CN107516296A - A kind of moving object detection tracking system and method based on FPGA - Google Patents
A kind of moving object detection tracking system and method based on FPGA Download PDFInfo
- Publication number
- CN107516296A CN107516296A CN201710554813.9A CN201710554813A CN107516296A CN 107516296 A CN107516296 A CN 107516296A CN 201710554813 A CN201710554813 A CN 201710554813A CN 107516296 A CN107516296 A CN 107516296A
- Authority
- CN
- China
- Prior art keywords
- data
- module
- ddr3
- image
- rgb
- 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.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 64
- 238000001514 detection method Methods 0.000 title claims abstract description 29
- 230000033001 locomotion Effects 0.000 claims abstract description 69
- 238000006243 chemical reaction Methods 0.000 claims abstract description 55
- 230000015654 memory Effects 0.000 claims abstract description 35
- 238000012545 processing Methods 0.000 claims abstract description 30
- 241001269238 Data Species 0.000 claims abstract description 28
- 238000005260 corrosion Methods 0.000 claims abstract description 22
- 230000007797 corrosion Effects 0.000 claims abstract description 22
- 239000000872 buffer Substances 0.000 claims abstract description 4
- 238000001914 filtration Methods 0.000 claims description 32
- 238000000605 extraction Methods 0.000 claims description 20
- 238000007639 printing Methods 0.000 claims description 15
- 230000000694 effects Effects 0.000 claims description 10
- 238000012546 transfer Methods 0.000 claims description 9
- 230000001360 synchronised effect Effects 0.000 claims description 8
- 238000006073 displacement reaction Methods 0.000 claims description 2
- 208000033749 Small cell carcinoma of the bladder Diseases 0.000 description 4
- 238000011161 development Methods 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 4
- 201000007710 urinary bladder small cell neuroendocrine carcinoma Diseases 0.000 description 4
- 238000004364 calculation method Methods 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- 238000004891 communication Methods 0.000 description 2
- 230000005055 memory storage Effects 0.000 description 2
- 230000000877 morphologic effect Effects 0.000 description 2
- 230000006399 behavior Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 239000013078 crystal Substances 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000011549 displacement method Methods 0.000 description 1
- 238000005538 encapsulation Methods 0.000 description 1
- 230000003628 erosive effect Effects 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000004377 microelectronic Methods 0.000 description 1
- 230000008447 perception Effects 0.000 description 1
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 1
- 238000005303 weighing Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/20—Image enhancement or restoration by the use of local operators
- G06T5/30—Erosion or dilatation, e.g. thinning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T1/00—General purpose image data processing
- G06T1/20—Processor architectures; Processor configuration, e.g. pipelining
-
- G06T5/70—
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/20—Analysis of motion
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20024—Filtering details
- G06T2207/20032—Median filtering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20036—Morphological image processing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30232—Surveillance
Abstract
The present invention relates to a kind of moving object detection tracking system and method based on FPGA, belong to technical field of image processing.Present invention view data first carries out data conversion and then divides two-way to export;Enter outside DDR3 memory modules by data input DDR3 fifo modules all the way to be cached, another way is cached after RGB turns gray proces, median filter process into DDR3 memory modules;The two frame greyscale image datas continuously exported from DDR3 buffers obtain the coordinate value of Motion mask after carrying out inter-frame difference computing, binary conversion treatment, corrosion expansion process, feed back to again on rgb format coloured image to obtain templates of moving objects, it is deposited into the storage of FPGA ram in slice, finally using its template matches of taking exercises, to realize the tracking of color motion target.The present invention has reached balance well in target identification speed, accuracy, cost, power consumption etc..
Description
Technical field
The present invention relates to a kind of moving object detection tracking system and method based on FPGA, belongs to image processing techniques neck
Domain.
Background technology
Vision is that people recognize the most important means in the world, machine vision and the key technology that perception is automatic field,
Computer Vision is then the basis of visually-perceptible, and as the hot issue of research and application and development and among these with motion
Target detection and the focus being tracked as in focus.
Video Supervision Technique rises and late 1970s, from the information of the simple displaying scene of beginning, reduces people's
Workload, to the detecting and tracking for the mobile target that gradually begins one's study, there is provided more information.At present, video object detection and with
Track has widely been used in the various aspects of people's life, such as traffic, national defence, Aero-Space etc..
At present, the research and realization of the detecting and tracking algorithm of moving target, usually used is computer platform.Although meter
Advanced language programming can be used by calculating machine platform, but computer platform has that volume is big, power consumption is high, processing speed is slow, poor real
The shortcomings of, these all constrain the further development of research.
With the development of modern microelectronic technology, field programmable gate array(That is FPGA)Device arises at the historic moment, and makes in the past
The digital display circuit of bulky complex is minimized, and makes it possible on-chip system.Substantial amounts of can configure is provided with inside FPGA to patrol
Volume resource, PLL resources, RAM resources, DSP operation unit etc. on piece.And there is parallel hardware configuration and pipeline processes side
Formula, the calculating speed on software can be broken through, realize the hardware-accelerated of algorithm.So FPGA to be used for the detection of moving target
Tracking is possible, and real time implementation, miniaturization or miniaturization for such vision processing system have great importance and practicality
Value.
The content of the invention
The technical problem to be solved in the present invention is:The present invention provides a kind of moving object detection tracking system based on FPGA
And method, to solve, the detecting and tracking system and method speed of service of the existing moving target based on computer platform is slow, volume
Greatly, the problem of power consumption is high, and the detecting and tracking system and method for the existing moving target based on ARM platforms slow-footed are asked
Topic, improve the accuracy rate of moving object detection;Realize moving object detection in the flat of speed, device volume and power consumption etc.
Weighing apparatus.
