CN108184062A - High speed tracing system and method based on multi-level heterogeneous parallel processing - Google Patents

High speed tracing system and method based on multi-level heterogeneous parallel processing Download PDF

Info

Publication number
CN108184062A
CN108184062A CN201711498701.2A CN201711498701A CN108184062A CN 108184062 A CN108184062 A CN 108184062A CN 201711498701 A CN201711498701 A CN 201711498701A CN 108184062 A CN108184062 A CN 108184062A
Authority
CN
China
Prior art keywords
module
image
speed
high speed
level
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN201711498701.2A
Other languages
Chinese (zh)
Other versions
CN108184062B (en
Inventor
王红伟
王加庆
刘力源
吴南健
秦琦
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Institute of Semiconductors of CAS
Original Assignee
Institute of Semiconductors of CAS
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 Institute of Semiconductors of CAS filed Critical Institute of Semiconductors of CAS
Priority to CN201711498701.2A priority Critical patent/CN108184062B/en
Publication of CN108184062A publication Critical patent/CN108184062A/en
Application granted granted Critical
Publication of CN108184062B publication Critical patent/CN108184062B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • 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
    • H04N23/67Focus control based on electronic image sensor signals
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00Image analysis
    • G06T7/20Analysis of motion
    • G06T7/246Analysis of motion using feature-based methods, e.g. the tracking of corners or segments
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04NPICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N23/00Cameras or camera modules comprising electronic image sensors; Control thereof
    • H04N23/80Camera processing pipelines; Components thereof
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04NPICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N23/00Cameras or camera modules comprising electronic image sensors; Control thereof
    • H04N23/80Camera processing pipelines; Components thereof
    • H04N23/81Camera processing pipelines; Components thereof for suppressing or minimising disturbance in the image signal generation
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04NPICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N5/00Details of television systems
    • H04N5/14Picture signal circuitry for video frequency region
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00Indexing scheme for image analysis or image enhancement
    • G06T2207/10Image acquisition modality
    • G06T2207/10016Video; Image sequence

Landscapes

  • Engineering & Computer Science (AREA)
  • Multimedia (AREA)
  • Signal Processing (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Studio Devices (AREA)
  • Image Processing (AREA)

Abstract

A kind of high speed tracing system and method based on multi-level heterogeneous parallel processing.The high speed tracing system includes the high-speed camera with autonomous focusing function, two-dimentional servo execution control response module, two-dimentional servo module, image data acquiring transmission module and image processing module.Main working process is:Image data acquiring transmission module acquires image data by high-speed camera, image data is packaged and is transferred to image processing module, image processing module carries out operation, and obtain the location information of target and the rotation angle information of servo module according to algorithm difference to image data;Servo module carries out the tracking of target according to the rotation angle transmitted from image processing module, and the high-speed camera of autonomous focusing function carries out auto-focusing according to target position information.The system and method for the present invention may be used on high-speed target identification and detection scene in production and living, can realize under complex scene to the robust tracking of high-speed moving object.

