samples include image samples, and a face image area and/or a head and shoulder image area in the image samples;
  • the first training module 660 is configured to train the preset first initial algorithm model based on the plurality of training samples to obtain the face and head and shoulder detection algorithm model.
  • the apparatus further includes:
  • the third obtaining module 670 is configured to acquire a plurality of training samples stored in advance, where the training samples include image samples, and the person-to-phone state information corresponding to the image samples;
  • the second training module 680 is configured to train the preset second initial algorithm model based on the plurality of training samples to obtain the call determination algorithm model.
  • the detecting module 640 includes:
  • a first determining sub-module 641 configured to determine a face image pending area and a head-and-shoulder image pending area in the detected image based on a preset face and head and shoulder detection algorithm model, and determine the face image pending area Corresponding confidence and a confidence level corresponding to the head and shoulder image pending area;
  • a second determining sub-module 642 configured to determine, according to a pre-stored correspondence between the location information and the weight, a weight corresponding to the face image pending area, and a weight corresponding to the head and shoulder image pending area;
  • the second determining sub-module 642 is further configured to: if the confidence level corresponding to the face image pending area is greater than a preset first confidence threshold, and the weight corresponding to the face image pending area is greater than a preset a first weight threshold, the face image pending area is used as a face image area;
  • the second determining sub-module 642 is further configured to: if the confidence level corresponding to the head-shoulder image pending area is greater than a preset second confidence threshold, and the weight corresponding to the head-shoulder image pending area is greater than a preset The second weight threshold is used as the head and shoulder image area.
  • the apparatus further includes:
  • a second determining module 690 configured to determine a currently detected face image area and/or a head and shoulder image area, the face image area and/or the head and shoulder image area in the previous frame detection image that is closest to the current time Similarity
  • the first sending module 630 is configured to:
  • the step of transmitting the first alert message to the server is performed.
  • the device further includes:
  • the fourth obtaining module 6100 is configured to acquire voice information detected by the voice input device.
  • the second sending module 6110 is configured to send a second alarm message to the server if the voice information includes a preset keyword.
  • the apparatus further includes:
  • the third determining module 6120 is configured to: if the preset detection trigger condition is reached, perform a step of detecting a face image area and a head and shoulder image area in the detection image based on a preset face and head and shoulder detection algorithm model ;
  • the preset detection trigger condition includes at least:
  • Character activity information is detected in the detected image.
  • the first determining module 620 is further configured to:
  • the first sending module 630 is further configured to:
  • the step of transmitting the first alarm message to the server is performed.
  • the terminal may determine the target detection area in the detection image, and detect the call status information of the person corresponding to the image of the target detection area according to the preset call determination algorithm model. If the character's phone status information is a call, the first alarm message is sent to the server, so that the user can be identified in time to make a call, and the alarm message can be sent in time to detect the call, so that the security is provided. The personnel are informed in time that there may be users who are deceived and processed to avoid loss of the user's property.
  • the device when the alarm message is sent by the foregoing embodiment, the device only exemplifies the division of each functional module. In actual applications, the function may be allocated by different functional modules according to requirements. Completion, that is, the internal structure of the device is divided into different functional modules to complete all or part of the functions described above.
  • the device for sending an alarm message provided by the foregoing embodiment is the same as the method for transmitting the alarm message. For the specific implementation process, refer to the method embodiment, and details are not described herein again.
  • FIG. 14 is a schematic structural diagram of a terminal involved in the embodiment of the present application.
  • the terminal may be used to implement the method for sending an alarm message provided in the foregoing embodiment. Specifically:
  • the terminal 900 may include an RF (Radio Frequency) circuit 110, a memory 120 including one or more computer readable storage media, an input unit 130, a display unit 140, a sensor 150, an audio circuit 160, and a WiFi (wireless fidelity, wireless).
  • the fidelity module 170 includes a processor 180 having one or more processing cores, and a power supply 190 and the like. It will be understood by those skilled in the art that the terminal structure shown in FIG. 14 does not constitute a limitation to the terminal, and may include more or less components than those illustrated, or a combination of certain components, or different component arrangements. among them:
  • the RF circuit 110 can be used for transmitting and receiving information or during a call, and receiving and transmitting signals. Specifically, after receiving downlink information of the base station, the downlink information is processed by one or more processors 180. In addition, the data related to the uplink is sent to the base station. .
  • the RF circuit 110 includes, but is not limited to, an antenna, at least one amplifier, a tuner, one or more oscillators, a Subscriber Identity Module (SIM) card, a transceiver, a coupler, an LNA (Low Noise Amplifier). , duplexer, etc.
  • RF circuitry 110 can also communicate with the network and other devices via wireless communication.
  • the wireless communication may use any communication standard or protocol, including but not limited to GSM (Global System of Mobile communication), GPRS (General Packet Radio Service), CDMA (Code Division Multiple Access). , Code Division Multiple Access), WCDMA (Wideband Code Division Multiple Access), LTE (Long Term Evolution), e-mail, SMS (Short Messaging Service), and the like.
  • GSM Global System of Mobile communication
  • GPRS General Packet Radio Service
  • CDMA Code Division Multiple Access
  • WCDMA Wideband Code Division Multiple Access
  • LTE Long Term Evolution
  • e-mail Short Messaging Service
  • the memory 120 can be used to store software programs and modules, and the processor 180 executes various functional applications and data processing by running software programs and modules stored in the memory 120.
  • the memory 120 may mainly include a storage program area and a storage data area, wherein the storage program area may store an operating system, an application required for at least one function (such as a sound playing function, an image playing function, etc.), and the like; the storage data area may be stored according to The data created by the use of the terminal 900 (such as audio data, phone book, etc.) and the like.
  • memory 120 can include high speed random access memory, and can also include non-volatile memory, such as at least one magnetic disk storage device, flash memory device, or other volatile solid state storage device. Accordingly, memory 120 may also include a memory controller to provide access to memory 120 by processor 180 and input unit 130.
  • the input unit 130 can be configured to receive input numeric or character information and to generate keyboard, mouse, joystick, optical or trackball signal inputs related to user settings and function controls.
  • input unit 130 can include touch-sensitive surface 131 as well as other input devices 132.
  • Touch-sensitive surface 131 also referred to as a touch display or trackpad, can collect touch operations on or near the user (such as a user using a finger, stylus, etc., on any suitable object or accessory on touch-sensitive surface 131 or The operation near the touch-sensitive surface 131) and driving the corresponding connecting device according to a preset program.
  • the touch-sensitive surface 131 can include two portions of a touch detection device and a touch controller.
  • the touch detection device detects the touch orientation of the user, and detects a signal brought by the touch operation, and transmits the signal to the touch controller; the touch controller receives the touch information from the touch detection device, converts the touch information into contact coordinates, and sends the touch information.
  • the processor 180 is provided and can receive commands from the processor 180 and execute them.
  • the touch-sensitive surface 131 can be implemented in various types such as resistive, capacitive, infrared, and surface acoustic waves.
  • the input unit 130 can also include other input devices 132.
  • other input devices 132 may include, but are not limited to, one or more of a physical keyboard, function keys (such as volume control buttons, switch buttons, etc.), trackballs, mice, joysticks, and the like.
  • the display unit 140 can be used to display information entered by the user or information provided to the user and various graphical user interfaces of the terminal 900, which can be composed of graphics, text, icons, video, and any combination thereof.
  • the display unit 140 may include a display panel 141.
  • the display panel 141 may be configured in the form of an LCD (Liquid Crystal Display), an OLED (Organic Light-Emitting Diode), or the like.
  • the touch-sensitive surface 131 may cover the display panel 141, and when the touch-sensitive surface 131 detects a touch operation thereon or nearby, it is transmitted to the processor 180 to determine the type of the touch event, and then the processor 180 according to the touch event The type provides a corresponding visual output on display panel 141.
  • touch-sensitive surface 131 and display panel 141 are implemented as two separate components to implement input and input functions, in some embodiments, touch-sensitive surface 131 can be integrated with display panel 141 for input. And output function.
  • Terminal 900 can also include at least one type of sensor 150, such as a light sensor, motion sensor, and other sensors.
  • the light sensor may include an ambient light sensor and a proximity sensor, wherein the ambient light sensor may adjust the brightness of the display panel 141 according to the brightness of the ambient light, and the proximity sensor may close the display panel 141 when the terminal 900 moves to the ear. / or backlight.
  • the gravity acceleration sensor can detect the magnitude of acceleration in all directions (usually three axes). When it is stationary, it can detect the magnitude and direction of gravity.
  • the terminal 900 can also be configured with gyroscopes, barometers, hygrometers, thermometers, infrared sensors and other sensors, here Let me repeat.
  • the audio circuit 160, the speaker 161, and the microphone 162 can provide an audio interface between the user and the terminal 900.
  • the audio circuit 160 can transmit the converted electrical data of the received audio data to the speaker 161 for conversion to the sound signal output by the speaker 161; on the other hand, the microphone 162 converts the collected sound signal into an electrical signal by the audio circuit 160. After receiving, it is converted into audio data, and then processed by the audio data output processor 180, transmitted to the terminal, for example, via the RF circuit 110, or outputted to the memory 120 for further processing.
  • the audio circuit 160 may also include an earbud jack to provide communication of the peripheral earphones with the terminal 900.
  • WiFi is a short-range wireless transmission technology
  • the terminal 900 can help users to send and receive emails, browse web pages, and access streaming media through the WiFi module 170, which provides wireless broadband Internet access for users.
  • FIG. 14 shows the WiFi module 170, it can be understood that it does not belong to the essential configuration of the terminal 900, and may be omitted as needed within the scope of not changing the essence of the invention.
  • the processor 180 is a control center of the terminal 900 that connects various portions of the entire handset using various interfaces and lines, by running or executing software programs and/or modules stored in the memory 120, and recalling data stored in the memory 120, The various functions and processing data of the terminal 900 are performed to perform overall monitoring of the mobile phone.
  • the processor 180 may include one or more processing cores; preferably, the processor 180 may integrate an application processor and a modem processor, where the application processor mainly processes an operating system, a user interface, an application, and the like.
  • the modem processor primarily handles wireless communications. It can be understood that the above modem processor may not be integrated into the processor 180.
  • the terminal 900 also includes a power source 190 (such as a battery) that supplies power to the various components.
  • the power source can be logically coupled to the processor 180 through a power management system to manage functions such as charging, discharging, and power management through the power management system.
  • Power supply 190 may also include any one or more of a DC or AC power source, a recharging system, a power failure detection circuit, a power converter or inverter, a power status indicator, and the like.
  • the terminal 900 may further include a camera, a Bluetooth module, and the like, and details are not described herein again.
  • the display unit of the terminal 900 is a touch screen display
  • the terminal 900 further includes a memory, and one or more programs, wherein one or more programs are stored in the memory and configured to be one or one
  • the above processor executes instructions including a method for the terminal to execute the above-described method of transmitting an alarm message.
  • the terminal may determine the target detection area in the detection image, and detect the call status information of the person corresponding to the image of the target detection area according to the preset call determination algorithm model. If the character's phone status information is a call, the first alarm message is sent to the server, so that the user can be identified in time to make a call, and the alarm message can be sent in time to detect the call, so that the security is provided. The personnel are informed in time that there may be users who are deceived and processed to avoid loss of the user's property.
  • a person skilled in the art may understand that all or part of the steps of implementing the above embodiments may be completed by hardware, or may be instructed by a program to execute related hardware, and the program may be stored in a computer readable storage medium.
  • the storage medium mentioned may be a read only memory, a magnetic disk or an optical disk or the like.