The technical scheme is that:A kind of moving object detection tracking system and method based on FPGA, including CMOS take the photograph
As head module, data conversion module, RGB, to turn gray scale module, medium filtering module, data input DDR3 fifo modules, data defeated
Go out DDR3 fifo modules, inter-frame difference and binarization block, corrosion expansion module, Motion mask extraction module, Motion mask
With module, VGA displays driving and serially printing module, DDR3 Read-write Catrol modules;
The CMOS camera module collection view data is transmitted to data conversion module, carries out data conversion and then divides two-way defeated
Go out;It is directly entered outside DDR3 memory modules by data input DDR3 fifo modules all the way to be cached, another way is passed through
RGB turns gray scale module RGB and turned after gray proces, medium filtering module median filter process again by data input DDR3 FIFO
Module enters outside DDR3 memory modules and cached, and wherein DDR3 Read-write Catrols module is used for the outside DDR3 of control data disengaging
Memory module;The two frame greyscale image datas continuously exported by data output DDR3 fifo modules from DDR3 memory modules
Inter-frame difference computing, binary conversion treatment, corrosion expansion process are carried out by inter-frame difference and binarization block, corrosion expansion module
The coordinate value of Motion mask is obtained afterwards, then by Motion mask extraction module, the coordinate value of Motion mask is fed back into rgb format
To obtain templates of moving objects on coloured image, the storage of FPGA ram in slice is deposited into, finally passes through Motion mask using it
Taken exercises template matches with module, to realize the tracking of color motion target, and its movement locus can be depicted and passed through
VGA displays driving and serially printing module carry out upper computer software and show and print movement locus coordinate values.
CMOS camera module:For by the rgb format image information of collection, including more bit image datas, 1 bit field
Synchronizing signal, 1 bit line synchronising signal, CMOS gathered data synchronizing clock signals, all deliver to data conversion module;
Data conversion module:The conversion of data bit width is carried out for 8 bit datas that will collect, be converted into 16 bit datas or
The bit data of person 24, and deliver to RGB respectively and turn gray scale module and DDR3 storage moulds are sent into by data input DDR3 fifo modules
Block is cached;Because the data of CMOS camera acquisition module collection are 8 bit datas, but in rgb format coloured image,
One pixel is generally 16 bit datas or 24 bit datas.So to carry out the conversion of data bit width.Equally, CMOS takes the photograph
As the synchronous acquisition clock of head is also corresponding slack-off 2 times or 3 times.The field line synchronising signal not dealt with will also pass through corresponding
Delay form new field line synchronising signal, these new data then are delivered into RGB turns gray scale module.
RGB turns gray scale module:, can for the rgb format image of 16 or 24 to be changed into 8 grayscale format images
Amount of calculation is reduced for later calculating.The greyscale image data converted is delivered into medium filtering module;
Medium filtering module:After rgb format coloured image is changed into grayscale format image, medium filtering is carried out, by intermediate value
View data after filtering delivers to data input DDR3 fifo modules;Because rgb format coloured image changes into grayscale format figure
As after, the inside can noisy presence, noise on image quality has a great impact, and progress medium filtering can not only remove orphan
Spot noise, moreover it is possible to keep the local edge of image, image will not produce significant fuzzy.Become using picture quality after medium filtering
The high processing more to prepare behind aspect.
Data input DDR3 fifo modules:This fifo module is mainly for the treatment of across clocked data transfer and data bit
Width conversion;Data after medium filtering are sent into DDR3 read-write clock zone and data bit width from synchronous acquisition clock zone
Change into the same bit wide length of same DDR3 reading-writing ports;Data after processing are delivered into outside DDR3 memory modules;
DDR3 memory modules:Because image data amount is big, FPGA piece memory storage resources use far from enough, outside you must use
Memory come cached with solve the problems, such as big data across clock transfer;
Data output DDR3 fifo modules:This fifo module turns mainly for the treatment of across clocked data transfer and data bit width
Change;Data are from DDR3 read-write clock zone feeding VGA read clocks domain and the same from DDR3 reading-writing ports data bit width
Bit wide change into the bit wide size of a pixel data;Data after processing are delivered into inter-frame difference and binarization block;
Inter-frame difference and binarization block:Including interframe subtraction module, binary conversion treatment module;Interframe subtraction module is by continuous two
The data that frame greyscale image data subtracts each other to obtain deliver to binary conversion treatment module and carry out binary conversion treatment;By inter-frame difference and two
Value module goes out moving target for Preliminary detection, and the data after processing are delivered into corrosion expansion module;If do not move mesh
Target is present, then the data of continuous two two field pictures correspondence position can be relatively.On the contrary, in some position of continuous two frame
View data can have a greater change, so tentatively can detect moving target by inter-frame difference and binarization block.
Corrode expansion module:For removing the noise spot in the image after inter-frame difference and binaryzation and to remove cavity existing
As;Data after processing are delivered into Motion mask extraction module;Image by difference and binaryzation output occurs that cavity is existing
As and noise spot, processing below can be had an impact, so to remove noise spot in image simultaneously by corroding expansion module
Remove cavitation.Data after processing are delivered into Motion mask extraction module.
Motion mask extraction module:Moving target is extracted according to block size, the moving target detected is big with template
Small deposit FPGA on-chip memories, to be supplied to Motion mask matching module;
Motion mask matching module:Each two field picture after being matched using obtained Motion mask, calculates optimal match point,
Reach tracking effect;And the data that processing can be gone out deliver to VGA displays driving and serially printing module;
VGA displays driving and serially printing module:For tracking result to be presented directly into user;VGA displays are used to show
Tracking effect and movement locus, serially printing are used to print movement locus numerical value.
A kind of moving object detection tracking based on FPGA, methods described step are:View data carries out data and turned
Change and then divide two-way to export;DDR3 memory modules are directly entered all the way to be cached, another way by RGB turn gray proces, in
Cached after value filtering processing into DDR3 memory modules;The two frame greyscale image datas continuously exported from DDR3 buffers
The coordinate value of Motion mask is obtained after carrying out inter-frame difference computing, binary conversion treatment, corrosion expansion process, then feeds back to RGB lattice
To obtain templates of moving objects on formula coloured image, the storage of FPGA ram in slice is deposited into, finally utilizes its template of taking exercises
Match somebody with somebody, to realize the tracking of color motion target, and its movement locus and upper computer software display movement locus can be depicted
Coordinate values.
When described image data carry out data conversion and then divide the two-way to export, the method for data conversion is spliced using displacement
Method.Such as 8 bit data inputs are required, 16 bit data outputs.Two 8 bit datas are exactly spliced into one 16 ratio
Special data.Then clock rate becomes original speed half all the time.
