CN107423709A - A kind of object detection method for merging visible ray and far infrared - Google Patents

A kind of object detection method for merging visible ray and far infrared Download PDF

Info

Publication number
CN107423709A
CN107423709A CN201710623646.9A CN201710623646A CN107423709A CN 107423709 A CN107423709 A CN 107423709A CN 201710623646 A CN201710623646 A CN 201710623646A CN 107423709 A CN107423709 A CN 107423709A
Authority
CN
China
Prior art keywords
target area
target
area
image block
detection method
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
Application number
CN201710623646.9A
Other languages
Chinese (zh)
Inventor
方武
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.)
Suzhou Institute of Trade and Commerce
Original Assignee
Suzhou Institute of Trade and Commerce
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 Suzhou Institute of Trade and Commerce filed Critical Suzhou Institute of Trade and Commerce
Priority to CN201710623646.9A priority Critical patent/CN107423709A/en
Publication of CN107423709A publication Critical patent/CN107423709A/en
Pending legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/22—Matching criteria, e.g. proximity measures
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/20—Analysis of motion
    • G06T7/254—Analysis of motion involving subtraction of images
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/40—Extraction of image or video features
    • G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
    • 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
    • G06T2207/10—Image acquisition modality
    • G06T2207/10048—Infrared image
    • 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
    • G06T2207/20021—Dividing image into blocks, subimages or windows
    • 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
    • G06T2207/20081—Training; Learning
    • 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
    • G06T2207/20212—Image combination
    • G06T2207/20221—Image fusion; Image merging
    • 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
    • G06T2207/20212—Image combination
    • G06T2207/20224—Image subtraction
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V2201/00—Indexing scheme relating to image or video recognition or understanding
    • G06V2201/07—Target detection

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Multimedia (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Data Mining & Analysis (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Evolutionary Biology (AREA)
  • Evolutionary Computation (AREA)
  • General Engineering & Computer Science (AREA)
  • Aiming, Guidance, Guns With A Light Source, Armor, Camouflage, And Targets (AREA)

Abstract

A kind of object detection method for merging visible ray and far infrared, including step：The visible images collected and far infrared graphic images are subjected to piecemeal, obtained image block is converted to the vector of N × 1；Structure mixing sampling matrix is acquired each image block after being compressed to the vector after conversion；Each pixel in each image block after compression is detected using Pixel-level background modeling algorithm, determines the target area and background area in image block；The target image detected is overlapped the target image after being merged with far infrared graphic images.Due to carrying out Sampling Compression to image by mixing sampling matrix, reduce calculating data volume, and according to subregion regularized learning algorithm speed, reduce average calculation times, it is overall to reduce number of pixels, the time of background modeling is effectively reduced, relative to traditional algorithm of target detection, an object of the application detection method makes memory size reduce 3/4ths, and processing time can reduce more than 40%.