Landscapes

  • Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Multimedia (AREA)
  • Health & Medical Sciences (AREA)
  • Human Computer Interaction (AREA)
  • General Health & Medical Sciences (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Oral & Maxillofacial Surgery (AREA)
  • Social Psychology (AREA)
  • Psychiatry (AREA)
  • Data Mining & Analysis (AREA)
  • Emergency Management (AREA)
  • Business, Economics & Management (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Signal Processing (AREA)
  • General Engineering & Computer Science (AREA)
  • Evolutionary Computation (AREA)
  • Evolutionary Biology (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Artificial Intelligence (AREA)
  • Image Analysis (AREA)

Abstract

一种发送报警消息的方法和装置,属于计算机技术领域。所述方法包括:获取摄像设备拍摄的检测图像(201);基于预设的人脸和头肩检测算法模型,在所述检测图像中检测人脸图像区域和头肩图像区域(202);根据人脸图像区域和头肩图像区域的检测结果,在所述检测图像中确定目标检测区域(203);根据预设的通电话判定算法模型,检测所述目标检测区域的图像对应的人物通电话状态信息,如果所述人物通电话状态信息为通电话,则向服务器发送第一报警消息(204)。采用该方法,可以使安保人员及时获知可能存在受到欺骗的用户,并进行处理,避免用户的财产受到损失。