Description

High speed tracing system and method based on multi-level heterogeneous parallel processing
Technical field
The present invention relates to programmable vision chip and high-speed object detection, tracking field more particularly to based on multi-level The high speed tracing system and method for heterogeneous parallel processing have the advantage of high speed, highly integrated, low-power consumption, low cost, can apply In tracking of the high-speed target of security monitoring, field of industrial production and sphere of life etc..
Background technology
With computer application and the development of machine vision, image is acquired by camera and carry out high-speed target detection and The SPEED VISION target following technology of tracking is achieved compared with much progress and is widely applied in industrial and agricultural production life.According to using The type of camera, the camera of Current vision Target Tracking System acquire image data using conventional image sensor, lead to picture It is low that the frame per second of acquisition is relatively low, picture obtains resolution ratio;According to the type of processor, Current vision Target Tracking System is based on calculating The tracking system of machine/general processor, since the specific aim of transmission speed limitation, shortage tracking is supported, system processing speed Degree is slow, can not apply in the SPEED VISION tracking field higher to requirement of real-time.
Invention content
In view of this, it is a primary object of the present invention to propose a kind of high speed tracing system and method, at least partly Ground solves at least one of above-mentioned technical problem.
To achieve these goals, as one aspect of the present invention, it is proposed that a kind of high speed tracing system, feature exist In the high speed tracing system includes high-speed image sampling device, two-dimentional servo performs control response module, two-dimentional servo mould Block, image data acquiring transmission module and image processing module;Wherein:
High-speed image sampling device has autonomous focusing function;
The picture signal of high-speed image sampling device acquisition is input to image procossing by image data acquiring transmission module Module;
Two-dimentional servo performs the two-dimentional servo mould of control information generation that control response module is exported based on image processing module The control instruction of block simultaneously sends it to the two-dimentional servo module;
Described image processing module is integrated with Pixel-level, block grade and distributed multi-stage heterogeneous parallel processing unit, has Image Feature Detection, feature extraction, tagsort and deep learning convolution algorithm processing function;Wherein, at described image The pixel processing unit PE arrays that reason module first passes through pixel level processor complete Pixel-level parallel algorithm, then by block grade simultaneously Block processing unit PPU in row processor accesses the image block in pixel level processor pixel processing unit PE arrays to complete office Portion's feature extraction algorithm finally realized using Distributed Calculation and carries out Fast Classification to the local feature of extraction, is generated described Information is controlled, so as to fulfill the parallel processing based on multi-level isomery.
As another aspect of the present invention, it is also proposed that a kind of high speed tracking side based on multi-level heterogeneous parallel processing Method uses high speed tracing system as described above to realize, which is characterized in that the high speed method for tracing includes following step Suddenly:
When in the visual field of the high-speed image sampling device with autonomous focusing function described in tracking target and appearing in, image Information collection transmission module receives image data from the high-speed image sampling device, is transferred to after image noise reduction is handled Image processing module;
Image processing module passes through carries out Pixel-level, block grade and distributed variable-frequencypump to described image data, obtains institute The location information of tracking target is stated, and sends location information to the two-dimentional servo and performs response module, is aligned simultaneously Confidence breath is judged, is obtained focusing coefficient and is transmitted to the high-speed image sampling device for carrying autonomous focusing function, completes The autonomous focusing of high-speed image sampling device;
Two-dimentional servo performs response module and the location information for tracking target is calculated, and obtains directional information and rotation Gyration information;Two-dimentional servo module receives the directional information and rotation angle that response module is performed from two-dimentional servo After information, rotated accordingly;
Meanwhile the tracking target information is shown on display module by image processing module.