The method that the RGB turns gray proces is that RGB changes into YCbCr format, then only takes its Y-component to obtain gray-scale map
Picture;The formula that RGB turns YCbCr is as follows:
Y= 0.183R + 0.614G + 0.062B + 16;
Cb = -0.101R – 0.338G + 0.439B + 128;
Cr = 0.439R – 0.399G – 0.040B + 128;
Because Verilog can not carry out floating-point operation, therefore first expand 256 times, in 8 bit-wises that move right, after conversion
Formula is:
Y= (47R + 157G + 16B + 4096)>> 8;
Cb = (26R – 86G + 112B + 32768)>> 8;
Cr = (112R – 102G – 10B + 32768)>> 8;
Finally, RGB Three-channel datas originally all take the numerical value of Y-component, and image is just converted into grayscale format from rgb format.
The method of the median filter process includes;First, three pixels of every row are ranked up respectively;Then, it is right
The sequence that three row pixels obtain is handled, i.e. minimum value in three maximums of extraction, the maximum in three minimum values, with
And the median of three medians;Finally, by obtained in second step three values, intermediate value is taken again, has thus finally given 9
The median of individual pixel.
The method of inter-frame difference computing and binary conversion treatment is exactly that data are subtracted each other and then are compared with threshold value, more than threshold value
Complete 1 is just set to as white, it is black to be otherwise just set to full 0, and wherein threshold value takes 25.
Corroding the method for expansion process is:Corrosion treatment is following three step, first, with 3 × 3 window modules to input two
Value image traversal;Then, AND operation all is done with 1 to the total data in 3 × 3 windows;Finally, as long as there is pixel
Value is 0, is as a result 0;1 is all, is as a result only 1;Expansion process equally point three steps;First, with 3 × 3 window modules to input
Binary image travels through;Then, inclusive-OR operation all is done with 1 to the total data in 3 × 3 windows;Finally, as long as there is a pixel
Numerical value be 1, be as a result 1;0 is all, is as a result just 0.
Templates of moving objects extraction method be:To numerical value in binary image for 1 transverse and longitudinal coordinate a little carry out
Statistics, calculate minimum row where target, maximum row and minimum row, maximum column and barycenter;Then this parameter is returned to
The coloured image of rgb format.
Motion mask matching method be:Absolute value sum algorithm, i.e. SAD matching algorithms;It is empty for RGB color image
Between, its formula is as follows:
C = ∑( |Rx– Rx+1| + |Gx– Gx+1| + |Bx– Bx+1| )
The minimum point of numerical value, is exactly optimal match point;
Wherein in above-mentioned formula represent two continuous frames image same position corresponding to tri- components of RGB numerical value subtract each other absolute value it
With R=red, G=green, B=blue.
The beneficial effects of the invention are as follows:
The present invention determines moving target using frame difference method, can adapt to the detection under different background.And it have passed through filtering, two
The processing such as value, morphology so that moving object detection is more accurate.FPGA parallel pipelines are taken when template matches
Processing mode, matching speed can be greatly speeded up, and be the matching carried out under RGB color image, colored fortune can be realized
The tracking of moving-target, and movement locus is depicted, and trajectory coordinates are shown on host computer;
Present system can be parallel, and overall operation speed is faster than PC platform;In terms of volume, the present invention adds for circuit board to be taken the photograph
It is certainly smaller than PC platform by volume as head is added some points encapsulation;In terms of power consumption, much lower than PC platform of the present invention;With arm platforms pair
Than faster than ARM platform in speed.
Brief description of the drawings
Fig. 1 is the system block diagram of the present invention;
Fig. 2 is flow chart of the method for the present invention.
Embodiment
Below in conjunction with the accompanying drawings and specific embodiment, the invention will be further described.
Embodiment 1:As shown in Figure 1-2, a kind of moving object detection tracking system and method based on FPGA, including CMOS
Camera module, data conversion module, RGB turn gray scale module, medium filtering module, data input DDR3 fifo modules, data
Export DDR3 fifo modules, inter-frame difference and binarization block, corrosion expansion module, Motion mask extraction module, Motion mask
Matching module, VGA displays driving and serially printing module, DDR3 Read-write Catrol modules;
The CMOS camera module collection view data is transmitted to data conversion module, carries out data conversion and then divides two-way defeated
Go out;It is directly entered outside DDR3 memory modules by data input DDR3 fifo modules all the way to be cached, another way is passed through
RGB turns gray scale module RGB and turned after gray proces, medium filtering module median filter process again by data input DDR3 FIFO
Module enters outside DDR3 memory modules and cached, and wherein DDR3 Read-write Catrols module is used for the outside DDR3 of control data disengaging
Memory module;The two frame greyscale image datas continuously exported by data output DDR3 fifo modules from DDR3 memory modules
Inter-frame difference computing, binary conversion treatment, corrosion expansion process are carried out by inter-frame difference and binarization block, corrosion expansion module
The coordinate value of Motion mask is obtained afterwards, then by Motion mask extraction module, the coordinate value of Motion mask is fed back into rgb format
To obtain templates of moving objects on coloured image, the storage of FPGA ram in slice is deposited into, finally passes through Motion mask using it
Taken exercises template matches with module, to realize the tracking of color motion target, and its movement locus can be depicted and passed through
VGA displays driving and serially printing module carry out upper computer software and show and print movement locus coordinate values.
As the further scheme of the present invention, in addition to Clock management module, SCCB communication modules;
Clock management module:Crystal oscillator 50MHZ clocks pass through Clock management DCM IP kernels on plate, produce four that whole system needs
Clock, it is global system clock, CMOS camera driving clock, DDR3 memory modules driving clock and VGA display modules respectively
Drive clock.
SCCB communication modules:Effect is for driving CMOS camera module.SCCB agreements one share 2 lines, respectively
It is clock line and data wire, according to the agreement of regulation, carries out data transmission.Using SCCB agreements, related CMOS register counts
Value delivers to CMOS camera module, to drive CMOS camera to work.
CMOS camera module:For by the rgb format image information of collection;Specifically include more bit image datas, 1
Bit field sync signal, 1 bit line synchronising signal, CMOS gathered data synchronizing clock signals, all deliver to data conversion module;
The movable information of the target color of CMOS camera shooting(Track and coordinate values etc.), dynamically it is presented on VGA displays simultaneously
It is transferred in host computer;The 500W high definitions CMOS that camera image acquisition module can produce using Omnivision companies
Camera OV5640, fpga chip can use the XC6SLX16 of the Spartan6 series of Xilinx companies product;
Data conversion module:The conversion of data bit width is carried out for 8 bit datas that will collect, be converted into 16 bit datas or
The bit data of person 24, and deliver to RGB respectively and turn gray scale module and DDR3 storage moulds are sent into by data input DDR3 fifo modules
Block is cached;Because the data of CMOS camera acquisition module collection are 8 bit datas, but in rgb format coloured image,
One pixel is generally 16 bit datas or 24 bit datas.So to carry out the conversion of data bit width.Equally, CMOS takes the photograph
As the synchronous acquisition clock of head is also corresponding slack-off 2 times or 3 times.The field line synchronising signal not dealt with will also pass through corresponding
Delay form new field line synchronising signal, these new data then are delivered into RGB turns gray scale module.