Description

一种发送报警消息的方法和装置
本申请要求于2017年01月23日提交国家知识产权局、申请号为201710050732.5、发明名称为“一种发送报警消息的方法和装置”的中国专利申请的优先权,其全部内容通过引用结合在本申请中。
技术领域
本申请涉及计算机技术领域,特别涉及一种发送报警消息的方法和装置。
背景技术
随着金融、通信行业的快速发展,电信诈骗在中国发展蔓延,出现频率越来越高。电信诈骗主要指犯罪分子利用电话致电移动终端的使用者,诱骗其至ATM(Automatic Teller Machine,自动出纳机)进行金融交易,并利用移动电话控制其操作,从而谋取利益。
为了避免电信诈骗给人们造成损失,目前,银行在ATM周围都会配备一些安保人员,安保人员会观察在ATM上办理业务的用户,如果安保人员判断该用户可能是被诈骗分子控制的用户,则对该用户进行提醒,并在必要时对其交易进行干涉。这样,可以在一定程度上阻止用户上当受骗,避免用户的财产受到损失。
在实现本申请的过程中,发明人发现相关技术至少存在以下问题:
安保人员的观察能力有限,尤其是在银行中的人较多时,无法注意到每一个受到欺骗的用户,导致用户的财产受到损失。
发明内容
为了解决相关技术的问题,本申请实施例提供了一种发送报警消息的方法。所述技术方案如下:
第一方面,提供了一种发送报警消息的方法,所述方法包括:
获取摄像设备拍摄的检测图像;
在所述检测图像中确定目标检测区域;
根据预设的通电话判定算法模型,检测所述目标检测区域的图像对应的人 物通电话状态信息,如果所述人物通电话状态信息为通电话,则向服务器发送第一报警消息。
可选的,所述在所述检测图像中确定目标检测区域,包括:
基于预设的人脸和头肩检测算法模型,在所述检测图像中检测人脸图像区域和头肩图像区域;
根据人脸图像区域和头肩图像区域的检测结果,在所述检测图像中确定目标检测区域。
可选的,所述根据人脸图像区域和头肩图像区域的检测结果,在所述检测图像中确定目标检测区域,包括:
如果检测到人脸图像区域和头肩图像区域,则根据预先存储的人脸图像区域、头肩图像区域和检测区域的位置关系,在所述检测图像中确定目标检测区域;
如果检测到人脸图像区域,未检测到头肩图像区域,则对检测到的人脸图像区域进行放大处理,将放大处理后的人脸图像区域作为目标检测区域;
如果检测到头肩图像区域,未检测到人脸图像区域,则对检测到的头肩图像区域进行缩小处理,将缩小处理后的头肩图像区域作为目标检测区域。
这样,提供了在不同情况下确定目标检测区域的实现方式,提高了目标检测区域的检出率。
可选的,所述方法还包括:
获取预先存储的多个训练样本,所述训练样本包括图像样本,及所述图像样本中的人脸图像区域和/或头肩图像区域;
基于所述多个训练样本,对预设的第一初始算法模型进行训练,得到所述人脸和头肩检测算法模型。
这样,提供了一种训练人脸和头肩检测算法模型的实现方式。
可选的,所述方法还包括:
获取预先存储的多个训练样本,所述训练样本包括图像样本,及所述图像样本对应的人物通电话状态信息;
基于所述多个训练样本,对预设的第二初始算法模型进行训练,得到所述通电话判定算法模型。
这样,提供了一种训练通电话判定算法模型的实现方式。
可选的,所述基于预设的人脸和头肩检测算法模型,在所述检测图像中检 测人脸图像区域和头肩图像区域,包括:
基于预设的人脸和头肩检测算法模型,在所述检测图像中确定人脸图像待定区域和头肩图像待定区域,并确定所述人脸图像待定区域对应的置信度和所述头肩图像待定区域对应的置信度;
根据预先存储的位置信息和权重的对应关系,确定所述人脸图像待定区域对应的权值,以及所述头肩图像待定区域对应的权值;
如果所述人脸图像待定区域对应的置信度大于预设的第一置信度阈值,且所述人脸图像待定区域对应的权值大于预设的第一权值阈值,则将所述人脸图像待定区域作为人脸图像区域;
如果所述头肩图像待定区域对应的置信度大于预设的第二置信度阈值,且所述头肩图像待定区域对应的权值大于预设的第二权值阈值,则将所述头肩图像待定区域作为头肩图像区域。
这样,结合位置信息来确定人脸图像区域和头肩图像区域,可以提高确定人脸图像区域和头肩图像区域的准确度。
可选的,所述方法还包括:
确定当前检测到的人脸图像区域和/或头肩图像区域,与当前时间最接近的上一帧检测图像中的人脸图像区域和/或头肩图像区域的相似度;
所述向服务器发送第一报警消息之前,还包括:
如果所述相似度不满足预设相似度条件,则执行向服务器发送第一报警消息的步骤。
这样,可以避免对同一对象重复发送报警消息。
可选的,所述方法还包括:
获取语音输入设备检测到的语音信息;
如果所述语音信息包含预设的关键词,则向所述服务器发送第二报警消息。
这样,可以根据语音信息来发送报警消息,可以实现多级报警,安保人员则可以根据不同的报警消息进行不同的处理,如关注或去现场查看等,从而可以提高用户体验。
可选的,所述基于预设的人脸和头肩检测算法模型,在所述检测图像中检测人脸图像区域和头肩图像区域之前,还包括:
如果达到预设的检测触发条件,则执行基于预设的人脸和头肩检测算法模 型,在所述检测图像中检测人脸图像区域和头肩图像区域的步骤;
所述预设的检测触发条件至少包括:
在所述检测图像中检测到人物活动信息;或者,
接收到目标设备发送的操作通知消息;或者,
接收到传感设备发送的对象检测通知。
可选的,所述方法还包括:
确定当前检测到的人脸图像区域和/或头肩图像区域,与预设时长内获取到的各帧检测图像中的人脸图像区域和/或头肩图像区域的各相似度;
所述向服务器发送第一报警消息之前,还包括:
如果得到的各相似度均不满足预设相似度条件,则执行向服务器发送第一报警消息的步骤。
第二方面,提供了一种发送报警消息的装置,所述装置包括:
第一获取模块,用于获取摄像设备拍摄的检测图像;
第一确定模块,用于在所述检测图像中确定目标检测区域;
第一发送模块,用于根据预设的通电话判定算法模型,检测所述目标检测区域的图像对应的人物通电话状态信息,如果所述人物通电话状态信息为通电话,则向服务器发送第一报警消息。
可选的,所述装置还包括:
检测模块,用于基于预设的人脸和头肩检测算法模型,在所述检测图像中检测人脸图像区域和头肩图像区域;
所述第一确定模块,用于:根据人脸图像区域和头肩图像区域的检测结果,在所述检测图像中确定目标检测区域。
可选的,所述第一确定模块,用于:
如果检测到人脸图像区域和头肩图像区域,则根据预先存储的人脸图像区域、头肩图像区域和检测区域的位置关系,在所述检测图像中确定目标检测区域;
如果检测到人脸图像区域,未检测到头肩图像区域,则对检测到的人脸图像区域进行放大处理,将放大处理后的人脸图像区域作为目标检测区域;
如果检测到头肩图像区域,未检测到人脸图像区域,则对检测到的头肩图像区域进行缩小处理,将缩小处理后的头肩图像区域作为目标检测区域。
可选的,所述装置还包括:
第二获取模块,用于获取预先存储的多个训练样本,所述训练样本包括图像样本,及所述图像样本中的人脸图像区域和/或头肩图像区域;
第一训练模块,用于基于所述多个训练样本,对预设的第一初始算法模型进行训练,得到所述人脸和头肩检测算法模型。
可选的,所述装置还包括:
第三获取模块,用于获取预先存储的多个训练样本,所述训练样本包括图像样本,及所述图像样本对应的人物通电话状态信息;
第二训练模块,用于基于所述多个训练样本,对预设的第二初始算法模型进行训练,得到所述通电话判定算法模型。
可选的,所述检测模块,包括:
第一确定子模块,用于基于预设的人脸和头肩检测算法模型,在所述检测图像中确定人脸图像待定区域和头肩图像待定区域,并确定所述人脸图像待定区域对应的置信度和所述头肩图像待定区域对应的置信度;
第二确定子模块,用于根据预先存储的位置信息和权重的对应关系,确定所述人脸图像待定区域对应的权值,以及所述头肩图像待定区域对应的权值;
所述第二确定子模块,还用于如果所述人脸图像待定区域对应的置信度大于预设的第一置信度阈值,且所述人脸图像待定区域对应的权值大于预设的第一权值阈值,则将所述人脸图像待定区域作为人脸图像区域;
所述第二确定子模块,还用于如果所述头肩图像待定区域对应的置信度大于预设的第二置信度阈值,且所述头肩图像待定区域对应的权值大于预设的第二权值阈值,则将所述头肩图像待定区域作为头肩图像区域。
可选的,所述装置还包括:
第二确定模块,用于确定当前检测到的人脸图像区域和/或头肩图像区域,与当前时间最接近的上一帧检测图像中的人脸图像区域和/或头肩图像区域的相似度;
所述第一发送模块,用于:
如果所述相似度不满足预设相似度条件,则执行向服务器发送第一报警消息的步骤。
可选的,所述装置还包括:
第四获取模块,用于获取语音输入设备检测到的语音信息;
第二发送模块,用于如果所述语音信息包含预设的关键词,则向所述服务器发送第二报警消息。
可选的,所述装置还包括:
第三确定模块,用于如果达到预设的检测触发条件,则执行基于预设的人脸和头肩检测算法模型,在所述检测图像中检测人脸图像区域和头肩图像区域的步骤;
所述预设的检测触发条件至少包括:
在所述检测图像中检测到人物活动信息;或者,
接收到目标设备发送的操作通知消息;或者,
接收到传感设备发送的对象检测通知。
可选的,所述第一确定模块,还用于:
确定当前检测到的人脸图像区域和/或头肩图像区域,与预设时长内获取到的各帧检测图像中的人脸图像区域和/或头肩图像区域的各相似度;
所述第一发送模块,还用于:
如果得到的各相似度均不满足预设相似度条件,则执行向服务器发送第一报警消息的步骤。
第三方面,提供了一种计算机可读存储介质,所述存储介质内存储有计算机程序,所述计算机程序被处理器执行时实现第一方面所述的方法步骤。
第四方面,提供了一种终端,所述终端包括:
一个或多个处理器;和
存储器;
所述存储器存储有一个或多个程序,所述一个或多个程序被配置成由所述一个或多个处理器执行,所述一个或多个程序包含用于进行如第一方面所述的方法步骤的指令。
本申请实施例提供的技术方案带来的有益效果是:
本申请实施例中,终端获取到摄像设备拍摄的检测图像后,可以在检测图像中确定目标检测区域,根据预设的通电话判定算法模型,检测目标检测区域的图像对应的人物通电话状态信息,如果人物通电话状态信息为通电话,则向服务器发送第一报警消息,这样,可以及时的识别出用户是否在打电话,并可 以在检测到打电话时,及时发送报警消息,以使安保人员及时获知可能存在受到欺骗的用户,并进行处理,避免用户的财产受到损失。
附图说明
为了更清楚地说明本申请实施例中的技术方案,下面将对实施例描述中所需要使用的附图作简单地介绍,显而易见地,下面描述中的附图仅仅是本申请的一些实施例,对于本领域普通技术人员来讲,在不付出创造性劳动的前提下,还可以根据这些附图获得其他的附图。
图1a是本申请实施例提供的一种系统框架图;
图1b是本申请实施例提供的一种发送报警消息的方法流程图;
图2是本申请实施例提供的一种发送报警消息的方法流程图;
图3a、图3b、图3c是本申请实施例提供的一种检测结果的示意图;
图4a、图4b、图4c是本申请实施例提供的目标检测区域的示意图;
图5是本申请实施例提供的一种界面显示示意图;
图6是本申请实施例提供的一种发送报警消息的装置结构示意图;
图7是本申请实施例提供的一种发送报警消息的装置结构示意图;
图8是本申请实施例提供的一种发送报警消息的装置结构示意图;
图9是本申请实施例提供的一种发送报警消息的装置结构示意图;
图10是本申请实施例提供的一种发送报警消息的装置结构示意图;
图11是本申请实施例提供的一种发送报警消息的装置结构示意图;
图12是本申请实施例提供的一种发送报警消息的装置结构示意图;
图13是本申请实施例提供的一种发送报警消息的装置结构示意图;
图14是本申请实施例提供的一种终端的结构示意图。
具体实施方式
为使本申请的目的、技术方案和优点更加清楚,下面将结合附图对本申请实施方式作进一步地详细描述。
本申请实施例提供了一种发送报警消息的方法,该方法的执行主体可以是终端,也可以是服务器。本实施例以执行主体为终端为例进行说明,其他情况与之类似。该终端可以是具有数据处理功能的终端,如计算机。该终端可以分别与摄像设备和某报警平台的后台服务器连接。其中,摄像设备可以安装在用 于办理资金业务的设备(如ATM)上,也可以安装在用于办理资金业务的设备周围。摄像设备可以进行图像拍摄,并可以将拍摄到的图像(即检测图像)实时发送给终端,终端接收到检测图像后,可以对检测图像进行分析,从而识别检测图像中的人物是否在打电话,如果是,则可以向报警平台的服务器发送报警消息。服务器则在接收到报警消息后,向安保人员的终端(如监控终端)发送警报提示信息,以提示安保人员进行处理,避免用户的财产受到损失。如图1a所示,为本实施例提供的系统框架图,其中包括摄像设备、终端和服务器。