Description

A kind of object detection method for merging visible ray and far infrared
Technical field
The present invention relates to technical field of image detection, and in particular to a kind of target detection for merging visible ray and far infrared Method.
Background technology
In computer vision correlation real-time vision system application field, the detection to target in the visual pattern of acquisition is first The step of wanting.The quality of algorithm of target detection is had influence on to further visual processes such as follow-up tracking and Activity recognitions.Due to The complicated and changeable of actual scene causes existing algorithm of target detection generally more complicated, computationally intensive, and memory size requires high, Be not suitable for the real-time vision system of resource-constrained.
Therefore, the problem of algorithm efficiency must be considered first by being directed to the algorithm of target detection of real-time vision system, use up It is likely to reduced amount of calculation and memory capacity.
The content of the invention
The application provides a kind of object detection method for merging visible ray and far infrared, including step：
The visible images collected and far infrared graphic images are subjected to piecemeal, obtained image block is converted to The vector of N × 1；
Structure mixing sampling matrix is acquired each image block after being compressed to the vector after conversion；
Each pixel in each image block after compression is detected using Pixel-level background modeling algorithm, it is determined that figure As the target area and background area in block；
The target image detected is overlapped the target image after being merged with far infrared graphic images.
In a kind of embodiment, structure mixing sampling matrix includes step：
75% high-density random sampling is carried out to the prediction target area in image block, obtains high-density sampling matrix；
25% low-density stochastical sampling is carried out to the projected background region in image block, obtains low-density sampling matrix；
High-density sampling matrix and low-density sampling matrix are merged into mixing sampling matrix.
In a kind of embodiment, prediction target area is the humidity province for being higher than mean temperature 25% in far infrared graphic images Domain.
In a kind of embodiment, during determining the target area and background area in image block, in addition to using different Strategy target area and background area are updated, and the parameter for mixing sampling matrix is adjusted according to the result of detection The step of section.
In a kind of embodiment, target area and background area are updated using different strategies, are specially：According to emerging Interesting region sets different sampled value M, and sample rate is improved in 1.2 times of the target area of former frame region, and is dropped in background area Low sampling rate, and, when background area brightness change it is smaller, the Gaussian Profile number of modeling is reduced, to reduce learning rate；When Background area brightness changes greatly, and Gaussian Profile number is improved, to improve learning rate.
In a kind of embodiment, target area and background area in image block are determined, is specially：Previous frame is detected Target area carries out matching detection after extension as the target area of present frame, wherein,
Pixel outside target area uses strict matching criterior；
Pixel in target area uses loose matching criterior.
In a kind of embodiment, the target area of target area that previous frame is detected as present frame after extension is entered Row matching detection, it is specially：The target area that previous frame is detected extends 15% target area as present frame, current Pixel outside the target area of frame reduces by 15% sampled point, increases by 15% sampled point in the target area of present frame.
In a kind of embodiment, the parameter for mixing sampling matrix is adjusted according to the result of detection, is specially：Next frame Target area sample rate lifting 15%, background area sample rate reduces by 15%.
In a kind of embodiment, in addition to post-processing is carried out to the target image after fusion and obtains final target image Step.
The object detection method of foundation above-described embodiment, due to carrying out Sampling Compression to image by mixing sampling matrix, Reduce calculating data volume, and according to subregion regularized learning algorithm speed, reduce average calculation times, and use not according to different zones Same sampled value, it is overall to reduce number of pixels, the time of background modeling is effectively reduced, is experimentally confirmed, the mesh of the application Mark detection method obtains preferable object detection results and has stronger anti-interference, is calculated relative to traditional target detection Method, memory size reduce 3/4ths, and processing time can reduce more than 40%.
Brief description of the drawings
Fig. 1 is target detection flow chart；
Fig. 2 builds schematic diagram for mixing sampling matrix；
Fig. 3 is the present invention and average every frame processing time comparison schematic diagram of other distinct methods.
Embodiment
The present invention is described in further detail below by embodiment combination accompanying drawing.
Compressive sensing theory breaches the requirement to sample number under tradition drawing Qwest theory, as long as signal is compressible It is or sparse, it is possible to by meeting that the observing matrix of certain condition is sampled the high dimensional signal after conversion, to obtain one Low-dimensional signal after individual sampling, then solve an optimization problem can perfectly reconstructed from a small amount of sampled value it is original Signal.
Background subtraction method is a kind of method ripe in object detection field Technical comparing, and application is quite varied.The party For method by subtracting each other to video image present frame and background model correspondence position pixel value, the absolute values being on duty are more than some threshold value When, the pixel is judged for object pixel, is otherwise background pixel.And handled by later image, obtain complete target image.
The algorithm of target detection of compressive sensing theory and background subtraction method is applied to fusion visible ray and far infrared In target detection, object region can be determined using far infrared thermal imaging, the reliability of target detection can be substantially improved. While retaining original image information, the pixel quantity of background modeling is greatly decreased, so as to improve efficiency of algorithm.
Based on this, this example provides a kind of object detection method for merging visible ray and far infrared, its flow chart such as Fig. 1 institutes Show, detailed process comprises the following steps.
S1：The visible images collected and far infrared graphic images are subjected to piecemeal, obtained image block is turned It is changed to the vector of N × 1.
In this step, image is carried out with far infrared graphic images size according to the visible images collected 8*8 piecemeals, obtained image block is converted to 64 × 1 vector.