Specifically, according to 1 bit(I.e. 1)Field sync signal, 1 bit line synchronising signal and synchronous acquisition clock, will be more
Bit image data carries out bit width conversion, is typically translated into the bit wide of a pixel number evidence, such as RGB565 of 16 bit wides or
The RGB888 of 24 bit wides.The data for producing 1 new bit field sync signal, 1 new bit line synchronising signal and 1 bit simultaneously have
Imitate signal.And will it is new caused by more bit image datas, 1 new bit field sync signal, 1 new bit line synchronising signal, same
Step collection clock and 1 bit data useful signal deliver to RGB and turn gray scale module and data input DDR3 fifo modules respectively.
RGB turns gray scale module:, can for the rgb format image of 16 or 24 to be changed into 8 grayscale format images
Amount of calculation is reduced for later calculating.The greyscale image data converted is delivered into medium filtering module;
Specifically, according to 1 new bit field sync signal, 1 new bit line synchronising signal, 1 bit data valid signal and
Synchronous acquisition clock, new more bit rgb format data are changed into grayscale format image.Rgb format turns grayscale format can be first
YCbCr format is changed into by RGB, then only takes its Y-component to obtain gray level image.The formula that RGB turns YCbCr is as follows:
Y= 0.183R + 0.614G + 0.062B + 16
Cb = -0.101R – 0.338G + 0.439B + 128
Cr = 0.439R – 0.399G – 0.040B + 128
Because Verilog can not carry out floating-point operation, therefore can first expand 256 times, in 8 bit-wises that move right, conversion
Formula afterwards is:
Y= (47R + 157G + 16B + 4096)>> 8
Cb = (26R – 86G + 112B + 32768)>> 8
Cr = (112R – 102G – 10B + 32768)>> 8
To synchronously it believe by transformed grayscale format data and caused 1 new bit field sync signal, 1 new bit-rows
Number, 1 new bit data useful signal and synchronous acquisition clock deliver to 3 × 3 window modules.
3 × 3 window modules:The module is and medium filtering module together cooperating.In the exploitation of altera corp
It can be built in external member using Shift Register IP kernels.In addition, also have by RAM or FIFO to build.
Medium filtering module:After rgb format coloured image is changed into grayscale format image, medium filtering is carried out, will
View data after medium filtering delivers to data input DDR3 fifo modules;Because rgb format coloured image changes into gray scale lattice
After formula image, the inside can noisy presence, noise on image quality has a great impact, and carrying out medium filtering can not only go
Except acnode noise, moreover it is possible to keep the local edge of image, image will not produce significant fuzzy.Use image matter after medium filtering
Processing of the quantitative change height more to prepare behind aspect.Using 3 × 3 window modules, the purpose of median filtering algorithm only require 3 × 3
The median of pel array.Specific method only needs three steps, and first, three pixels of every row are arranged respectively
Sequence.Then, the sequence obtained to three row pixels is handled, i.e. minimum value in three maximums of extraction, in three minimum values
Maximum, and the median of three medians.Finally, by obtained in second step three values, intermediate value is taken again, thus
The median of 9 pixels is finally given.Using Verilog HDL parallel characteristics, this process only needs three clocks.
Likewise, field sync signal, line synchronising signal and data valid signal also want corresponding three clocks of delay to obtain new field
Synchronizing signal, new line synchronising signal and new data valid signal.And these new data-signals are delivered into data input
DDR3 fifo modules.
Data input DDR3 fifo modules:This fifo module is mainly for the treatment of across clocked data transfer and data bit
Width conversion;Data after medium filtering are sent into DDR3 read-write clock zone and data bit width from synchronous acquisition clock zone
Change into the same bit wide length of same DDR3 reading-writing ports;Data after processing are delivered into outside DDR3 memory modules;
DDR3 memory modules:Because image data amount is big, FPGA piece memory storage resources use far from enough, outside you must use
Memory come cached with solve the problems, such as big data across clock transfer;
DDR3 Read-write Catrol modules:The main function of this module is the outside DDR3 memory modules of control data disengaging, in FPGA
Corresponding IP kernel can be typically integrated in development kit, has corresponding user interface.As long as these user interfaces are correctly controlled, just
DDR3 memory modules can very easily be read and write.Here DDR3 memory modules are arranged to 2 input ports and 3 output ports, and 2
What individual input port was respectively that a port enters is rgb format coloured picture view data, and what a port was entered is by grey scale change
And the grayscale format view data after medium filtering;3 output ports be respectively 1 port go out be rgb format colour
View data, what 2 ports went out is two continuous frames greyscale image data.These data deliver to data output DDR3 fifo modules.
Data output DDR3 fifo modules:This fifo module is mainly for the treatment of across clocked data transfer and data bit
Width conversion;Data are sent into VGA read clocks domain from DDR3 read-write clock zone and data bit width from DDR3 reading-writing ports
The same bit wide changes into the bit wide size of a pixel data;Data after processing are delivered into interframe subtraction module;
Interframe subtraction module:The effect of this module is to detect moving target.Specific formula is as follows:
D(x,y) = |Fx(x,y) – Fx+1(x,y)|
The data that two continuous frames greyscale image data subtracts each other to obtain are delivered into binary conversion treatment module.
Binary conversion treatment module:This module is to more preferably more obviously come out moving target indication.If greater than
Threshold value is set to 1, is otherwise set to 0.Specific formula is as follows:
Data after binaryzation are delivered into corrosion expansion process module.
The data that two continuous frames greyscale image data subtracts each other to obtain are delivered to binary conversion treatment module and entered by interframe subtraction module
Row binary conversion treatment;Go out moving target for Preliminary detection by inter-frame difference and binarization block, the data after processing are sent
To corrosion expansion module;If without the presence of moving target, then the data of continuous two two field pictures correspondence position, which can compare, to be connect
Closely.On the contrary, some location drawing picture data in continuous two frame can have a greater change, so by inter-frame difference and binaryzation
Module tentatively can detect moving target.