终端可以包括收发器、处理器和存储器。收发器可以用于接收摄像设备拍摄的检测图像;处理器可以为CPU(Central Processing Unit,中央处理单元)等,可以基于预设的人脸和头肩检测算法模型,在检测图像中检测人脸图像区域和头肩图像区域,然后根据人脸图像区域和头肩图像区域的检测结果,在检测图像中确定目标检测区域,进而根据预设的通电话判定算法模型,检测目标检测区域的图像对应的人物通电话状态信息,如果人物通电话状态信息为通电话,则可以控制收发器向服务器发送第一报警消息;存储器可以为RAM(Random Access Memory,随机存取存储器)、Flash(闪存)等,可以用于存储接收到的数据、处理过程所需的数据、处理过程中生成的数据等,如检测图像、人脸和头肩检测算法模型、以及通电话判定算法模型等。另外,该终端还可以包括蓝牙和电源等部件。
下面将结合具体实施方式,以办理资金业务的设备为ATM为例,对图1b所示的处理流程进行详细的说明,内容如下:
步骤101,获取摄像设备拍摄的检测图像。
在实施中,ATM周围的摄像设备在开机状态下,可以执行图像拍摄处理,并可以将拍摄到的图像(即检测图像)实时发送给终端,终端则可以接收摄像设备发送的检测图像,并可以对检测图像进行存储。另外,摄像设备还可以将其对应的ATM的标识发送给终端,终端则可以将检测图像与该ATM的标识进行对应的存储。
步骤102,在检测图像中确定目标检测区域。
在实施中,终端获取到摄像设备拍摄的检测图像后,可以在获取到的检测图像中,确定用于确定人物通电话状态信息的目标检测区域。
步骤103,根据预设的通电话判定算法模型,检测目标检测区域的图像对 应的人物通电话状态信息,如果人物通电话状态信息为通电话,则向服务器发送第一报警消息。
在实施中,终端中可以存储有预设的通电话判定算法模型,该通电话判定算法模型可以是通过训练得到的。终端确定出目标检测区域后,可以根据预设的通电话判定算法模型,检测目标检测区域的图像对应的人物通电话状态信息。其中,人物通电话状态信息可以包括通电话和不通电话。终端可以统计预设时长内人物通电话状态信息为通电话的检测图像的帧数,如果该帧数大于预设阈值,则可以向服务器发送报警消息(可称为第一报警消息),也即如果人物通电话状态信息为通电话、且预设时长内检测到的人物通电话状态为通电话的检测图像的帧数大于预设阈值,则可以向服务器发送报警消息。或者,终端也可以统计预设时长内人物通电话状态信息为通电话的检测图像的帧数,如果该帧数在预设时长内获取到的检测图像中的总帧数中所占的比例大于预设比例阈值,则可以向服务器发送第一报警消息,也即如果人物通电话状态信息为通电话、且预设时长内检测到的人物通电话状态为通电话的检测图像的帧数在预设时长内获取到的检测图像中的总帧数中所占的比例大于预设比例阈值,则可以向服务器发送第一报警消息。
第一报警消息中可以携带有ATM的标识。服务器接收到第一报警消息后,可以向安保人员的终端发送携带有ATM的标识的第一警报通知消息,用于提醒安保人员对正在使用目标设备(其中,目标设备为ATM的标识对应的设备)的用户进行关注。
本申请实施例中,终端获取到摄像设备拍摄的检测图像后,可以在检测图像中确定目标检测区域,根据预设的通电话判定算法模型,检测目标检测区域的图像对应的人物通电话状态信息,如果人物通电话状态信息为通电话,则向服务器发送第一报警消息,这样,可以及时的识别出用户是否在打电话,并可以在检测到打电话时,及时发送报警消息,以使安保人员及时获知可能存在受到欺骗的用户,并进行处理,避免用户的财产受到损失。
下面将结合具体实施方式,以办理资金业务的设备为ATM为例,对图2所示的处理流程进行详细的说明,内容可以如下:
步骤201,获取摄像设备拍摄的检测图像。
在实施中,ATM周围的摄像设备在开机状态下,可以执行图像拍摄处理, 并可以将拍摄到的图像(即检测图像)实时发送给终端,终端则可以接收摄像设备发送的检测图像,并可以对检测图像进行存储。另外,摄像设备还可以将其对应的ATM的标识发送给终端,终端则可以将检测图像与该ATM的标识进行对应的存储。
步骤202,基于预设的人脸和头肩检测算法模型,在检测图像中检测人脸图像区域和头肩图像区域。
在实施中,终端中可以存储预设的人脸和头肩检测算法模型,该人脸和头肩检测算法模型可以是预先通过训练得到的。终端获取到检测图像后,可以在获取到的检测图像中检测人脸图像区域和头肩图像区域。具体的,可以根据检测图像中各像素点的像素值(如red通道值、green通道值和blue通道值),以及人脸和头肩检测算法模型,在检测图像中,确定可能是人脸图像的区域(可称为人脸图像待定区域),以及可能是头肩图像的区域(可称为头肩图像待定区域),并可以确定人脸图像待定区域对应的置信度,以及头肩图像待定区域对应的置信度。其中,置信度可以用于反映检测到的图像为人脸图像(或头肩图像)的概率。终端中还可以存储有人脸图像区域对应的第一置信度阈值,以及头肩图像区域对应的第二置信度阈值。第一置信度阈值可以与第二置信度阈值相同,也可以与第二置信度阈值不同。
确定出人脸图像待定区域对应的置信度、头肩图像待定区域对应的置信度、第一置信度阈值和第二置信度阈值后,终端可以分别将人脸图像待定区域对应的置信度与第一置信度阈值进行比较,将头肩图像待定区域对应的置信度与第二置信度阈值进行比较,如果终端判定检测到的人脸图像待定区域对应的置信度大于或等于第一置信度阈值,则可以将该人脸图像待定区域作为人脸图像区域,否则,判定该人脸图像待定区域不是人脸图像区域。如果终端判定检测到的头肩图像待定区域对应的置信度大于或等于第二置信度阈值,则可以将该头肩图像待定区域作为头肩图像区域,否则,判定该头肩图像待定区域不是头肩图像区域。
可选的,人脸和头肩检测算法模型可能会存在误差,因此,可以根据检测到的人脸图像区域和/或头肩图像区域的位置信息,对检测结果进行筛选,相应的处理过程可以如下:基于预设的人脸和头肩检测算法模型,在检测图像中确定人脸图像待定区域和头肩图像待定区域,并确定人脸图像待定区域对应的置信度和头肩图像待定区域对应的置信度;根据预先存储的位置信息和权重的对 应关系,确定人脸图像待定区域对应的权值,以及头肩图像待定区域对应的权值;如果人脸图像待定区域对应的置信度大于预设的第一置信度阈值,且人脸图像待定区域对应的权值大于预设的第一权值阈值,则将人脸图像待定区域作为人脸图像区域;如果头肩图像待定区域对应的置信度大于预设的第二置信度阈值,且头肩图像待定区域对应的权值大于预设的第二权值阈值,则将头肩图像待定区域作为头肩图像区域。
在实施中,终端可以基于预设的人脸和头肩检测算法模型,在检测图像中确定人脸图像待定区域和头肩图像待定区域,并确定人脸图像待定区域对应的置信度和头肩图像待定区域对应的置信度,具体过程与上述类似,不再赘述。确定出人脸图像待定区域和头肩图像待定区域后,终端可以基于人脸和头肩检测算法模型,确定人脸图像待定区域的位置信息和头肩图像待定区域的位置信息。终端中可以预先存储有位置信息与权值的对应关系,该对应关系可以包括人脸图像待定区域的位置信息与权值的对应关系,以及头肩图像待定区域的位置信息与权值的对应关系。确定出人脸图像待定区域的位置信息和头肩图像待定区域的位置信息后,终端可以从上述对应关系中,获取检测到的人脸图像待定区域的位置信息对应的权值,以及检测到的头肩图像待定区域的位置信息对应的权值。其中,权值可以用于反映检测到的人脸图像待定区域(或头肩图像待定区域)的重要程度。其中,如果人脸图像待定区域(或头肩图像待定区域)的位置信息为检测图像中的中间位置,则对应的权值较高,如果人脸图像待定区域(或头肩图像待定区域)的位置信息为检测图像中的边缘位置,则对应的权值较低。终端中还可以存储有人脸图像区域对应的第一权值阈值,以及头肩图像区域对应的第二权值阈值。第一权值阈值可以与第二权值阈值相同,也可以与第二权值阈值不同。
对于人脸图像待定区域,确定出人脸图像待定区域对应的置信度、第一置信度阈值后,终端可以将人脸图像待定区域对应的置信度与第一置信度阈值进行比较,如果终端判定检测到的人脸图像待定区域对应的置信度小于第一置信度阈值,则可以判定该人脸图像待定区域不是人脸图像区域,如果终端判定检测到的人脸图像待定区域对应的置信度大于或等于第一置信度阈值,则终端可以进一步将人脸图像待定区域对应的权值与第一权值阈值进行比较,如果该人脸图像待定区域对应的权值大于或等于第一权值阈值,则可将该人脸图像待定区域作为人脸图像区域;否则,可以判定该人脸图像待定区域不是人脸图像区 域。对于头肩待定区域,确定出头肩图像待定区域对应的置信度、第二置信度阈值后,终端可以将头肩图像待定区域对应的置信度与第二置信度阈值进行比较,如果终端判定检测到的头肩图像待定区域对应的置信度小于第二置信度阈值,则可以判定该头肩图像待定区域不是头肩图像区域,如果终端判定检测到的头肩图像待定区域对应的置信度大于或等于第二置信度阈值,则终端可以进一步将头肩图像待定区域对应的权值与第二权值阈值进行比较,如果该头肩图像待定区域对应的权值大于或等于第二权值阈值,则可以将该头肩图像待定区域作为头肩图像区域,否则,判定该头肩图像待定区域不是头肩图像区域。这样,可以有效的避免误检的情况,提高检测人脸图像区域和头肩图像区域的准确度。
可选的,获取到检测图像后,终端首先可以检测ATM周围是否有人,此种情况下,终端可以在检测到有人时执行步骤202的处理,相应的处理过程可以如下:如果达到预设的检测触发条件,则执行基于预设的人脸和头肩检测算法模型,在所述检测图像中检测人脸图像区域和头肩图像区域的步骤。
在实施中,预设的检测触发条件可以是多种多样的。本实施例提供了几种可行的处理方式:
方式一、终端可以在检测图像中检测到人物活动信息时,执行步骤202的处理。
在实施中,终端可以获取某ATM的摄像设备拍摄到的检测图像,然后可以根据检测图像建立对应的背景模型,并且,终端还可以根据检测图像,周期性的更新背景模型。后续终端每获取到一帧检测图像,终端可以将检测图像与背景模型进行对比,确定前景图像,并可以对前景图像进行二值化,生成二值前景图像,根据二值前景图像,判断检测图像中是否存在人物。另外,终端除了根据前景图像判断检测图像中是否存在人物,还可以通过判断检测图像中的运动区域的大小判断ATM周围是否有人。具体的,获取到该检测图像后,终端可以获取当前获取的检测图像的上一帧检测图像,进而,可以根据相邻两帧检测图像中各像素点的灰度值,计算相邻两帧检测图像中,相同位置的像素点的灰度值的差值(可称为差异度)。得到各像素点的差异度后,终端可以在各像素点中,确定差异度大于预设阈值的像素点,得到该检测图像的运动区域。如果终端在预设时长内的检测图像中,均检测到前景图像,且检测图像中的运动区域均大于预设阈值,则可以判定ATM周围有人。
方式二、终端可以在接收到目标设备发送的操作通知消息时,执行步骤202的处理。
其中,目标设备可以是用于办理资金业务的设备,如ATM。
在实施中,用户在目标设备中进行操作时,目标设备可以检测到用户输入的指令,然后可以向终端发送操作通知消息,操作通知消息中可以携带有目标设备的设备标识,相应的,终端可以接收目标设备发送的操作通知消息。终端获取到检测图像后,可以判断是否接收到目标设备发送的操作通知消息,如果接收到该操作通知消息,则终端可以执行步骤202的处理。
方式三、终端可以在接收到传感设备发送的对象检测通知时,执行步骤202的处理。
在实施中,ATM周围,或ATM中可以配置有传感器(如红外传感器),当用户靠近ATM时,该传感器可以检测到相应的检测信号,然后可以向终端发送对象检测通知,对象检测通知中可以携带有目标设备的设备标识,相应的,终端可以接收传感设备发送的对象检测通知。终端获取到检测图像后,可以判断是否接收到传感设备发送的对象检测通知,如果终端接收到该对象检测通知后,则终端可以执行步骤202的处理。
可选的,人脸和头肩检测算法模型可以是预先训练得到的,相应的,训练人脸和头肩检测算法模型的处理过程可以如下:获取预先存储的多个训练样本,训练样本包括图像样本,及图像样本中的人脸图像区域和/或头肩图像区域;基于多个训练样本,对预设的第一初始算法模型进行训练,得到人脸和头肩检测算法模型。
在实施中,终端可以采用高性能的深度学习网络架构(如faster-rcnn)和高效的ZF(Matthew D.Zeiler and Rob Fergus)改进型网络模型来训练人脸和头肩检测器。训练样本可以包括训练正样本和训练负样本。其中,训练正样本可以包括图像样本,及图像样本中的人脸图像区域和/或头肩图像区域,图像样本可以是不同角度下的人脸图像和头肩图像;人脸图像区域和/或头肩图像区域可以是用人脸图像区域和/或头肩图像区域的坐标信息表示的。训练负样本可以包括图像样本,以及图像样本中非人脸和头肩的区域。终端可以基于多个训练样本,以及预设的训练算法,对预设的第一初始算法模型进行训练,得到人脸和头肩检测算法模型。基于深度学习技术训练出的人脸和头肩检测算法模型的目标检出率和检准率均有大幅提升。