S2：Structure mixing sampling matrix is acquired each image block after being compressed to the vector after conversion.
, can not be with because completely random calculation matrix carries out stochastical sampling for each pixel in image sequence Maximal efficiency obtains image useful information, as shown in Fig. 2 the mode of structure mixing sampling matrix is in this example：
75% high-density random sampling is carried out to the prediction target area in image block, obtains high-density sampling matrix St, to retain the useful information of target area, wherein, prediction target area is the area that temperature is higher in far infrared graphic images Domain (the higher temperature province referred to higher than mean temperature 25% of temperature)；
25% low-density stochastical sampling is carried out to the projected background region in image block, obtains low-density sampling matrix Sb, wherein, projected background region refers to predict other regions beyond target area；
By high-density sampling matrix StWith low-density sampling matrix SbMerge into mixing sampling matrix Sm：Sm=St∪Sb。
By mixing sampling matrix SmTo being sampled to the vector after conversion, the size of image can be so compressed.
S3：Each pixel in each image block after compression is detected using Pixel-level background modeling algorithm, really Determine the target area and background area in image block.
Wherein, each pixel in each image block after compression is detected using Pixel-level background modeling algorithm, Concrete mode is：A sample set is stored for each pixel, sampled value is exactly the past pixel of the pixel in sample set The pixel value of value and its neighbours' point, each new pixel value and sample set then be compared to judge whether to belong to background Point.In model, background model is that each background dot stores a sample set, then by each new pixel value and sample set Be compared to judge whether to belong to background dot, if a new observed value belong to background dot so it should with sample set Sampled value relatively.
It is shown below, note v (x) is the pixel value at x points；M (x) is that (sample set size is for background sample collection at x N)；SR(v (x)) is the R centered on x, and for the region of radius, (parameter uses district grid strategy, is adjusted in prediction target area R It is small 20%).Parameter is arranged to N=20, #min=2, R=20.If following formula establishment (parameter uses district grid strategy, 20%) prediction target area T is tuned up, it is judged that x points belong to background dot.
M (x)={ v1,v2,……vN}；
{SR(v(x))∩{v1,v2,……vN}}≥#min；
Pixel-level background modeling method has the characteristics that to calculate simple, Detection results preferably and reply noise is stable, is adapted to In the application scenarios that the amounts of calculation such as embedded vision system are small and memory size requirement is low.
It is determined that during target area and background area in image block, in addition to using different strategies to target Region and background area are updated, and the step of the parameter for mixing sampling matrix is adjusted according to the result of detection, its In, target area and background area are updated using different strategies, are specially：Different adopt is set according to interest region Sample value M, sample rate is improved in 1.2 times of the target area of former frame region, and sample rate is reduced in background area, and, work as the back of the body Scene area brightness change is smaller, the Gaussian Profile number of modeling is reduced, to reduce learning rate；When background area brightness change compared with Greatly, Gaussian Profile number is improved, to improve learning rate.
Further, the target area and background area in image block are determined, is specially：The target area that previous frame is detected Domain carries out matching detection after extension as the target area of present frame, wherein, the pixel outside target area is using tight The matching criterior of lattice；
Pixel in target area uses loose matching criterior.
Further, target area previous frame detected is matched after extension as the target area of present frame Detection, it is specially：The target area that previous frame is detected extends 15% target area as present frame, in the mesh of present frame The pixel marked outside region reduces by 15% sampled point, increases by 15% sampled point in the target area of present frame, can improve target The effect of detection.
Further, the parameter for mixing sampling matrix is adjusted according to the result of detection, is specially：Next frame target area Domain sample rate lifting 15%, background area sample rate reduces by 15%.
S4：The target image detected is overlapped the target image after being merged with far infrared graphic images.
Specifically, carrying out Canny rim detections to the target image detected, edge image I is obtainedc, then using one Determine weights α (image averaging gray value/255) superposition far infrared graphic images ItThe image I after final fusion is obtained afterwardsm： Im=α It+(1-α)Ic。
S5：Post-processing is carried out to the target image after fusion and obtains final target image.
The bianry image template Morg of target image is can obtain according to step S1-S4.3 × 3 are carried out to two-value template Morg Morphology opening operation, it is Ms to obtain result, then it is M to obtain result after removing isolated point by 3 × 3 erosion operations.The process The loss of partial target pixel is result in, takes the processing method as follows based on morphology object reconstruction to retain as far as possible more More target images：
Wherein, F is the final result after foreground extraction, noise filtering.The size of structural element SE in equation is big The small target size for depending on detection.Experiment finds that using 5 × 5 structural element preferable object detection results can be reached. Carrying out cavity filling to the foreground target F being partitioned into using structural element combination Assimilation filling can make target more complete.Finally The result counted by target sizes removes the fritter for being less than 40 pixels, to reach the purpose for eliminating noise.
The application selects embedded vision platform to carry out target detection test, and the design parameter of vision system test platform is such as Shown in table 1, according to table 2 and table 3 and Fig. 3, according to the comparison with other existing algorithms, the present invention can obtain preferable target Testing result and there is stronger anti-interference, relative to other traditional algorithm of target detection, memory size reduces about four / tri-, processing time can reduce more than 40%.
The vision system test platform parameter of table 1
The Reliability comparotive of 2 four kinds of algorithms of table
The memory size of the distinct methods of table 3 compares
Use above specific case is illustrated to the present invention, is only intended to help and is understood the present invention, not limiting The system present invention.For those skilled in the art, according to the thought of the present invention, can also make some simple Deduce, deform or replace.