Corrode expansion module:For removing the noise spot in the image after inter-frame difference and binaryzation and to remove cavity existing
As;Data after processing are delivered into Motion mask extraction module;Image by difference and binaryzation output occurs that cavity is existing
As and noise spot, processing below can be had an impact, thus need Morphological scale-space reduce cavity and remove isolated noise
Point.Here Morphological scale-space is exactly to first carry out erosion operation to carry out expansion process again.Corrosion expansion process needs also exist for 3 × 3
Window module.Corrosion treatment is specifically divided into three steps:First, the binary image of input is traveled through with 3 × 3 window modules;Then,
AND operation all is done with 1 to the total data in 3 × 3 windows;Finally, it is as a result 0 as long as the value for having a pixel is 0.It is all
1, as a result it is only 1.Expansion process equally point three steps:First, the binary image of input is traveled through with 3 × 3 window modules;So
Afterwards, inclusive-OR operation all is done with 1 to the total data in 3 × 3 windows;Finally, as long as the numerical value for having a pixel is 1, as a result for
1.0 is all, is as a result just 0.Data after treated are delivered to moving target recognition template.
Motion mask extraction module:Moving target is extracted according to block size, the moving target detected is big with template
Small deposit FPGA on-chip memories, to be supplied to Motion mask matching module;Specific method is
1 transverse and longitudinal coordinate a little counted, calculate minimum row where target, maximum row and minimum row, maximum column and matter
The heart.Then the coloured image of this parameter feedback to rgb format, such moving target are just extracted, but from
From the aspect of resource considers two in track speed and FPGA pieces, this module is not conducive to excessive.Can according to actual conditions and resource come
Select the template of suitable size.The template of extraction delivers to the storage of FPGA ram in slice and finally delivers to Motion mask from ram in slice storage
Matching module.
Motion mask matching module:Each two field picture after being matched using obtained Motion mask, calculates optimal
With point, reach tracking effect;And the data that processing can be gone out deliver to VGA displays driving and serially printing module;Template matches are calculated
Method has many kinds, used here as most suitable FPGA matching algorithm, i.e. absolute value sum algorithm(SAD).It is colored for rgb format
Image space, its formula are as follows:
C = ∑( |Rx– Rx+1| + |Gx– Gx+1| + |Bx– Bx+1| )
FPGA parallel behavior can be utilized herein, take flowing water technology to accelerate computing.Then, it is exactly to similarity measurements
The result of calculation of amount is compared, and finds the position of SAD minimum points, and the point is exactly optimal match point.Finally by data result
Deliver to VGA display modules and serially printing module.
VGA display modules:The effect of the module is for driving VGA displays, and final result is presented on into VGA displays
On.First, the resolution sizes of VGA displays are determined, then provide corresponding driving clock again, field sync signal, row is synchronously
Signal and process handle obtained final image data above, and moving target quilt can be just shown on rgb format coloured image
The effect of tracking and the movement locus of moving target.
Serially printing module:The effect of the module is that the track data of moving target motion is shown on upper computer software,
That is moving target center-of-mass coordinate value.The baud rate of serial ports is first determined, correctly description, which sends and receives sequential, just can correctly exist
Continually changing center-of-mass coordinate numerical value is printed in upper computer software.
A kind of moving object detection tracking based on FPGA, methods described step are:View data carries out data and turned
Change and then divide two-way to export;DDR3 memory modules are directly entered all the way to be cached, another way by RGB turn gray proces, in
Cached after value filtering processing into DDR3 memory modules;The two frame greyscale image datas continuously exported from DDR3 buffers
The coordinate value of Motion mask is obtained after carrying out inter-frame difference computing, binary conversion treatment, corrosion expansion process, then feeds back to RGB lattice
To obtain templates of moving objects on formula coloured image, the storage of FPGA ram in slice is deposited into, finally utilizes its template of taking exercises
Match somebody with somebody, to realize the tracking of color motion target, and its movement locus and upper computer software display movement locus can be depicted
Coordinate values.
As the further scheme of the present invention, when described image data carry out data conversion and then divide the two-way to export, data
The method of conversion is using displacement splicing method.Such as 8 bit data inputs are required, 16 bit data outputs.It is exactly two 8
Bit data is spliced into 16 bit datas.Then clock rate becomes original speed half all the time.
As the further scheme of the present invention, the method that the RGB turns gray proces is that RGB changes into YCbCr format, then
Its Y-component is only taken to obtain gray level image;The formula that RGB turns YCbCr is as follows:
Y= 0.183R + 0.614G + 0.062B + 16;
Cb = -0.101R – 0.338G + 0.439B + 128;
Cr = 0.439R – 0.399G – 0.040B + 128;
Because Verilog can not carry out floating-point operation, therefore first expand 256 times, in 8 bit-wises that move right, after conversion
Formula is:
Y = (47R + 157G + 16B + 4096)>> 8;
Cb = (26R – 86G + 112B + 32768)>> 8;
Cr = (112R – 102G – 10B + 32768)>> 8;
Finally, RGB Three-channel datas originally all take the numerical value of Y-component, and image is just converted into grayscale format from rgb format.
As the further scheme of the present invention, the method for the median filter process includes;First, respectively to the three of every row
Individual pixel is ranked up;Then, the sequence obtained to three row pixels is handled, that is, extracts the minimum value in three maximums,
Maximum in three minimum values, and the median of three medians;Finally, by obtained in second step three values, again
Intermediate value is taken, has thus finally given the median of 9 pixels.
As the further scheme of the present invention, the method for inter-frame difference computing and binary conversion treatment is exactly then data are subtracted each other
It is compared with threshold value, complete 1 is just set to more than threshold value as white, it is black to be otherwise just set to full 0, and wherein threshold value takes 25.
As the further scheme of the present invention, corroding the method for expansion process is:Corrosion treatment is following three step, first,
The binary image of input is traveled through with 3 × 3 window modules;Then, "AND" fortune is all done with 1 to the total data in 3 × 3 windows
Calculate;Finally, it is as a result 0 as long as the value for having a pixel is 0;1 is all, is as a result only 1;Expansion process equally point three steps;It is first
First, the binary image of input is traveled through with 3 × 3 window modules;Then, the total data in 3 × 3 windows is all done with 1
Inclusive-OR operation;Finally, it is as a result 1 as long as the numerical value for having a pixel is 1;0 is all, is as a result just 0.