步骤203,根据人脸图像区域和头肩图像区域的检测结果,在检测图像中确定目标检测区域。
在实施中,终端在检测图像中进行人脸图像区域和头肩图像区域的检测后,可能检测到人脸图像区域和头肩图像区域,如图3a所示,也可能只检测到人脸图像区域,或只检测到头肩图像区域,如图3b、图3c所示。终端对人脸图像区域和头肩图像区域进行检测后,终端可以基于得到的检测结果,确定检测图像中的目标检测区域。
可选的,基于检测结果的不同,终端在检测图像中确定目标检测区域的方式也不同,相应的,步骤203的处理过程可以如下:如果检测到人脸图像区域和头肩图像区域,则根据预先存储的人脸图像区域、头肩图像区域和检测区域的位置关系,在检测图像中确定目标检测区域;如果检测到人脸图像区域,未检测到头肩图像区域,则对检测到的人脸图像区域进行放大处理,将放大处理后的人脸图像区域作为目标检测区域;如果检测到头肩图像区域,未检测到人脸图像区域,则对检测到的头肩图像区域进行缩小处理,将缩小处理后的头肩图像区域作为目标检测区域。
在实施中,如果终端检测到人脸图像区域和头肩图像区域,则可以根据预先存储的人脸图像区域、头肩图像区域和检测区域的位置关系,以及确定出的人脸图像区域、头肩图像区域,在检测图像中确定目标检测区域。例如,可以确定人脸图像区域中的人眼位置,以及头肩图像区域中的脖颈位置,然后可以在人眼位置以下、脖颈位置以上的区域中,确定目标检测区域,如图4a所示,目标检测区域可以为长方形区域。
如果终端只检测到人脸图像区域,未检测到头肩图像区域,则可以根据预设的放大系数,在检测图像中对人脸图像区域进行放大处理,将放大处理后的人脸图像区域作为目标检测区域,其中,人脸图像区域的宽度对应的放大系数,可以大于人脸图像区域的长度对应的放大系数,如图4b所示。也就是说,目标检测区域是检测图像中的包含人脸图像区域的区域。
如果终端只检测到头肩图像区域,未检测到人脸图像区域,则可以根据预设的缩小系数,在检测图像中对检测到的头肩图像区域进行缩小处理,将缩小处理后的头肩图像区域作为目标检测区域,其中,头肩图像区域的宽度对应的缩小系数,可以小于人脸图像区域的长度对应的缩小系数,如图4c所示。另外,如果未检测到人脸图像区域和头肩图像区域,则终端可以不进行处理。也 就是说,目标检测区域是头肩图像区域中的区域。
步骤204,根据预设的通电话判定算法模型,检测目标检测区域的图像对应的人物通电话状态信息,如果人物通电话状态信息为通电话,则向服务器发送第一报警消息。
在实施中,终端中可以存储有预设的通电话判定算法模型,该通电话判定算法模型可以是通过训练得到的。终端确定目标检测区域后,可以根据预设的通电话判定算法模型,检测目标检测区域的图像对应的人物通电话状态信息。其中,人物通电话状态信息可以包括通电话和不通电话。终端可以统计预设时长内人物通电话状态信息为通电话的检测图像的帧数,如果该帧数大于预设阈值,则可以向服务器发送报警消息(可称为第一报警消息),也即如果人物通电话状态信息为通电话、且预设时长内检测到的人物通电话状态为通电话的检测图像的帧数大于预设阈值,则可以向服务器发送报警消息。或者,终端也可以统计预设时长内人物通电话状态信息为通电话的检测图像的帧数,如果该帧数在预设时长内获取到的检测图像中的总帧数中所占的比例大于预设比例阈值,则可以向服务器发送第一报警消息,也即如果人物通电话状态信息为通电话、且预设时长内检测到的人物通电话状态为通电话的检测图像的帧数在预设时长内获取到的检测图像中的总帧数中所占的比例大于预设比例阈值,则可以向服务器发送第一报警消息。
第一报警消息中可以携带有ATM的标识。服务器接收到第一报警消息后,可以向安保人员的终端发送携带有ATM的标识的第一警报通知消息,用于提醒安保人员对正在使用目标设备(其中,目标设备为ATM的标识对应的设备)的用户进行关注。终端还可以根据检测到的人脸图像区域或头肩图像区域,在检测图像中添加标记图像,用于圈定检测图像中的人物。如果人物通电话状态信息为通电话,则终端可以将添加有标记图像的检测图像发送给服务器,服务器则可以将添加有标记图像的检测图像发送给安保人员的终端,以便于安保人员查找相应的用户,如图5所示,也就是说,第一报警消息中还可以携带有添加有标记图像的检测图像,相应的,第一警报通知消息中还可以携带有添加有标记图像的检测图像。另外,终端在发送第一报警消息的同时,还可以通过语音播报设备,输出预设的防诈骗语音信息,以及时提醒该用户防止上当受骗。
可选的,终端训练通电话判定算法模型的处理过程可以如下:获取预先存储的多个训练样本,训练样本包括图像样本,及图像样本对应的人物通电话状 态信息;基于多个训练样本,对预设的第二初始算法模型进行训练,得到通电话判定算法模型。
在实施中,终端可以采用深度学习网络改进型模型(如googlenet-loss1)来训练通电话判定算法模型。训练样本可以包括训练正样本和训练负样本。其中,训练正样本中的图像样本可以是包含通电话的人物的图像,图像样本可以是不同姿态打电话的图像,如左手打电话、右手打电话、反手打电话、脸和肩夹住电话等不同姿态,即训练正样本对应的人物通电话状态信息可以为通电话。训练负样本可以是不包含通电话的人物的图像,即训练负样本对应的人物通电话状态信息可以为不通电话。终端中可以预设存储有第二初始算法模型,其中,第二初始算法模型中可以包含待定参数,终端可以根据预先存储的多个训练样本,对预设的第二初始算法模型进行训练,得到待定参数的训练值,即可以得到通电话判定算法模型。基于深度学习技术的训练得到的通电话判定算法模型的检测准确率相比传统方法(如SVM)有较大提升,能排除因头发、眼睛等干扰造成的误检。
可选的,可以结合语音识别技术来识别可能受到欺骗的用户,相应的处理过程可以如下:获取语音输入设备检测到的语音信息;如果语音信息包含预设的关键词,则向服务器发送第二报警消息。
在实施中,语音输入设备可以安装在ATM上,也可以安装在ATM周围。当与ATM的距离在一定范围内的某人发出声音时,语音输入设备可以检测到相应的语音信息,并可以将检测到的语音信息发送给终端,相应的,终端可以接收语音输入设备发送的语音信息。终端接收到语音信号后,可以通过基于预先存储的语音识别算法,来判定该语音信息中是否包括预设的关键词,如卡号、转账等。其中,语音识别算法可以是终端采用相关技术中的语音识别算法模型训练得到的。例如,终端可以根据相关技术中MFCC声纹特征和HMM隐马尔可夫模型对电信诈骗中常出现的敏感词汇(即关键词)进行建模,并可以提取数据库中的声音文件的声纹特征,然后可以建立每个关键词对应的HMM模型,进而可以利用MFCC声纹特征对每一个HMM模型训练,得到语音识别算法。
如果终端通过语音识别算法,识别出语音信息中包含预设的关键词,则可以向服务器发送报警消息(可称为第二报警消息)。服务器接收到该第二报警消息后,可以向安保人员的终端发送携带有目标设备标识的第二警报通知消 息,用于提醒安保人员对正在使用目标设备的用户进行交易干涉,以及时阻止用户向诈骗分子转账。
可选的,终端还可以对检测到的人物进行持续跟踪,以避免重复发送报警消息,相应的处理过程可以如下:确定当前检测到的人脸图像区域和/或头肩图像区域,与当前时间最接近的上一帧检测图像中的人脸图像区域和/或头肩图像区域的相似度;如果相似度不满足预设相似度条件,则执行向服务器发送第一报警消息的步骤。
在实施中,终端检测到人脸图像区域和/或头肩图像区域后,可以确定当前检测到的人脸图像区域和/或头肩图像区域,与当前时间最接近的上一帧检测图像中的人脸图像区域和/或头肩图像区域的相似度。具体的,检测到人脸图像区域和/或头肩图像区域后,终端可以生成人脸图像区域和/或头肩图像区域对应的目标框。以人脸图像区域对应的目标框为例,确定出目标框后,终端可以确定该目标框的属性信息,其中,该属性信息可以包括目标框内像素点的灰度值、目标框的尺寸信息和目标框的位置信息,进而,可以将当前接收到的检测图像中的目标框的属性信息,与当前时间最接近的上一帧检测图像中的目标框的属性信息进行对比,确定二者的相似度,该相似度可以作为当前检测到的人脸图像区域,与当前时间最接近的上一帧检测图像中的人脸图像区域的相似度。
例如,可以计算目标框内像素点的差值的平均值,得到灰度差异度;可以计算尺寸比值,得到尺寸差异度;还可以根据坐标信息,计算位置差异度。如果灰度差异度小于预设的第一差异度阈值,且尺寸差异度小于预设的第二差异度阈值,且位置差异度小于预设的第三差异度阈值,则可以判定二者的相似度为高,否则,可以判定二者的相似度为低。
确定出相似度后,如果终端判定相似度为低,则说明当前检测图像中的用户可能与上一帧检测图像中的用户不相同,终端可以执行向服务器发送第一报警消息的步骤,也即检测到目标检测区域的图像对应的人物通电话状态信息后,如果人物通电话状态信息为通电话,且确定出的上述相似度不满足预设相似度条件,则执行向服务器发送第一报警消息的步骤。如果终端判定相似度为高,则说明当前检测图像中的用户可能与上一帧图像中的用户相同,基于上述处理,终端已向服务器发送对应该用户的报警消息,因此,无需再次向服务器发送报警消息,从而可以避免对于同一用户反复报警的情况。
在实际中,可以优先用人脸图像区域对应的目标框进行判断,如果未检测 到人脸图像区域,则可以用头肩图像区域对应的目标框进行判断,具体的处理过程与上述类似,不再赘述。
需要说明的是,上文的标记图像可以为终端生成的目标框图像,如果终端在上一帧检测图像中检测到人脸图像区域和/或头肩图像区域,而在当前的检测图像中未检测到人脸图像区域和/或头肩图像区域,可以将上一帧检测图像对应的目标框图像,添加到当前的检测图像中,以便安保人员的终端进行显示,如果终端在连续的预设数目帧检测图像中,都未检测到人脸图像区域和/或头肩图像区域,则可以停止在检测图像中添加目标框图像。在实际中,人脸和头肩检测算法模型存在检测误差,会导致检测结果不是连续的,相应的,目标框图像的显示也不是连续的,这样,安保人员看到的检测图像中的目标框图像会是时有时无,甚至闪烁的,而基于本处理,可以提高显示目标框图像的连续性,能有效的提高用户体验。
可选的,为防止预设时长内针对同一用户重复发送报警消息,相应的处理过程可以如下:确定当前检测到的人脸图像区域和/或头肩图像区域,与预设时长内获取到的各帧检测图像中的人脸图像区域和/或头肩图像区域的各相似度;如果得到的各相似度均不满足预设相似度条件,则执行向服务器发送第一报警消息的步骤。
在实施中,检测出当前获取到的检测图像中的人脸图像区域和/或头肩图像区域后,终端可以获取预设时长内获取到的各帧检测图像中的人脸图像区域和/或头肩图像区域,进而,可以按照上述方式,确定当前获取到的检测图像中的人脸图像区域和/或头肩图像区域,与预设时长内获取到的各帧检测图像中的人脸图像区域和/或头肩图像区域的各个相似度,如果得到的各个相似度均不满足预设相似度条件(此种情况说明当前检测图像中的用户可能与预设时长内获取到的其他帧检测图像中用户不同),则终端可以执行向服务器发送第一报警消息的步骤,也即检测到目标检测区域的图像对应的人物通电话状态信息后,如果人物通电话状态信息为通电话,且确定出的上述各个相似度均不满足预设相似度条件,则执行向服务器发送第一报警消息的步骤。如果得到的各相似度中存在满足预设相似度条件的相似度(此种情况说明,当前检测图像中的用户可能与预设时长内获取到的其他帧检测图像中的某用户是同一用户),则基于上述处理,终端已向服务器发送对应该用户的报警消息,因此,无需再次向服务器发送报警消息,从而可以避免对于同一用户反复报警的情况。
本申请实施例中,终端获取到摄像设备拍摄的检测图像后,可以在检测图像中确定目标检测区域,根据预设的通电话判定算法模型,检测目标检测区域的图像对应的人物通电话状态信息,如果人物通电话状态信息为通电话,则向服务器发送第一报警消息,这样,可以及时的识别出用户是否在打电话,并可以在检测到打电话时,及时发送报警消息,以使安保人员及时获知可能存在受到欺骗的用户,并进行处理,避免用户的财产受到损失。
基于相同的技术构思,本申请实施例还提供了一种发送报警消息的装置,如图6所示,该装置包括:
第一获取模块610,用于获取摄像设备拍摄的检测图像;
第一确定模块620,用于在所述检测图像中确定目标检测区域;
第一发送模块630,用于根据预设的通电话判定算法模型,检测所述目标检测区域的图像对应的人物通电话状态信息,如果所述人物通电话状态信息为通电话,则向服务器发送第一报警消息。
可选的,如图7所示,所述装置还包括:
检测模块640,用于基于预设的人脸和头肩检测算法模型,在所述检测图像中检测人脸图像区域和头肩图像区域;