Compared with prior art, the beneficial effects of the invention are as follows:
(1) high-speed camera with autonomous focusing function and integrated high speed imaging sensor can acquisition frame rate be The image data of 1000fps;In the scene that may be used on the high-speed target identification and detection in production and living;High speed camera has concurrently Autonomous focusing function can automatically lead to target after the image processing module of multi-level heterogeneous parallel processing carries out image procossing It crosses focusing function and collects visual field middle position;
(2) while Image Acquisition transmission module is acquired image data, noise reduction process is carried out to image data, is added Speed is based on identification of the multi-level heterogeneous parallel processing image processing module to target;
(3) image processing module based on multi-level heterogeneous parallel processing has Pixel-level, block grade and distributed parallel Processing capacity can carry out flexible image block selection and be operable with multiple-instruction multiple-data pattern, accelerate at image data The speed of target identification in reason;
(4) the high speed tracing system based on multi-level heterogeneous parallel processing have high speed, highly integrated, low-power consumption, it is low into This advantage, can apply in fields such as production and living, scientific research, sports;
(5) it by designing corresponding target tracking algorithm, can realize under complex scene to high-speed moving object Robust tracking.
Description of the drawings
Fig. 1 is the structure diagram of the high speed tracing system of the present invention;
Fig. 2 is work flow diagram of the high speed tracing system of the present invention in tracing process;
Fig. 3 is flow of the image processing module of the present invention when carrying out multi-level heterogeneous parallel processing to image data Figure.
Specific embodiment
The invention discloses a kind of high speed tracing systems, perform control including high-speed image sampling device, two-dimentional servo and ring Module, two-dimentional servo module, image data acquiring transmission module and image processing module are answered, wherein:
High-speed image sampling device has autonomous focusing function, for example, high-speed camera, and pixel device is using at a high speed Cmos image sensor, picture read-out speed are 1000fps.Its optical lens uses the optical frames with autonomous focusing function After image processing module handles image data, autonomous, Real-time Focusing is realized according to environment and target algorithm for head, Ensure the image quality of high-speed image sampling device.
Two-dimentional servo performs control response module, and image processing module is connect with two-dimentional servo module, preferably through Bus mode is attached.
Two-dimentional servo module including the upper motor with step function and lower motor, has the high speed image independently focused Harvester is mechanically fixed on motor, and upper motor is installed on lower motor.Upper motor realizes that high speed image is adopted Acquisition means are rotated up and down, and lower motor realizes the three-dimensional rotation of entire two-dimentional servo module by rotating left and right.
Image processing module, framework are integrated with Pixel-level, block grade and distributed multi-stage heterogeneous parallel processing device, have Image Feature Detection, feature extraction, tagsort and deep learning algorithm convolution processing function, can efficiently, it is neatly complete Into each quasi-representative and the vision processing algorithm that at present develops;Its instruction set is regards based on the special of multi-tier Heterogeneous parallel processing architecture Feel process instruction collection, support Pixel-level, block grade and distributed parallel operation programming, and devise block processing unit circuit, Processing unit circuit, track circuit and cache control circuit are to realize instruction set function.
Described image processing module includes Pixel-level parallel processor, block grade parallel processor and distributed variable-frequencypump, By time-sharing multiplex, Pixel-level parallel processing and distributed variable-frequencypump are completed using identical hardware.Wherein at Pixel-level Pixel processing unit PE (ProcessingElement) array for managing device completes such as filtering, morphology, background subtraction, characteristic point The Pixel-levels parallel algorithms such as detection, feature generation, then at block processing unit (PPU) the access pixel in block grade parallel processor The image block in unit PE subarrays is managed to complete local shape factor algorithm, the spy to extracting is realized using Distributed Calculation Sign carries out Fast Classification, so as to fulfill the parallel processing based on multi-level isomery.