Claims (9)

1. a kind of object detection method for merging visible ray and far infrared, it is characterised in that including step：
The visible images collected and far infrared graphic images are subjected to piecemeal, obtained image block is converted into N × 1 Vector；
Structure mixing sampling matrix is acquired each image block after being compressed to the vector after conversion；
Each pixel in each image block after compression is detected using Pixel-level background modeling algorithm, determines image block In target area and background area；
The target image detected is overlapped the target image after being merged with far infrared graphic images.
2. object detection method as claimed in claim 1, it is characterised in that the structure mixing sampling matrix includes step：
75% high-density random sampling is carried out to the prediction target area in image block, obtains high-density sampling matrix；
25% low-density stochastical sampling is carried out to the projected background region in image block, obtains low-density sampling matrix；
The high-density sampling matrix and low-density sampling matrix are merged into mixing sampling matrix.
3. object detection method as claimed in claim 2, it is characterised in that the prediction target area is far infrared heat It is higher than the temperature province of mean temperature 25% in image.
4. object detection method as claimed in claim 1, it is characterised in that the target area determined in image block and the back of the body During scene area, in addition to different strategies is used to be updated target area and background area, and according to detection As a result the step of parameter for mixing sampling matrix being adjusted.
5. object detection method as claimed in claim 4, it is characterised in that it is described using different strategies to target area and Background area is updated, and is specially：Different sampled value M is set according to interest region, 1.2 times of the target area of former frame Region improve sample rate, and background area reduce sample rate, and, when background area brightness change it is smaller, reduce modeling Gaussian Profile number, to reduce learning rate；When background area, brightness changes greatly, and improves Gaussian Profile number, is learned with improving Practise speed.
6. object detection method as claimed in claim 4, it is characterised in that the target area determined in image block and the back of the body Scene area, it is specially：The target area that previous frame is detected is matched after extension as the target area of present frame Detection, wherein,
Pixel outside target area uses strict matching criterior；
Pixel in target area uses loose matching criterior.
7. object detection method as claimed in claim 6, it is characterised in that the target area warp for detecting previous frame The target area progress matching detection as present frame after extension is crossed, is specially：The target area that previous frame is detected extends 15% target area as present frame, the pixel outside the target area of present frame reduces by 15% sampled point, in present frame Target area in increase by 15% sampled point.
8. object detection method as claimed in claim 4, it is characterised in that the result according to detection samples square to mixing The parameter of battle array is adjusted, and is specially：Next frame target area sample rate lifting 15%, background area sample rate reduces by 15%.
9. object detection method as claimed in claim 1, it is characterised in that also include after being carried out to the target image after fusion The step of phase handles to obtain final target image.
CN201710623646.9A 2017-07-27 2017-07-27 A kind of object detection method for merging visible ray and far infrared Pending CN107423709A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201710623646.9A CN107423709A (en) 2017-07-27 2017-07-27 A kind of object detection method for merging visible ray and far infrared

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201710623646.9A CN107423709A (en) 2017-07-27 2017-07-27 A kind of object detection method for merging visible ray and far infrared

Publications (1)

Publication Number Publication Date
CN107423709A true CN107423709A (en) 2017-12-01

Family

ID=60430235

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201710623646.9A Pending CN107423709A (en) 2017-07-27 2017-07-27 A kind of object detection method for merging visible ray and far infrared

Country Status (1)

Country Link
CN (1) CN107423709A (en)

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108428224A (en) * 2018-01-09 2018-08-21 中国农业大学 Animal body surface temperature checking method and device based on convolutional Neural net
CN110278029A (en) * 2019-06-25 2019-09-24 Oppo广东移动通信有限公司 Data transfer control method and Related product
CN111931754A (en) * 2020-10-14 2020-11-13 深圳市瑞图生物技术有限公司 Method and system for identifying target object in sample and readable storage medium
CN112132753A (en) * 2020-11-06 2020-12-25 湖南大学 Infrared image super-resolution method and system for multi-scale structure guide image
CN112257626A (en) * 2020-10-29 2021-01-22 辽宁工程技术大学 Method and system for sampling remote sensing data

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103778618A (en) * 2013-11-04 2014-05-07 国家电网公司 Method for fusing visible image and infrared image
US20140233796A1 (en) * 2013-02-15 2014-08-21 Omron Corporation Image processing device, image processing method, and image processing program
CN104123734A (en) * 2014-07-22 2014-10-29 西北工业大学 Visible light and infrared detection result integration based moving target detection method
CN104599290A (en) * 2015-01-19 2015-05-06 苏州经贸职业技术学院 Video sensing node-oriented target detection method
CN105095898A (en) * 2015-09-06 2015-11-25 苏州经贸职业技术学院 Real-time vision system oriented target compression sensing method
CN105654511A (en) * 2015-12-29 2016-06-08 浙江大学 Quick detecting and tracking method for weak moving object