所述第一确定模块620,用于:根据人脸图像区域和头肩图像区域的检测结果,在所述检测图像中确定目标检测区域。
可选的,所述第一确定模块620,用于:
如果检测到人脸图像区域和头肩图像区域,则根据预先存储的人脸图像区域、头肩图像区域和检测区域的位置关系,在所述检测图像中确定目标检测区域;
如果检测到人脸图像区域,未检测到头肩图像区域,则对检测到的人脸图像区域进行放大处理,将放大处理后的人脸图像区域作为目标检测区域;
如果检测到头肩图像区域,未检测到人脸图像区域,则对检测到的头肩图像区域进行缩小处理,将缩小处理后的头肩图像区域作为目标检测区域。
可选的,如图8所示,所述装置还包括:
第二获取模块650,用于获取预先存储的多个训练样本,所述训练样本包括图像样本,及所述图像样本中的人脸图像区域和/或头肩图像区域;
第一训练模块660,用于基于所述多个训练样本,对预设的第一初始算法 模型进行训练,得到所述人脸和头肩检测算法模型。
可选的,如图9所示,所述装置还包括:
第三获取模块670,用于获取预先存储的多个训练样本,所述训练样本包括图像样本,及所述图像样本对应的人物通电话状态信息;
第二训练模块680,用于基于所述多个训练样本,对预设的第二初始算法模型进行训练,得到所述通电话判定算法模型。
可选的,如图10所示,所述检测模块640,包括:
第一确定子模块641,用于基于预设的人脸和头肩检测算法模型,在所述检测图像中确定人脸图像待定区域和头肩图像待定区域,并确定所述人脸图像待定区域对应的置信度和所述头肩图像待定区域对应的置信度;
第二确定子模块642,用于根据预先存储的位置信息和权重的对应关系,确定所述人脸图像待定区域对应的权值,以及所述头肩图像待定区域对应的权值;
所述第二确定子模块642,还用于如果所述人脸图像待定区域对应的置信度大于预设的第一置信度阈值,且所述人脸图像待定区域对应的权值大于预设的第一权值阈值,则将所述人脸图像待定区域作为人脸图像区域;
所述第二确定子模块642,还用于如果所述头肩图像待定区域对应的置信度大于预设的第二置信度阈值,且所述头肩图像待定区域对应的权值大于预设的第二权值阈值,则将所述头肩图像待定区域作为头肩图像区域。
可选的,如图11所示,所述装置还包括:
第二确定模块690,用于确定当前检测到的人脸图像区域和/或头肩图像区域,与当前时间最接近的上一帧检测图像中的人脸图像区域和/或头肩图像区域的相似度;
所述第一发送模块630,用于:
如果所述相似度不满足预设相似度条件,则执行向服务器发送第一报警消息的步骤。
可选的,如图12所示,所述装置还包括:
第四获取模块6100,用于获取语音输入设备检测到的语音信息;
第二发送模块6110,用于如果所述语音信息包含预设的关键词,则向所述服务器发送第二报警消息。
可选的,如图13所示,所述装置还包括:
第三确定模块6120,用于如果达到预设的检测触发条件,则执行基于预设的人脸和头肩检测算法模型,在所述检测图像中检测人脸图像区域和头肩图像区域的步骤;
所述预设的检测触发条件至少包括:
在所述检测图像中检测到人物活动信息;或者,
接收到目标设备发送的操作通知消息;或者,
接收到传感设备发送的对象检测通知。
可选的,所述第一确定模块620,还用于:
确定当前检测到的人脸图像区域和/或头肩图像区域,与预设时长内获取到的各帧检测图像中的人脸图像区域和/或头肩图像区域的各相似度;
所述第一发送模块630,还用于:
如果得到的各相似度均不满足预设相似度条件,则执行向服务器发送第一报警消息的步骤。
本申请实施例中,终端获取到摄像设备拍摄的检测图像后,可以在检测图像中确定目标检测区域,根据预设的通电话判定算法模型,检测目标检测区域的图像对应的人物通电话状态信息,如果人物通电话状态信息为通电话,则向服务器发送第一报警消息,这样,可以及时的识别出用户是否在打电话,并可以在检测到打电话时,及时发送报警消息,以使安保人员及时获知可能存在受到欺骗的用户,并进行处理,避免用户的财产受到损失。
需要说明的是:上述实施例提供的发送报警消息的装置在发送报警消息时,仅以上述各功能模块的划分进行举例说明,实际应用中,可以根据需要而将上述功能分配由不同的功能模块完成,即将设备的内部结构划分成不同的功能模块,以完成以上描述的全部或者部分功能。另外,上述实施例提供的发送报警消息的装置与发送报警消息的方法实施例属于同一构思,其具体实现过程详见方法实施例,这里不再赘述。
请参考图14,其示出了本申请实施例所涉及的终端的结构示意图,该终端可以用于实施上述实施例中提供的发送报警消息的方法。具体来讲:
终端900可以包括RF(Radio Frequency,射频)电路110、包括有一个或一个以上计算机可读存储介质的存储器120、输入单元130、显示单元140、传感器150、音频电路160、WiFi(wireless fidelity,无线保真)模块170、包括有 一个或者一个以上处理核心的处理器180、以及电源190等部件。本领域技术人员可以理解,图14中示出的终端结构并不构成对终端的限定,可以包括比图示更多或更少的部件,或者组合某些部件,或者不同的部件布置。其中:
RF电路110可用于收发信息或通话过程中,信号的接收和发送,特别地,将基站的下行信息接收后,交由一个或者一个以上处理器180处理;另外,将涉及上行的数据发送给基站。通常,RF电路110包括但不限于天线、至少一个放大器、调谐器、一个或多个振荡器、用户身份模块(SIM)卡、收发信机、耦合器、LNA(Low Noise Amplifier,低噪声放大器)、双工器等。此外,RF电路110还可以通过无线通信与网络和其他设备通信。所述无线通信可以使用任一通信标准或协议,包括但不限于GSM(Global System of Mobile communication,全球移动通讯系统)、GPRS(General Packet Radio Service,通用分组无线服务)、CDMA(Code Division Multiple Access,码分多址)、WCDMA(Wideband Code Division Multiple Access,宽带码分多址)、LTE(Long Term Evolution,长期演进)、电子邮件、SMS(Short Messaging Service,短消息服务)等。
存储器120可用于存储软件程序以及模块,处理器180通过运行存储在存储器120的软件程序以及模块,从而执行各种功能应用以及数据处理。存储器120可主要包括存储程序区和存储数据区,其中,存储程序区可存储操作系统、至少一个功能所需的应用程序(比如声音播放功能、图像播放功能等)等;存储数据区可存储根据终端900的使用所创建的数据(比如音频数据、电话本等)等。此外,存储器120可以包括高速随机存取存储器,还可以包括非易失性存储器,例如至少一个磁盘存储器件、闪存器件、或其他易失性固态存储器件。相应地,存储器120还可以包括存储器控制器,以提供处理器180和输入单元130对存储器120的访问。
输入单元130可用于接收输入的数字或字符信息,以及产生与用户设置以及功能控制有关的键盘、鼠标、操作杆、光学或者轨迹球信号输入。具体地,输入单元130可包括触敏表面131以及其他输入设备132。触敏表面131,也称为触摸显示屏或者触控板,可收集用户在其上或附近的触摸操作(比如用户使用手指、触笔等任何适合的物体或附件在触敏表面131上或在触敏表面131附近的操作),并根据预先设定的程式驱动相应的连接装置。可选的,触敏表面131可包括触摸检测装置和触摸控制器两个部分。其中,触摸检测装置检测 用户的触摸方位,并检测触摸操作带来的信号,将信号传送给触摸控制器;触摸控制器从触摸检测装置上接收触摸信息,并将它转换成触点坐标,再送给处理器180,并能接收处理器180发来的命令并加以执行。此外,可以采用电阻式、电容式、红外线以及表面声波等多种类型实现触敏表面131。除了触敏表面131,输入单元130还可以包括其他输入设备132。具体地,其他输入设备132可以包括但不限于物理键盘、功能键(比如音量控制按键、开关按键等)、轨迹球、鼠标、操作杆等中的一种或多种。
显示单元140可用于显示由用户输入的信息或提供给用户的信息以及终端900的各种图形用户接口,这些图形用户接口可以由图形、文本、图标、视频和其任意组合来构成。显示单元140可包括显示面板141,可选的,可以采用LCD(Liquid Crystal Display,液晶显示器)、OLED(Organic Light-Emitting Diode,有机发光二极管)等形式来配置显示面板141。进一步的,触敏表面131可覆盖显示面板141,当触敏表面131检测到在其上或附近的触摸操作后,传送给处理器180以确定触摸事件的类型,随后处理器180根据触摸事件的类型在显示面板141上提供相应的视觉输出。虽然在图14中,触敏表面131与显示面板141是作为两个独立的部件来实现输入和输入功能,但是在某些实施例中,可以将触敏表面131与显示面板141集成而实现输入和输出功能。
终端900还可包括至少一种传感器150,比如光传感器、运动传感器以及其他传感器。具体地,光传感器可包括环境光传感器及接近传感器,其中,环境光传感器可根据环境光线的明暗来调节显示面板141的亮度,接近传感器可在终端900移动到耳边时,关闭显示面板141和/或背光。作为运动传感器的一种,重力加速度传感器可检测各个方向上(一般为三轴)加速度的大小,静止时可检测出重力的大小及方向,可用于识别手机姿态的应用(比如横竖屏切换、相关游戏、磁力计姿态校准)、振动识别相关功能(比如计步器、敲击)等;至于终端900还可配置的陀螺仪、气压计、湿度计、温度计、红外线传感器等其他传感器,在此不再赘述。
音频电路160、扬声器161,传声器162可提供用户与终端900之间的音频接口。音频电路160可将接收到的音频数据转换后的电信号,传输到扬声器161,由扬声器161转换为声音信号输出;另一方面,传声器162将收集的声音信号转换为电信号,由音频电路160接收后转换为音频数据,再将音频数据输出处理器180处理后,经RF电路110以发送给比如另一终端,或者将音频 数据输出至存储器120以便进一步处理。音频电路160还可能包括耳塞插孔,以提供外设耳机与终端900的通信。
WiFi属于短距离无线传输技术,终端900通过WiFi模块170可以帮助用户收发电子邮件、浏览网页和访问流式媒体等,它为用户提供了无线的宽带互联网访问。虽然图14示出了WiFi模块170,但是可以理解的是,其并不属于终端900的必须构成,完全可以根据需要在不改变发明的本质的范围内而省略。
处理器180是终端900的控制中心,利用各种接口和线路连接整个手机的各个部分,通过运行或执行存储在存储器120内的软件程序和/或模块,以及调用存储在存储器120内的数据,执行终端900的各种功能和处理数据,从而对手机进行整体监控。可选的,处理器180可包括一个或多个处理核心;优选的,处理器180可集成应用处理器和调制解调处理器,其中,应用处理器主要处理操作系统、用户界面和应用程序等,调制解调处理器主要处理无线通信。可以理解的是,上述调制解调处理器也可以不集成到处理器180中。
终端900还包括给各个部件供电的电源190(比如电池),优选的,电源可以通过电源管理系统与处理器180逻辑相连,从而通过电源管理系统实现管理充电、放电、以及功耗管理等功能。电源190还可以包括一个或一个以上的直流或交流电源、再充电系统、电源故障检测电路、电源转换器或者逆变器、电源状态指示器等任意组件。
尽管未示出,终端900还可以包括摄像头、蓝牙模块等,在此不再赘述。具体在本实施例中,终端900的显示单元是触摸屏显示器,终端900还包括有存储器,以及一个或者一个以上的程序,其中一个或者一个以上程序存储于存储器中,且经配置以由一个或者一个以上处理器执行包含该终端用于执行上述发送报警消息的方法的指令。
本申请实施例中,终端获取到摄像设备拍摄的检测图像后,可以在检测图像中确定目标检测区域,根据预设的通电话判定算法模型,检测目标检测区域的图像对应的人物通电话状态信息,如果人物通电话状态信息为通电话,则向服务器发送第一报警消息,这样,可以及时的识别出用户是否在打电话,并可以在检测到打电话时,及时发送报警消息,以使安保人员及时获知可能存在受到欺骗的用户,并进行处理,避免用户的财产受到损失。
本领域普通技术人员可以理解实现上述实施例的全部或部分步骤可以通 过硬件来完成,也可以通过程序来指令相关的硬件完成,所述的程序可以存储于一种计算机可读存储介质中,上述提到的存储介质可以是只读存储器,磁盘或光盘等。
以上所述仅为本申请的较佳实施例,并不用以限制本申请,凡在本申请的精神和原则之内,所作的任何修改、等同替换、改进等,均应包含在本申请的保护范围之内。