Display module realizes the connection with image processing module by HDMI (DP) interface.It is according to image processing module To the operation result of image data, display target and background information, and target is marked in real time.
Power entry module is that all modules of whole system carry out distribution, ensures the steady operation of system.
In said program, the high-speed camera machine high speed image of autonomous focusing function and integrated high speed imaging sensor passes Sensor is with the high speed imaging sensor more than or equal to 1000fps Image Acquisition speed;Its autonomous focusing function based on The image processor block of multi-level heterogeneous parallel processing has handled image data and has obtained after target position information according to environment And target requirement and carry out automatic focusing.
In said program, image data acquiring transmission module is to the height of autonomous focusing function and integrated high speed imaging sensor Imaging sensor in fast video camera connects entirely, and the high speed of the autonomous focusing function of real-time reception and integrated high speed imaging sensor is taken the photograph Camera the image collected data simultaneously carry out simple noise reduction process to data, and treated image data is passed by fiber optic communication It is defeated by the image processor block based on multi-level heterogeneous parallel processing.
In a preferred embodiment, it is total that the image processor block based on multi-level heterogeneous parallel processing includes system Line, sensor control block, sensor caching, program cache, network interface, for MPP calculating core and MPU microprocessors.Sensor caching is for storing the data from imaging sensor, in order to which processor repeatedly accesses.Program The algorithm routine for visual processes is stored in caching.Network interface is used to communicate with PC.Calculating core is multi-level isomery Parallel processing architecture provides main calculated performance and can realize that multiple-instruction multiple-data works.Referred to according to the control of MPU It enables, identical or entirely different manipulation of image data can be carried out between calculating core.
In said program, image processing module includes processor, and instruction set is based on multi-tier Heterogeneous parallel processing architecture Special visual processes instruction set, support the programming of Pixel-level, block grade and distributed parallel operation, and it is single to devise block processing First circuit, processing unit circuit and cache control circuit are to realize instruction set function;
In said program, Pixel-level parallel processing and distributed variable-frequencypump use identical hardware by time-sharing multiplex To complete;Preferably, Pixel-level parallel processing includes filtering, morphology, background subtraction, characteristic point detection and feature generation etc. Operation.
In said program, two-dimentional servo performs response module and the image processor mould of multi-level heterogeneous parallel processing Block connects, and two-dimentional servo performs response control after target position information is obtained after carrying out image real time transfer, calculates servo mould Block needs rotation angle, and send respective direction information with it is rotary pulsed.
In said program, two-dimentional servo module can realize full ambient engine 360 degree rotation, and perform response control with two-dimentional servo Module connect, when reflect rotation angle directional information with it is rotary pulsed be sent to servo module after, servo module will carry out phase The rotation in direction is answered, ensures that the high speed camera being mounted on two-dimentional servo module in real time includes target in camera fields of view.
In said program, display module passes through HDMI (DP) interfaces and the image procossing based on multi-level heterogeneous parallel processing Device module connects, and after the completion of image real time transfer, the image processor block of multi-level heterogeneous parallel processing is showing target Show and shown in module.
In said program, the high-speed camera machine of autonomous focusing function and integrated high speed imaging sensor in system can lead to It crosses USB3.0 and carries out real-time Communication for Power with host computer, the processor based on multi-level heterogeneous parallel processing can pass through PCIe protocol, net Mouth agreement carries out real-time Communication for Power with host computer.
The invention also discloses a kind of high speed method for tracing based on multi-level heterogeneous parallel processing, use as described above High speed tracing system realize that the high speed method for tracing includes the following steps:
When in the visual field of the high-speed image sampling device with autonomous focusing function described in tracking target and appearing in, image Information collection transmission module receives image data from the high-speed image sampling device, is transferred to after image noise reduction is handled Image processing module;
Image processing module passes through carries out Pixel-level, block grade and distributed variable-frequencypump to described image data, obtains institute The location information of tracking target is stated, and sends location information to the two-dimentional servo and performs response module, is aligned simultaneously Confidence breath is judged, is obtained focusing coefficient and is transmitted to the high-speed image sampling device for carrying autonomous focusing function, completes The autonomous focusing of high-speed image sampling device;
Two-dimentional servo performs response module and the location information for tracking target is calculated, and obtains directional information and rotation Gyration information;Two-dimentional servo module receives the directional information and rotation angle that response module is performed from two-dimentional servo After information, rotated accordingly;
Meanwhile the tracking target information is shown on display module by image processing module.
Preferably, above-mentioned high speed method for tracing is further comprising the steps of:
When showing the tracking target on display module, judge whether primary tracking target loses every some cycles It loses, if lost, two-dimentional servo performs control response module and two-dimentional servo module cooperation performs tracking action, searches The trace of the tracking target;If it is not, then two-dimentional servo performs control response module and two-dimentional servo module cooperation performs lock Fixed action, by the tracking target lock-on in the visual field of the high-speed image sampling device with autonomous focusing function.
In a preferred embodiment, the present invention is using the high speed camera with autonomous focusing function and with multi-level Heterogeneous parallel processing framework can field programming SPEED VISION target tracking processor;In order to preferably support autonomous focusing function High speed camera and with multi-level heterogeneous parallel processing framework can be between field programming SPEED VISION target tracking processor Work makes the high speed camera of autonomous focusing function under conditions of being acquired with high frame rate image, can the high speed of field programming regard Feel that target tracking processor can carry out high frame rate image processing, devise the high-speed target based on multi-level heterogeneous parallel processing Tracing system.The system has high integration, high-performance, low cost, the characteristic of low-power consumption, and passes through and design corresponding mesh Tracing algorithm is marked, is realized under complex scene to the robust tracking of high-speed moving object.
To make the objectives, technical solutions, and advantages of the present invention clearer, below in conjunction with specific embodiment, and reference Attached drawing, the present invention is described in further detail.
High-speed target tracing system overall structure figure based on multi-level heterogeneous parallel processing is as shown in Figure 1.Using this hair The method for tracing that bright high speed tracing system realizes high-speed target is, 1 in Fig. 1 occurs in tracking target, with from homophony In the visual field of the high speed camera of burnt function;Image information collecting transmission module is integrated in 1 in Fig. 1 from the high-speed cmos of camera Image data is received in imaging sensor, 2 be transferred to after simple image noise reduction process in Fig. 1, based on multi-level isomery The image processing module of parallel processing;Image processing module based on multi-level heterogeneous parallel processing passes through carries out pixel to image Grade, block grade and distributed variable-frequencypump, obtain the location information of high-speed target, and will be respectively transmitted after location information in Fig. 1 4, two-dimentional servo performs response module, while location information is judged, obtains focusing coefficient and passes in Fig. 1 1, in the camera of autonomous focusing function, complete the autonomous focusing of high speed camera;4 in Fig. 1, two-dimentional servo performs response control mould Block calculates the location information of target, obtains directional information and rotation angle information;5 in Fig. 1, two-dimentional servo module connects It receives after two-dimentional servo performs the directional information of response module and rotation angle pulse information, is revolved accordingly Turn;2 in Fig. 1 simultaneously, the image processing module based on multi-level heterogeneous parallel processing will be in high-speed target presentation of information to Fig. 1 3, on display module.
1. the high speed camera with autonomous focusing function