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20140233796A1 (en) * 2013-02-15 2014-08-21 Omron Corporation Image processing device, image processing method, and image processing program
CN103778618A (en) * 2013-11-04 2014-05-07 国家电网公司 Method for fusing visible image and infrared image
CN104123734A (en) * 2014-07-22 2014-10-29 西北工业大学 Visible light and infrared detection result integration based moving target detection method
CN104599290A (en) * 2015-01-19 2015-05-06 苏州经贸职业技术学院 Video sensing node-oriented target detection method
CN105095898A (en) * 2015-09-06 2015-11-25 苏州经贸职业技术学院 Real-time vision system oriented target compression sensing method
CN105654511A (en) * 2015-12-29 2016-06-08 浙江大学 Quick detecting and tracking method for weak moving object

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
吕梅柏: "双波段成像跟踪***设计", 《中国优秀硕士学位论文全文数据库_工程科技II辑》 *

Cited By (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108428224A (en) * 2018-01-09 2018-08-21 中国农业大学 Animal body surface temperature checking method and device based on convolutional Neural net
CN108428224B (en) * 2018-01-09 2020-05-22 中国农业大学 Animal body surface temperature detection method and device based on convolutional neural network
CN110278029A (en) * 2019-06-25 2019-09-24 Oppo广东移动通信有限公司 Data transfer control method and Related product
CN110278029B (en) * 2019-06-25 2020-12-22 Oppo广东移动通信有限公司 Data transmission control method and related product
CN111931754A (en) * 2020-10-14 2020-11-13 深圳市瑞图生物技术有限公司 Method and system for identifying target object in sample and readable storage medium
CN112257626A (en) * 2020-10-29 2021-01-22 辽宁工程技术大学 Method and system for sampling remote sensing data
CN112132753A (en) * 2020-11-06 2020-12-25 湖南大学 Infrared image super-resolution method and system for multi-scale structure guide image
CN112132753B (en) * 2020-11-06 2022-04-05 湖南大学 Infrared image super-resolution method and system for multi-scale structure guide image

Similar Documents

Publication Publication Date Title
CN107423709A (en) A kind of object detection method for merging visible ray and far infrared
WO2021208275A1 (en) Traffic video background modelling method and system
CN104392468B (en) Based on the moving target detecting method for improving visual background extraction
WO2016192494A1 (en) Image processing method and device
CN105404847B (en) A kind of residue real-time detection method
CN104268505A (en) Automatic cloth defect point detection and recognition device and method based on machine vision
CN107833242A (en) One kind is based on marginal information and improves VIBE moving target detecting methods
CN107204004B (en) Aluminum electrolysis cell fire eye video dynamic feature identification method and system
CN105354847A (en) Fruit surface defect detection method based on adaptive segmentation of sliding comparison window
CN105095898B (en) A kind of targeted compression cognitive method towards real-time vision system
EP3200442B1 (en) Method and apparatus for image processing
CN110580709A (en) Target detection method based on ViBe and three-frame differential fusion
CN111062331B (en) Image mosaic detection method and device, electronic equipment and storage medium
CN108038856B (en) Infrared small target detection method based on improved multi-scale fractal enhancement
CN107993254A (en) Moving target detecting method based on disassociation frame calculus of finite differences
CN109949308A (en) A kind of space Relative Navigation target rapid extracting method of anti-starlight interference
CN111860143A (en) Real-time flame detection method for inspection robot
TW201032180A (en) Method and device for keeping image background by multiple gauss models
CN116485885A (en) Method for removing dynamic feature points at front end of visual SLAM based on deep learning
CN108492306A (en) A kind of X-type Angular Point Extracting Method based on image outline
CN109978916A (en) Vibe moving target detecting method based on gray level image characteristic matching
CN108010050B (en) Foreground detection method based on adaptive background updating and selective background updating
CN113657264A (en) Forest fire smoke root node detection method based on fusion of dark channel and KNN algorithm
CN113096103A (en) Intelligent smoke image sensing method for emptying torch
CN105205485B (en) Large scale image partitioning algorithm based on maximum variance algorithm between multiclass class

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
RJ01 Rejection of invention patent application after publication
RJ01 Rejection of invention patent application after publication

Application publication date: 20171201