Claims (22)

  1. 一种发送报警消息的方法,其特征在于,所述方法包括:
    获取摄像设备拍摄的检测图像;
    在所述检测图像中确定目标检测区域;
    根据预设的通电话判定算法模型,检测所述目标检测区域的图像对应的人物通电话状态信息,如果所述人物通电话状态信息为通电话,则向服务器发送第一报警消息。
  2. 根据权利要求1所述的方法,其特征在于,所述在所述检测图像中确定目标检测区域,包括:
    基于预设的人脸和头肩检测算法模型,在所述检测图像中检测人脸图像区域和头肩图像区域;
    根据人脸图像区域和头肩图像区域的检测结果,在所述检测图像中确定目标检测区域。
  3. 根据权利要求2所述的方法,其特征在于,所述根据人脸图像区域和头肩图像区域的检测结果,在所述检测图像中确定目标检测区域,包括:
    如果检测到人脸图像区域和头肩图像区域,则根据预先存储的人脸图像区域、头肩图像区域和检测区域的位置关系,在所述检测图像中确定目标检测区域;
    如果检测到人脸图像区域,未检测到头肩图像区域,则对检测到的人脸图像区域进行放大处理,将放大处理后的人脸图像区域作为目标检测区域;
    如果检测到头肩图像区域,未检测到人脸图像区域,则对检测到的头肩图像区域进行缩小处理,将缩小处理后的头肩图像区域作为目标检测区域。
  4. 根据权利要求2所述的方法,其特征在于,所述方法还包括:
    获取预先存储的多个训练样本,所述训练样本包括图像样本,及所述图像样本中的人脸图像区域和/或头肩图像区域;
    基于所述多个训练样本,对预设的第一初始算法模型进行训练,得到所述人脸和头肩检测算法模型。
  5. 根据权利要求2所述的方法,其特征在于,所述基于预设的人脸和头肩检测算法模型,在所述检测图像中检测人脸图像区域和头肩图像区域,包括:
    基于预设的人脸和头肩检测算法模型,在所述检测图像中确定人脸图像待 定区域和头肩图像待定区域,并确定所述人脸图像待定区域对应的置信度和所述头肩图像待定区域对应的置信度;
    根据预先存储的位置信息和权重的对应关系,确定所述人脸图像待定区域对应的权值,以及所述头肩图像待定区域对应的权值;
    如果所述人脸图像待定区域对应的置信度大于预设的第一置信度阈值,且所述人脸图像待定区域对应的权值大于预设的第一权值阈值,则将所述人脸图像待定区域作为人脸图像区域;
    如果所述头肩图像待定区域对应的置信度大于预设的第二置信度阈值,且所述头肩图像待定区域对应的权值大于预设的第二权值阈值,则将所述头肩图像待定区域作为头肩图像区域。
  6. 根据权利要求2所述的方法,其特征在于,所述方法还包括:
    确定当前检测到的人脸图像区域和/或头肩图像区域,与当前时间最接近的上一帧检测图像中的人脸图像区域和/或头肩图像区域的相似度;
    所述向服务器发送第一报警消息之前,还包括:
    如果所述相似度不满足预设相似度条件,则执行向服务器发送第一报警消息的步骤。
  7. 根据权利要求2所述的方法,其特征在于,所述基于预设的人脸和头肩检测算法模型,在所述检测图像中检测人脸图像区域和头肩图像区域之前,还包括:
    如果达到预设的检测触发条件,则执行基于预设的人脸和头肩检测算法模型,在所述检测图像中检测人脸图像区域和头肩图像区域的步骤;
    所述预设的检测触发条件至少包括:
    在所述检测图像中检测到人物活动信息;或者,
    接收到目标设备发送的操作通知消息;或者,
    接收到传感设备发送的对象检测通知。
  8. 根据权利要求2所述的方法,其特征在于,所述方法还包括:
    确定当前检测到的人脸图像区域和/或头肩图像区域,与预设时长内获取到的各帧检测图像中的人脸图像区域和/或头肩图像区域的各相似度;
    所述向服务器发送第一报警消息之前,还包括:
    如果得到的各相似度均不满足预设相似度条件,则执行向服务器发送第一报警消息的步骤。
  9. 根据权利要求1所述的方法,其特征在于,所述方法还包括:
    获取预先存储的多个训练样本,所述训练样本包括图像样本,及所述图像样本对应的人物通电话状态信息;
    基于所述多个训练样本,对预设的第二初始算法模型进行训练,得到所述通电话判定算法模型。
  10. 根据权利要求1所述的方法,其特征在于,所述方法还包括:
    获取语音输入设备检测到的语音信息;
    如果所述语音信息包含预设的关键词,则向所述服务器发送第二报警消息。
  11. 一种发送报警消息的装置,其特征在于,所述装置包括:
    第一获取模块,用于获取摄像设备拍摄的检测图像;
    第一确定模块,用于在所述检测图像中确定目标检测区域;
    第一发送模块,用于根据预设的通电话判定算法模型,检测所述目标检测区域的图像对应的人物通电话状态信息,如果所述人物通电话状态信息为通电话,则向服务器发送第一报警消息。
  12. 根据权利要求11所述的装置,其特征在于,所述装置还包括:
    检测模块,用于基于预设的人脸和头肩检测算法模型,在所述检测图像中检测人脸图像区域和头肩图像区域;
    所述第一确定模块,用于:根据人脸图像区域和头肩图像区域的检测结果,在所述检测图像中确定目标检测区域。
  13. 根据权利要求12所述的装置,其特征在于,所述第一确定模块,用于:
    如果检测到人脸图像区域和头肩图像区域,则根据预先存储的人脸图像区域、头肩图像区域和检测区域的位置关系,在所述检测图像中确定目标检测区域;
    如果检测到人脸图像区域,未检测到头肩图像区域,则对检测到的人脸图像区域进行放大处理,将放大处理后的人脸图像区域作为目标检测区域;
    如果检测到头肩图像区域,未检测到人脸图像区域,则对检测到的头肩图像区域进行缩小处理,将缩小处理后的头肩图像区域作为目标检测区域。
  14. 根据权利要求12所述的装置,其特征在于,所述装置还包括:
    第二获取模块,用于获取预先存储的多个训练样本,所述训练样本包括图像样本,及所述图像样本中的人脸图像区域和/或头肩图像区域;
    第一训练模块,用于基于所述多个训练样本,对预设的第一初始算法模型进行训练,得到所述人脸和头肩检测算法模型。
  15. 根据权利要求12所述的装置,其特征在于,所述检测模块,包括:
    第一确定子模块,用于基于预设的人脸和头肩检测算法模型,在所述检测图像中确定人脸图像待定区域和头肩图像待定区域,并确定所述人脸图像待定区域对应的置信度和所述头肩图像待定区域对应的置信度;
    第二确定子模块,用于根据预先存储的位置信息和权重的对应关系,确定所述人脸图像待定区域对应的权值,以及所述头肩图像待定区域对应的权值;
    所述第二确定子模块,还用于如果所述人脸图像待定区域对应的置信度大于预设的第一置信度阈值,且所述人脸图像待定区域对应的权值大于预设的第一权值阈值,则将所述人脸图像待定区域作为人脸图像区域;
    所述第二确定子模块,还用于如果所述头肩图像待定区域对应的置信度大于预设的第二置信度阈值,且所述头肩图像待定区域对应的权值大于预设的第二权值阈值,则将所述头肩图像待定区域作为头肩图像区域。
  16. 根据权利要求12所述的装置,其特征在于,所述装置还包括:
    第二确定模块,用于确定当前检测到的人脸图像区域和/或头肩图像区域,与当前时间最接近的上一帧检测图像中的人脸图像区域和/或头肩图像区域的相似度;
    所述第一发送模块,用于:
    如果所述相似度不满足预设相似度条件,则执行向服务器发送第一报警消息的步骤。
  17. 根据权利要求12所述的装置,其特征在于,所述装置还包括:
    第三确定模块,用于如果达到预设的检测触发条件,则执行基于预设的人脸和头肩检测算法模型,在所述检测图像中检测人脸图像区域和头肩图像区域的步骤;
    所述预设的检测触发条件至少包括:
    在所述检测图像中检测到人物活动信息;或者,
    接收到目标设备发送的操作通知消息;或者,
    接收到传感设备发送的对象检测通知。
  18. 根据权利要求12所述的装置,其特征在于,所述第一确定模块,还用于:
    确定当前检测到的人脸图像区域和/或头肩图像区域,与预设时长内获取到的各帧检测图像中的人脸图像区域和/或头肩图像区域的各相似度;
    所述第一发送模块,还用于:
    如果得到的各相似度均不满足预设相似度条件,则执行向服务器发送第一报警消息的步骤。
  19. 根据权利要求11所述的装置,其特征在于,所述装置还包括:
    第三获取模块,用于获取预先存储的多个训练样本,所述训练样本包括图像样本,及所述图像样本对应的人物通电话状态信息;
    第二训练模块,用于基于所述多个训练样本,对预设的第二初始算法模型进行训练,得到所述通电话判定算法模型。
  20. 根据权利要求11所述的装置,其特征在于,所述装置还包括:
    第四获取模块,用于获取语音输入设备检测到的语音信息;
    第二发送模块,用于如果所述语音信息包含预设的关键词,则向所述服务器发送第二报警消息。
  21. 一种计算机可读存储介质,其特征在于,所述存储介质内存储有计算机程序,所述计算机程序被处理器执行时实现权利要求1-10任一所述的方法步骤。
  22. 一种终端,其特征在于,所述终端包括:
    一个或多个处理器;和
    存储器;
    所述存储器存储有一个或多个程序,所述一个或多个程序被配置成由所述一个或多个处理器执行,所述一个或多个程序包含用于进行如权利要求1-10任一所述的方法步骤的指令。
PCT/CN2018/073775 2017-01-23 2018-01-23 一种发送报警消息的方法和装置 WO2018133874A1 (zh)