High speed camera with autonomous focusing function is by high speed imaging sensor, the optical lens with autonomous focusing function It is formed with high-speed image sampling transmission module, high-speed image sampling and the transmission of 1000fps can be completed, and can completed online certainly Homophony is burnt;Carry out data acquisition transmission after high speed camera to imaging sensor for image information collecting transmission module, and can be to figure As data carry out simple noise reduction process.
2. the image processing module based on multi-level heterogeneous parallel processing
Image processing module based on multi-level heterogeneous parallel processing is by pixel level processor, block level processor and distribution Formula parallel processing and MPU microprocessors are formed.
The image processing module of multi-level heterogeneous parallel processing can carry out original pixel-level image parallel processing, Local map As the image Parallel Processing and distributed variable-frequencypump between block.Image processing module is cached by imaging sensor, program cache, System bus and the calculating compositions such as core and MPU microprocessors.Wherein, imaging sensor buffer memory is acquired from imaging sensor The image arrived, the program that program cache storage handles image, system bus is as image processing program with calculating core Link and control calculate core carry out image procossing.It calculates to contain in core and can carry out Pixel-level parallel processing function Pixel level processor, can be into the block level processor of row block grade local shape factor, by time-sharing multiplex, pixel level processor and block Level processor accesses identical hardware to complete, and MPU is mainly used for system administration.Image based on multi-level heterogeneous parallel processing The main flow of processing module is:Imaging sensor is handled by Image Acquisition and after being transferred to sensor caching by Pixel-level Device completes the parallel algorithms of Pixel-levels such as filtering, feature generation, and then block level processor access is stored in pixel level processor Image block in array simultaneously provides program control instruction and performs local feature extraction algorithm and obtain partial statistics characteristic;In calculation block The feature that the mode of Distributed Parallel Computing extracts block level processor may be used after processing to classify.So as to fulfill picture Plain grade image processor and the multi-level isomerism parallel image processing operations of block level processor and distributed variable-frequencypump.
3. display module
Display module connects the image processing module of multi-level heterogeneous parallel processing, is responsible for display and determines high-speed target by it Image and its location information.
4. two-dimentional servo performs response module
Location information and angle information are sent to two-dimentional servo and held by the image processing module of multi-level heterogeneous parallel processing Row response module, two-dimentional servo perform the direction that response module is rotated by two-dimentional servo module needs are calculated With need the pulse that rotates and control two-dimentional servo module.
5. two-dimentional servo module
Space is divided into (x, y, z) three-dimensional planar, lower motor is the bottom of at by two-dimentional servo module by upper motor and lower motor form Plane is that (x, y) plane is moved, and the position that upper motor is rotated according to lower motor can be respectively in (x, z), (y, z) plane Movement, it is final to realize that two-dimentional servo module rotates freely in the region of 360 degree of three dimensions gamut.
6. Power entry module
Each module needs to stablize and continuous voltage, and in order to make system steady operation, Power entry module is the other each modules of system Supply stable power supply.
As a kind of application of the present invention, such as the target even high-speed motion of automatic identification and tracing movement Target, the target are that the automatic Classification and Identification of neural network classifier in the image processing module by the present invention comes out, and are being opened After the high-speed target tracing system for moving the present invention, system is automatically from the image of the high speed camera capture with autonomous focusing function In identify moving target, and complete it is subsequent tracking and locking work.
Particular embodiments described above has carried out the purpose of the present invention, technical solution and advantageous effect further in detail Describe in detail bright, it should be understood that the above is only a specific embodiment of the present invention, is not intended to restrict the invention, it is all Within the spirit and principles in the present invention, any modification, equivalent substitution, improvement and etc. done should be included in the protection of the present invention Within the scope of.