As the further scheme of the present invention, the method for templates of moving objects extraction is:It is 1 to numerical value in binary image
Transverse and longitudinal coordinate a little counted, calculate minimum row where target, maximum row and minimum row, maximum column and matter
The heart;Then this parameter is returned to the coloured image of rgb format.
As the further scheme of the present invention, the method for Motion mask matching is:Absolute value sum algorithm, i.e. SAD matchings
Algorithm;It is as follows for RGB color image space, its formula:
C = ∑( |Rx– Rx+1| + |Gx– Gx+1| + |Bx– Bx+1| )
The minimum point of numerical value, is exactly optimal match point;
Wherein in above-mentioned formula represent two continuous frames image same position corresponding to tri- components of RGB numerical value subtract each other absolute value it
With R=red, G=green, B=blue.
The specific embodiment of the present invention is explained in detail above in conjunction with accompanying drawing, but the present invention is not limited to above-mentioned reality
Example is applied, can also be on the premise of present inventive concept not be departed from those of ordinary skill in the art's possessed knowledge
Various changes can be made.
Claims (10)
- A kind of 1. moving object detection tracking system based on FPGA, it is characterised in that:Turn including CMOS camera module, data Mold changing block, RGB turn gray scale module, medium filtering module, data input DDR3 fifo modules, data output DDR3 FIFO moulds Block, inter-frame difference and binarization block, corrosion expansion module, Motion mask extraction module, Motion mask matching module, VGA show Show driving and serially printing module, DDR3 Read-write Catrol modules;The CMOS camera module collection view data is transmitted to data conversion module, carries out data conversion and then divides two-way defeated Go out;It is directly entered outside DDR3 memory modules by data input DDR3 fifo modules all the way to be cached, another way is passed through RGB turns gray scale module RGB and turned after gray proces, medium filtering module median filter process again by data input DDR3 FIFO Module enters outside DDR3 memory modules and cached, and wherein DDR3 Read-write Catrols module is used for the outside DDR3 of control data disengaging Memory module;The two frame greyscale image datas continuously exported by data output DDR3 fifo modules from DDR3 memory modules Inter-frame difference computing, binary conversion treatment, corrosion expansion process are carried out by inter-frame difference and binarization block, corrosion expansion module The coordinate value of Motion mask is obtained afterwards, then by Motion mask extraction module, the coordinate value of Motion mask is fed back into rgb format To obtain templates of moving objects on coloured image, the storage of FPGA ram in slice is deposited into, finally passes through Motion mask using it Taken exercises template matches with module, to realize the tracking of color motion target, and its movement locus can be depicted and passed through VGA displays driving and serially printing module carry out upper computer software and show and print movement locus coordinate values.
- 2. the moving object detection tracking system according to claim 1 based on FPGA, it is characterised in that:CMOS camera module:For by the rgb format image information of collection, including more bit image datas, 1 bit field synchronization Signal, 1 bit line synchronising signal, CMOS gathered data synchronizing clock signals, all deliver to data conversion module;Data conversion module:The conversion of data bit width is carried out for 8 bit datas that will collect, be converted into 16 bit datas or The bit data of person 24, and deliver to RGB respectively and turn gray scale module and DDR3 storage moulds are sent into by data input DDR3 fifo modules Block is cached;RGB turns gray scale module:For the rgb format image of 16 or 24 to be changed into 8 grayscale format images, will convert Greyscale image data deliver to medium filtering module;Medium filtering module:After rgb format coloured image is changed into grayscale format image, medium filtering is carried out, by intermediate value View data after filtering delivers to data input DDR3 fifo modules;Data input DDR3 fifo modules:This fifo module turns mainly for the treatment of across clocked data transfer and data bit width Change;Data after medium filtering are sent into DDR3 read-write clock zone from synchronous acquisition clock zone and data bit width is converted The bit wide length the same into same DDR3 reading-writing ports;Data after processing are delivered into outside DDR3 memory modules;DDR3 memory modules:For it is data cached with solve big data across clock transfer;Data output DDR3 fifo modules:This fifo module turns mainly for the treatment of across clocked data transfer and data bit width Change;Data are from DDR3 read-write clock zone feeding VGA read clocks domain and the same from DDR3 reading-writing ports data bit width Bit wide change into the bit wide size of a pixel data;Data after processing are delivered into inter-frame difference and binarization block;Inter-frame difference and binarization block:Including interframe subtraction module, binary conversion treatment module;Interframe subtraction module is by continuous two The data that frame greyscale image data subtracts each other to obtain deliver to binary conversion treatment module and carry out binary conversion treatment;By inter-frame difference and two Value module goes out moving target for Preliminary detection, and the data after processing are delivered into corrosion expansion module;Corrode expansion module:For removing the noise spot in the image after inter-frame difference and binaryzation and removing cavitation;Will Data after processing deliver to Motion mask extraction module;Motion mask extraction module:Moving target is extracted according to block size, the moving target detected is deposited with template size Enter FPGA on-chip memories, to be supplied to Motion mask matching module;Motion mask matching module:Each two field picture after being matched using obtained Motion mask, calculates optimal match point, Reach tracking effect;And the data that processing can be gone out deliver to VGA displays driving and serially printing module;VGA displays driving and serially printing module:For tracking result to be presented directly into user;VGA displays are used to show Tracking effect and movement locus, serially printing are used to print movement locus numerical value.
- A kind of 3. moving object detection tracking based on FPGA, it is characterised in that:Then view data carries out data conversion Divide two-way output;It is directly entered DDR3 memory modules all the way to be cached, another way turns gray proces, medium filtering by RGB Cached after processing into DDR3 memory modules;The two frame greyscale image datas continuously exported from DDR3 buffers carry out frame Between calculus of differences, binary conversion treatment, corrosion expansion process after obtain the coordinate value of Motion mask, then feed back to rgb format colour To obtain templates of moving objects on image, the storage of FPGA ram in slice is deposited into, finally using its template matches of taking exercises, with The tracking of color motion target is realized, and its movement locus and upper computer software display movement locus number of coordinates can be depicted Value.
- 4. the moving object detection tracking according to claim 3 based on FPGA, it is characterised in that:Described image number During according to carrying out data conversion and then divide the two-way to export, the method for data conversion is using displacement splicing method.