Priority Applications (2)

Application Number Priority Date Filing Date Title
US16/480,126 US11386698B2 (en) 2017-01-23 2018-01-23 Method and device for sending alarm message
EP18741996.5A EP3572974B1 (en) 2017-01-23 2018-01-23 Method and device for sending warning message

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201710050732.5 2017-01-23
CN201710050732.5A CN108345819B (zh) 2017-01-23 2017-01-23 一种发送报警消息的方法和装置

Publications (1)

Publication Number Publication Date
WO2018133874A1 true WO2018133874A1 (zh) 2018-07-26

Family

ID=62908530

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2018/073775 WO2018133874A1 (zh) 2017-01-23 2018-01-23 一种发送报警消息的方法和装置

Country Status (4)

Country Link
US (1) US11386698B2 (zh)
EP (1) EP3572974B1 (zh)
CN (1) CN108345819B (zh)
WO (1) WO2018133874A1 (zh)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112825138A (zh) * 2019-11-21 2021-05-21 佳能株式会社 图像处理设备、图像处理方法、摄像设备及机器可读介质
CN113709564A (zh) * 2020-05-21 2021-11-26 云米互联科技(广东)有限公司 基于5g电视的预警方法、5g电视及可读存储介质

Families Citing this family (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR102576388B1 (ko) * 2018-02-21 2023-09-08 엘지전자 주식회사 디스플레이 장치 및 그의 동작 방법
CN110895661A (zh) * 2018-09-12 2020-03-20 杭州海康威视数字技术股份有限公司 一种行为识别方法、装置及设备
CN109684990B (zh) * 2018-12-20 2023-05-30 天津天地伟业信息系统集成有限公司 一种基于视频的打电话行为检测方法
CN112307839B (zh) * 2019-07-31 2024-10-18 深圳熙卓科技有限公司 一种异常人员检测的方法及终端设备
CN110874582A (zh) * 2019-11-18 2020-03-10 浙江大华技术股份有限公司 电信诈骗的确定方法及装置、存储介质、电子装置
CN111222493B (zh) * 2020-01-20 2023-07-28 北京捷通华声科技股份有限公司 一种视频处理方法及装置
CN111382726B (zh) * 2020-04-01 2023-09-01 浙江大华技术股份有限公司 工程作业检测方法以及相关装置
CN114155483A (zh) * 2021-11-11 2022-03-08 鸿富锦精密电子(郑州)有限公司 监控告警方法、装置、存储介质及计算机设备
CN114495491A (zh) * 2021-12-31 2022-05-13 深圳云天励飞技术股份有限公司 跨线出行路线确定方法、装置、计算机设备及介质

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103366506A (zh) * 2013-06-27 2013-10-23 北京理工大学 一种驾驶员行车途中接打手机行为的自动监控装置及方法
CN104573659A (zh) * 2015-01-09 2015-04-29 安徽清新互联信息科技有限公司 一种基于svm的驾驶员接打电话监控方法
CN104966059A (zh) * 2015-06-15 2015-10-07 安徽创世科技有限公司 基于智能监控系统的驾驶员开车打电话检测方法
US20150286885A1 (en) * 2014-04-04 2015-10-08 Xerox Corporation Method for detecting driver cell phone usage from side-view images
CN105913022A (zh) * 2016-04-11 2016-08-31 深圳市飞瑞斯科技有限公司 一种基于视频分析的手持式打电话状态判别方法及系统

Family Cites Families (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US9952046B1 (en) * 2011-02-15 2018-04-24 Guardvant, Inc. Cellular phone and personal protective equipment usage monitoring system
TW201447817A (zh) * 2013-06-04 2014-12-16 Utechzone Co Ltd 偵測人員使用手持裝置的方法、裝置及其影像警報系統
US9721173B2 (en) * 2014-04-04 2017-08-01 Conduent Business Services, Llc Machine learning approach for detecting mobile phone usage by a driver
TWI550440B (zh) * 2014-12-09 2016-09-21 由田新技股份有限公司 人員使用手持裝置的偵測方法以及系統
TWI603270B (zh) * 2014-12-11 2017-10-21 由田新技股份有限公司 偵測人員使用手持裝置的方法及裝置
TWI520076B (zh) * 2014-12-11 2016-02-01 由田新技股份有限公司 偵測人員使用手持裝置的方法及裝置
CN105046725B (zh) * 2015-07-10 2017-03-29 清华大学 低比特率视频通话中基于模型和对象的头肩图像重建方法

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103366506A (zh) * 2013-06-27 2013-10-23 北京理工大学 一种驾驶员行车途中接打手机行为的自动监控装置及方法
US20150286885A1 (en) * 2014-04-04 2015-10-08 Xerox Corporation Method for detecting driver cell phone usage from side-view images
CN104573659A (zh) * 2015-01-09 2015-04-29 安徽清新互联信息科技有限公司 一种基于svm的驾驶员接打电话监控方法
CN104966059A (zh) * 2015-06-15 2015-10-07 安徽创世科技有限公司 基于智能监控系统的驾驶员开车打电话检测方法
CN105913022A (zh) * 2016-04-11 2016-08-31 深圳市飞瑞斯科技有限公司 一种基于视频分析的手持式打电话状态判别方法及系统

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
See also references of EP3572974A4 *

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112825138A (zh) * 2019-11-21 2021-05-21 佳能株式会社 图像处理设备、图像处理方法、摄像设备及机器可读介质
EP3826293A1 (en) * 2019-11-21 2021-05-26 Canon Kabushiki Kaisha Detection of head and face separately and selection of result for use in detection of facial features
US11670112B2 (en) 2019-11-21 2023-06-06 Canon Kabushiki Kaisha Image processing apparatus, image processing method, and image capture apparatus
CN113709564A (zh) * 2020-05-21 2021-11-26 云米互联科技(广东)有限公司 基于5g电视的预警方法、5g电视及可读存储介质
CN113709564B (zh) * 2020-05-21 2023-12-12 云米互联科技(广东)有限公司 基于5g电视的预警方法、5g电视及可读存储介质

Also Published As

Publication number Publication date
CN108345819B (zh) 2020-09-15
EP3572974B1 (en) 2023-11-22
EP3572974A1 (en) 2019-11-27
US11386698B2 (en) 2022-07-12
US20210319211A1 (en) 2021-10-14
CN108345819A (zh) 2018-07-31
EP3572974A4 (en) 2020-03-04

Similar Documents

Publication Publication Date Title
WO2018133874A1 (zh) 一种发送报警消息的方法和装置
CN109828688B (zh) 屏幕损坏处理方法、移动终端及计算机可读存储介质
US9906406B2 (en) Alerting method and mobile terminal
WO2021057267A1 (zh) 图像处理方法及终端设备
CN108830062B (zh) 人脸识别方法、移动终端及计算机可读存储介质
CN109195213B (zh) 移动终端屏幕控制方法、移动终端及计算机可读存储介质
WO2019206077A1 (zh) 视频通话处理方法及移动终端
WO2019154184A1 (zh) 生物特征识别方法及移动终端
CN107885448B (zh) 应用触摸操作的控制方法、移动终端及可读存储介质
CN110275794B (zh) 终端的屏幕显示方法、终端以及计算机可读存储介质
CN113452845B (zh) 识别异常电话号码的方法和电子设备
CN110009004B (zh) 图像数据处理方法、计算机设备和存储介质
CN107896287B (zh) 手机号码风险监测方法、移动终端及计算机可读存储介质
WO2021104266A1 (zh) 对象显示方法及电子设备
CN109165534A (zh) 终端控制方法、终端及计算机可读存储介质
CN109462696A (zh) 信息提示方法、移动终端及计算机可读存储介质
CN110427149B (zh) 终端的操作方法及终端
CN108924355A (zh) 信号频率调整方法、移动终端及计算机可读存储介质
CN110418004A (zh) 截图处理方法、终端及计算机可读存储介质
CN110401806A (zh) 一种移动终端的视频通话方法、移动终端及存储介质
CN111443818B (zh) 一种屏幕亮度调控方法、设备及计算机可读存储介质
CN110399780B (zh) 一种人脸检测方法、装置及计算机可读存储介质
CN109711850B (zh) 一种安全支付方法、设备及计算机可读存储介质
CN110083294A (zh) 屏幕截取方法、终端和计算机可读存储介质
CN109889660A (zh) 临时信息记录方法、存储介质和移动终端

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 18741996

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

WWE Wipo information: entry into national phase

Ref document number: 2018741996

Country of ref document: EP