Claims (10)

1. a kind of high speed tracing system, which is characterized in that the high speed tracing system includes high-speed image sampling device, two dimension is watched Clothes perform control response module, two-dimentional servo module, image data acquiring transmission module and image processing module;Wherein:
High-speed image sampling device has autonomous focusing function;
The picture signal of high-speed image sampling device acquisition is input to image processing module by image data acquiring transmission module;
Two-dimentional servo performs control response module and generates two-dimentional servo module based on the control information that image processing module exports Control instruction simultaneously sends it to the two-dimentional servo module;
Described image processing module is integrated with Pixel-level, block grade and distributed multi-stage heterogeneous parallel processing unit, has image Feature detection, feature extraction, tagsort and deep learning convolution algorithm processing function;Wherein, described image processing mould The pixel processing unit PE arrays that block first passes through pixel level processor complete Pixel-level parallel algorithm, are then located parallel by block grade Block processing unit PPU in reason device accesses the image block in pixel level processor pixel processing unit PE arrays to complete local spy Extraction algorithm is levied, is finally realized using Distributed Calculation and Fast Classification is carried out to the local feature of extraction, generate the control Letter opens, so as to fulfill the parallel processing based on multi-level isomery.
2. high speed tracing system according to claim 1, which is characterized in that the high-speed image sampling device is takes the photograph at a high speed Camera, pixel device use high-speed cmos imaging sensor, and picture read-out speed is more than or equal to 1000fps;
Preferably, described image data acquisition transmission module connects the imaging sensor in the high-speed image sampling device entirely It connects, real-time reception described image sensor the image collected data simultaneously carry out noise reduction process, then pass through to described image data Fiber-optic communications traffic gives described image processing module.
3. high speed tracing system according to claim 1, which is characterized in that the two dimension servo performs control response module Image data processing module is connect with two-dimentional servo module by bus mode.
4. high speed tracing system according to claim 1, which is characterized in that the two dimension servo module includes having stepping The upper motor of function and lower motor, the high-speed image sampling device are fixed on the upper motor, and upper motor is installed on lower electricity On machine, so as to which the upper motor realizes being rotated up and down for the high-speed image sampling device, lower motor, which is realized, to be rotated left and right, from And entire two-dimentional servo module can realize three-dimensional rotation.
5. high speed tracing system according to claim 1, which is characterized in that described image processing module includes processor, Its instruction set is the special visual processes instruction set based on multi-tier Heterogeneous parallel processing architecture, supports Pixel-level, block grade and divides The programming of cloth parallel work-flow, and block processing unit circuit, processing unit circuit and cache control circuit are devised to realize Instruction set function;
Delay preferably, described image processing module further includes system bus, sensor control block, sensor caching, program It deposits and network interface, wherein, the sensor caching is for storing the data from high-speed image sampling device;Described program is delayed The algorithm routine for visual processes is stored in depositing;The network interface is used to communicate with PC.
6. high speed tracing system according to claim 1, which is characterized in that the Pixel-level parallel processing and distribution are simultaneously Row processing is completed by time-sharing multiplex using identical hardware;
Preferably, the Pixel-level parallel processing includes filtering, morphology, background subtraction, characteristic point detection and feature generation.
7. high speed tracing system according to claim 1, which is characterized in that the high speed tracing system further includes display mould Block, the display module are connect by HDMI/DP interfaces with image processing module;
The information that the display module to image obtain after operation according to image processing module, display target and background letter Breath, and target is marked in real time.
8. high speed tracing system according to claim 1, which is characterized in that the high speed tracing system further includes distribution mould Block, the Power entry module carry out distribution for system, ensure the steady operation of system.
9. high speed tracing system according to claim 1, which is characterized in that the high-speed image sampling device can pass through USB3.0 and host computer carry out real-time Communication for Power, and the processor in described image processing module can pass through PCIe protocol and/or net Mouth agreement carries out real-time Communication for Power with host computer.
10. a kind of high speed method for tracing based on multi-level heterogeneous parallel processing, uses such as claim 1 to 9 any one The high speed tracing system is realized, which is characterized in that the high speed method for tracing includes the following steps:
When in the visual field of the high-speed image sampling device with autonomous focusing function described in tracking target and appearing in, image information Acquisition transmission module receives image data from the high-speed image sampling device, and image is transferred to after image noise reduction is handled Processing module;
Image processing module passes through carries out Pixel-level, block grade and distributed variable-frequencypump to described image data, obtains described chase after The location information of track target, and send location information to the two-dimentional servo and perform response module, while align confidence Breath is judged, is obtained focusing coefficient and is transmitted to the high-speed image sampling device for carrying autonomous focusing function, completes at a high speed The autonomous focusing of image collecting device;
Two-dimentional servo performs response module and the location information for tracking target is calculated, and obtains directional information and rotation angle Spend information;Two-dimentional servo module receives the directional information and rotation angle information that response module is performed from two-dimentional servo Afterwards, it is rotated accordingly;
Meanwhile the tracking target information is shown on display module by image processing module;
Preferably, the high speed method for tracing is further comprising the steps of:
When showing the tracking target on display module, judge whether primary tracking target loses every some cycles, such as Fruit has been lost, then two-dimentional servo execution control response module and two-dimentional servo module cooperation performs tracking and act, described in search Track the trace of target;It is moved if it is not, then two-dimentional servo execution control response module performs to lock with two-dimentional servo module cooperation Make, by the tracking target lock-on in the visual field of the high-speed image sampling device with autonomous focusing function.
CN201711498701.2A 2017-12-29 2017-12-29 High-speed tracking system and method based on multi-level heterogeneous parallel processing Active CN108184062B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201711498701.2A CN108184062B (en) 2017-12-29 2017-12-29 High-speed tracking system and method based on multi-level heterogeneous parallel processing