- 5. the moving object detection tracking according to claim 3 based on FPGA, it is characterised in that:The RGB turns The method of gray proces is that RGB changes into YCbCr format, then only takes its Y-component to obtain gray level image;RGB turns YCbCr public affairs Formula is as follows:Y= 0.183R + 0.614G + 0.062B + 16;Cb = -0.101R – 0.338G + 0.439B + 128;Cr = 0.439R – 0.399G – 0.040B + 128;Because Verilog can not carry out floating-point operation, therefore first expand 256 times, in 8 bit-wises that move right, after conversion Formula is:Y= (47R + 157G + 16B + 4096)>> 8;Cb = (26R – 86G + 112B + 32768)>> 8;Cr = (112R – 102G – 10B + 32768)>> 8;Finally, RGB Three-channel datas originally all take the numerical value of Y-component, and image is just converted into grayscale format from rgb format.
- 6. the moving object detection tracking according to claim 3 based on FPGA, it is characterised in that:The intermediate value filter The method of ripple processing includes;First, three pixels of every row are ranked up respectively;Then, the sequence three row pixels obtained Handled, i.e. minimum value in three maximums of extraction, the maximum in three minimum values, and the centre of three medians Value;Finally, by obtained in second step three values, intermediate value is taken again, has thus finally given the median of 9 pixels.
- 7. the moving object detection tracking according to claim 3 based on FPGA, it is characterised in that:Frame-to-frame differences partite transport Calculate and the method for binary conversion treatment is exactly that data are subtracted each other and then are compared with threshold value, complete 1 is just set to more than threshold value as white, it is no It is black to be then just set to full 0, and wherein threshold value takes 25.
- 8. the moving object detection tracking according to claim 3 based on FPGA, it is characterised in that:Corrode at expansion The method of reason is:Corrosion treatment is following three step, and first, the binary image of input is traveled through with 3 × 3 window modules;Then, AND operation all is done with 1 to the total data in 3 × 3 windows;Finally, it is as a result 0 as long as the value for having a pixel is 0;It is all 1, as a result it is only 1;Expansion process equally point three steps;First, the binary image of input is traveled through with 3 × 3 window modules;So Afterwards, inclusive-OR operation all is done with 1 to the total data in 3 × 3 windows;Finally, as long as the numerical value for having a pixel is 1, as a result for 1;0 is all, is as a result just 0.
- 9. the moving object detection tracking according to claim 3 based on FPGA, it is characterised in that:Moving target mould Plate extraction method be:To numerical value in binary image for 1 transverse and longitudinal coordinate a little count, where calculating target Minimum row, maximum row and minimum row, maximum column and barycenter;Then this parameter is returned to the coloured image of rgb format.
- 10. the moving object detection tracking according to claim 3 based on FPGA, it is characterised in that:Motion mask The method of matching is:Absolute value sum algorithm, i.e. SAD matching algorithms;It is as follows for RGB color image space, its formula:C = ∑( |Rx – Rx+1| + |Gx – Gx+1| + |Bx – Bx+1| )The minimum point of numerical value, is exactly optimal match point;Wherein in above-mentioned formula represent two continuous frames image same position corresponding to tri- components of RGB numerical value subtract each other absolute value it With R=red, G=green, B=blue.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710554813.9A CN107516296A (en) | 2017-07-10 | 2017-07-10 | A kind of moving object detection tracking system and method based on FPGA |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710554813.9A CN107516296A (en) | 2017-07-10 | 2017-07-10 | A kind of moving object detection tracking system and method based on FPGA |
Publications (1)
Publication Number | Publication Date |
---|---|
CN107516296A true CN107516296A (en) | 2017-12-26 |
Family
ID=60722346
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710554813.9A Pending CN107516296A (en) | 2017-07-10 | 2017-07-10 | A kind of moving object detection tracking system and method based on FPGA |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107516296A (en) |
Cited By (15)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108154489A (en) * | 2018-01-15 | 2018-06-12 | 大峡谷照明系统(苏州)股份有限公司 | Image interactive control method and system |
CN108400816A (en) * | 2018-02-27 | 2018-08-14 | 中国科学技术大学 | High-speed light identifing source based on wireless light communication, tracking with to Barebone |
CN109240965A (en) * | 2018-08-01 | 2019-01-18 | 清华大学 | Fpga logic capture processing display external member and its application method |
CN110006795A (en) * | 2019-04-30 | 2019-07-12 | 华北电力大学(保定) | Grain testing apparatus, method and FPGA |
CN110503081A (en) * | 2019-08-30 | 2019-11-26 | 山东师范大学 | Act of violence detection method, system, equipment and medium based on inter-frame difference |
CN110728617A (en) * | 2019-09-30 | 2020-01-24 | 上海电机学院 | FPGA-based dynamic target identification and real-time tracking system |
CN111340841A (en) * | 2020-03-19 | 2020-06-26 | 昆明理工大学 | Real-time multi-moving-object detection method and system |
CN111340835A (en) * | 2020-03-27 | 2020-06-26 | 天津光电通信技术有限公司 | FPGA-based video image edge detection system |
CN112967310A (en) * | 2021-02-04 | 2021-06-15 | 成都国翼电子技术有限公司 | FPGA-based template matching acceleration method |
CN113223047A (en) * | 2021-03-05 | 2021-08-06 | 兰州大学 | FPGA-based template matching target tracking method and tracking system |
CN113409355A (en) * | 2021-05-13 | 2021-09-17 | 杭州电子科技大学 | Moving target identification system and method based on FPGA |
CN113610221A (en) * | 2021-06-29 | 2021-11-05 | 西安电子科技大学 | Variable expansion convolution operation hardware system based on FPGA |
CN114217953A (en) * | 2021-11-18 | 2022-03-22 | 成都理工大学 | Target positioning system and identification method based on FPGA image processing |
CN115190238A (en) * | 2022-06-20 | 2022-10-14 | 中国人民解放军战略支援部队航天工程大学 | Ground demonstration verification system for detection and tracking of satellite-borne moving target |
WO2024001764A1 (en) * | 2022-06-30 | 2024-01-04 | 中兴通讯股份有限公司 | Image processing method and device, storage medium, and electronic device |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2008085731A1 (en) * | 2007-01-03 | 2008-07-17 | Tvia, Inc. | Digital color management method and system |
CN103826105A (en) * | 2014-03-14 | 2014-05-28 | 贵州大学 | Video tracking system and realizing method based on machine vision technology |
CN106204660A (en) * | 2016-07-26 | 2016-12-07 | 华中科技大学 | A kind of Ground Target Tracking device of feature based coupling |
-
2017
- 2017-07-10 CN CN201710554813.9A patent/CN107516296A/en active Pending
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2008085731A1 (en) * | 2007-01-03 | 2008-07-17 | Tvia, Inc. | Digital color management method and system |
CN103826105A (en) * | 2014-03-14 | 2014-05-28 | 贵州大学 | Video tracking system and realizing method based on machine vision technology |
CN106204660A (en) * | 2016-07-26 | 2016-12-07 | 华中科技大学 | A kind of Ground Target Tracking device of feature based coupling |
Non-Patent Citations (2)
Title |
---|
刘骏: "《基于FPGA的运动目标检测和跟踪系统设计》", 《CNKI数据库》 * |
姜宇: "《基于FPGA的运动目标检测系统的研究》", 《CNKI数据库》 * |
Cited By (19)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108154489A (en) * | 2018-01-15 | 2018-06-12 | 大峡谷照明系统(苏州)股份有限公司 | Image interactive control method and system |
CN108400816A (en) * | 2018-02-27 | 2018-08-14 | 中国科学技术大学 | High-speed light identifing source based on wireless light communication, tracking with to Barebone |
CN109240965B (en) * | 2018-08-01 | 2021-02-12 | 清华大学 | FPGA logic capturing processing display suite and use method thereof |
CN109240965A (en) * | 2018-08-01 | 2019-01-18 | 清华大学 | Fpga logic capture processing display external member and its application method |
CN110006795A (en) * | 2019-04-30 | 2019-07-12 | 华北电力大学(保定) | Grain testing apparatus, method and FPGA |
CN110006795B (en) * | 2019-04-30 | 2024-02-13 | 华北电力大学(保定) | Particle detection device and method and FPGA |
CN110503081A (en) * | 2019-08-30 | 2019-11-26 | 山东师范大学 | Act of violence detection method, system, equipment and medium based on inter-frame difference |
CN110728617A (en) * | 2019-09-30 | 2020-01-24 | 上海电机学院 | FPGA-based dynamic target identification and real-time tracking system |
CN111340841A (en) * | 2020-03-19 | 2020-06-26 | 昆明理工大学 | Real-time multi-moving-object detection method and system |
CN111340835A (en) * | 2020-03-27 | 2020-06-26 | 天津光电通信技术有限公司 | FPGA-based video image edge detection system |
CN112967310A (en) * | 2021-02-04 | 2021-06-15 | 成都国翼电子技术有限公司 | FPGA-based template matching acceleration method |
CN112967310B (en) * | 2021-02-04 | 2023-07-14 | 成都国翼电子技术有限公司 | Template matching acceleration method based on FPGA |
CN113223047A (en) * | 2021-03-05 | 2021-08-06 | 兰州大学 | FPGA-based template matching target tracking method and tracking system |
CN113409355A (en) * | 2021-05-13 | 2021-09-17 | 杭州电子科技大学 | Moving target identification system and method based on FPGA |
CN113610221A (en) * | 2021-06-29 | 2021-11-05 | 西安电子科技大学 | Variable expansion convolution operation hardware system based on FPGA |
CN113610221B (en) * | 2021-06-29 | 2024-02-13 | 西安电子科技大学 | FPGA-based variable expansion convolution operation hardware system |
CN114217953A (en) * | 2021-11-18 | 2022-03-22 | 成都理工大学 | Target positioning system and identification method based on FPGA image processing |
CN115190238A (en) * | 2022-06-20 | 2022-10-14 | 中国人民解放军战略支援部队航天工程大学 | Ground demonstration verification system for detection and tracking of satellite-borne moving target |
WO2024001764A1 (en) * | 2022-06-30 | 2024-01-04 | 中兴通讯股份有限公司 | Image processing method and device, storage medium, and electronic device |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107516296A (en) | A kind of moving object detection tracking system and method based on FPGA | |
Zhuang et al. | Underwater image enhancement with hyper-laplacian reflectance priors | |
CN107133969B (en) | A kind of mobile platform moving target detecting method based on background back projection | |
CN103379266A (en) | High-definition web camera with video semantic analysis function | |
CN108171734B (en) | ORB feature extraction and matching method and device | |
CN101729919B (en) | System for full-automatically converting planar video into stereoscopic video based on FPGA | |
CN103258332A (en) | Moving object detection method resisting illumination variation | |
CN104933418B (en) | A kind of crowd's demographic method of double image machine | |
CN104506872B (en) | A kind of method and device of converting plane video into stereoscopic video | |
CN107197121B (en) | A kind of electronic image stabilization method based on on-board equipment | |
CN104396236B (en) | The Video coding based on motion compensation is assisted using depth information | |
CN109389617A (en) | A kind of motion estimate based on piece heterogeneous system and method for tracing and system | |
CN108447060A (en) | Front and back scape separation method based on RGB-D images and its front and back scene separation device | |
CN108491857A (en) | A kind of multiple-camera target matching method of ken overlapping | |
CN103400153B (en) | Serial filtering matching method and system for real-time image identification | |
CN108282597A (en) | A kind of real-time target tracing system and method based on FPGA | |
CN108810506A (en) | A kind of Penetrating Fog enhancing image processing method and system based on FPGA | |
Xiao et al. | EVA $^{2} $: Event-Assisted Video Frame Interpolation via Cross-Modal Alignment and Aggregation | |
CN116913178B (en) | Spliced screen linkage system and video splicing method | |
Biswas et al. | HALSIE: Hybrid Approach to Learning Segmentation by Simultaneously Exploiting Image and Event Modalities | |
CN108447072A (en) | A kind of image partition method and system of crystal grain | |
CN113065506A (en) | Human body posture recognition method and system | |
CN109978787B (en) | Image processing method based on biological visual computing model | |
Das Biswas et al. | HALSIE: Hybrid Approach to Learning Segmentation by Simultaneously Exploiting Image and Event Modalities | |
CN108898566A (en) | A kind of low-luminance color video enhancement method using space-time illumination figure |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20171226 |
|
RJ01 | Rejection of invention patent application after publication |