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201711498701.2A CN108184062B (en) 2017-12-29 2017-12-29 High-speed tracking system and method based on multi-level heterogeneous parallel processing

Publications (2)

Publication Number Publication Date
CN108184062A true CN108184062A (en) 2018-06-19
CN108184062B CN108184062B (en) 2020-08-25

Family

ID=62549629

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201711498701.2A Active CN108184062B (en) 2017-12-29 2017-12-29 High-speed tracking system and method based on multi-level heterogeneous parallel processing

Country Status (1)

Country Link
CN (1) CN108184062B (en)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109741237A (en) * 2018-12-28 2019-05-10 中国科学院半导体研究所 Large-scale image data processing system and method
CN111932589A (en) * 2020-08-11 2020-11-13 沈阳帝信人工智能产业研究院有限公司 Data processing method, system and equipment
WO2023201842A1 (en) * 2022-04-19 2023-10-26 珠海欧比特宇航科技股份有限公司 High-bandwidth signal processing system, device, method, and storage medium

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20030095711A1 (en) * 2001-11-16 2003-05-22 Stmicroelectronics, Inc. Scalable architecture for corresponding multiple video streams at frame rate
CN101814137A (en) * 2010-03-25 2010-08-25 浙江工业大学 Driver fatigue monitor system based on infrared eye state identification
CN103020890A (en) * 2012-12-17 2013-04-03 中国科学院半导体研究所 Visual processing device based on multi-layer parallel processing
CN103544714A (en) * 2013-10-22 2014-01-29 中国科学院半导体研究所 Visual tracking system and method based on high-speed image sensor

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20030095711A1 (en) * 2001-11-16 2003-05-22 Stmicroelectronics, Inc. Scalable architecture for corresponding multiple video streams at frame rate
CN101814137A (en) * 2010-03-25 2010-08-25 浙江工业大学 Driver fatigue monitor system based on infrared eye state identification
CN103020890A (en) * 2012-12-17 2013-04-03 中国科学院半导体研究所 Visual processing device based on multi-layer parallel processing
CN103544714A (en) * 2013-10-22 2014-01-29 中国科学院半导体研究所 Visual tracking system and method based on high-speed image sensor

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109741237A (en) * 2018-12-28 2019-05-10 中国科学院半导体研究所 Large-scale image data processing system and method
CN109741237B (en) * 2018-12-28 2020-10-23 中国科学院半导体研究所 Large-scale image data processing system and method
CN111932589A (en) * 2020-08-11 2020-11-13 沈阳帝信人工智能产业研究院有限公司 Data processing method, system and equipment
WO2023201842A1 (en) * 2022-04-19 2023-10-26 珠海欧比特宇航科技股份有限公司 High-bandwidth signal processing system, device, method, and storage medium

Also Published As

Publication number Publication date
CN108184062B (en) 2020-08-25

Similar Documents

Publication Publication Date Title
CN110580732B (en) 3D information acquisition device
CN108154494B (en) A kind of image fusion system based on low-light and infrared sensor
CN107659774B (en) Video imaging system and video processing method based on multi-scale camera array
CN107610108B (en) Image processing method and device
CN105210368B (en) Background difference extraction element and background difference extracting method
CN104820418B (en) A kind of embedded vision system and its application method for mechanical arm
WO2019238114A1 (en) Three-dimensional dynamic model reconstruction method, apparatus and device, and storage medium
CN108184062A (en) High speed tracing system and method based on multi-level heterogeneous parallel processing
CN104184961A (en) Mobile device and system used for generating panoramic video
CN104270570A (en) Binocular video camera and image processing method thereof
CN105809711B (en) A kind of pig movement big data extracting method and its system based on video frequency tracking
CN105592269B (en) Moving target captures and tracking equipment and moving target capture and tracking
JP2011508293A (en) Vision sensor, system and method
EP1946567A4 (en) Device for generating three dimensional surface models of moving objects
Zheng et al. Deep learning for event-based vision: A comprehensive survey and benchmarks
CN111242025A (en) Action real-time monitoring method based on YOLO
CN106354264A (en) Real-time man-machine interaction system based on eye tracking and a working method of the real-time man-machine interaction system
CN107944420A (en) The photo-irradiation treatment method and apparatus of facial image
Popovic et al. Real-time implementation of Gaussian image blending in a spherical light field camera
CN107845056A (en) Fish eye images panorama generation method based on cylinder model
CN105791686B (en) Multiple mobile object captures and tracking equipment and multiple mobile object capture and tracking
CN107807777A (en) Multinuclear embedded television tracker human-computer interaction device and method
CN109089048B (en) Multi-lens panoramic linkage device and method
Wu et al. Real-time human posture reconstruction in wireless smart camera networks
CN104751455A (en) Crop image dense matching method and system

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant