CN107103609A - Niblack power equipment Infrared Image Segmentations based on particle group optimizing - Google Patents
Niblack power equipment Infrared Image Segmentations based on particle group optimizing Download PDFInfo
- Publication number
- CN107103609A CN107103609A CN201710249006.6A CN201710249006A CN107103609A CN 107103609 A CN107103609 A CN 107103609A CN 201710249006 A CN201710249006 A CN 201710249006A CN 107103609 A CN107103609 A CN 107103609A
- Authority
- CN
- China
- Prior art keywords
- mrow
- infrared image
- msub
- threshold value
- neighborhood
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/136—Segmentation; Edge detection involving thresholding
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Transforming Light Signals Into Electric Signals (AREA)
- Image Analysis (AREA)
Abstract
The present invention relates to a kind of Niblack power equipment Infrared Image Segmentations based on particle group optimizing, comprise the following steps:1) infrared image is obtained, the infrared image is divided into q nonoverlapping Continuous Rectangular neighborhoods, the gray average and grey scale variance of each rectangular neighborhood is calculated;2) it is directed to each rectangular neighborhood, a threshold value optimizing interval for corresponding to the rectangular neighborhood is obtained according to setting step-length, form q dimension population solution spaces, and using inter-class variance as particle cluster algorithm fitness function, automatic searching corresponds to the optimum segmentation threshold value T of each rectangular neighborhood in the q ties up population solution space*, the optimum segmentation threshold value T*So that inter-class variance is maximum;3) according to step 2) obtain each rectangular neighborhood optimum segmentation threshold value to each rectangular neighborhood carry out binary conversion treatment.Compared with prior art, the present invention solves the problems, such as to cause infrared image over-segmentation using traditional global threshold dividing method.
Description
Technical field
The present invention relates to a kind of image processing method, more particularly, to a kind of Niblack electric power based on particle group optimizing
Equipment Infrared Image Segmentation.
Background technology
In recent years, Transformer Substation Online Monitoring System is widely applied, and thermal infrared imager, visible light camera shoot and set
Standby visible ray and infrared image beam back master control room and carry out manual analysis, and although this method reduces the work of artificial gathered data
Amount, but be the failure to break away from the dependence to Artificial Diagnosis.With continuing to develop for artificial intelligence and image processing techniques, intelligent diagnostics
Technology starts to be applied to Fault Diagnosis for Electrical Equipment.Intelligent diagnosing method is broadly divided into three steps, is looked for first from infrared image
Go out device target region, i.e. area-of-interest (ROI), the information of correlation is then extracted from region, finally the letter to extracting
Breath classification is so as to complete Fault Diagnosis for Electrical Equipment.Can one step of wherein most critical be ROI acquisition, accurately obtain ROI
Whether accurate determine that power equipment temperature field information is extracted to a certain extent.It is general to be obtained using threshold segmentation method
ROI, this method has the advantages that simple to operate, arithmetic speed is fast.Domestic and foreign scholars have done numerous studies to it, and such as Otsu is proposed
One-dimensional maximum variance between clusters, the minimal error threshold value based on Bayes minimal error sorting criterions of the proposition such as Kittler
Method, the Threshold segmentation innovatory algorithm based on maximum entropy that Kapur etc. is provided, Kennedy and Eberhart propose jointly based on
The particle swarm optimization algorithm of group collaboration, and the Optimal improvements image segmentation algorithm based on particle cluster algorithm, etc..Above-mentioned calculation
Method is mostly based on global threshold, for noise is big, contrast is low, lack of homogeneity infrared image is difficult to target device well
Distinguished with background.
The content of the invention
It is an object of the present invention to overcome the above-mentioned drawbacks of the prior art and provide one kind is excellent based on population
The Niblack power equipment Infrared Image Segmentations of change.
The purpose of the present invention can be achieved through the following technical solutions:
A kind of Niblack power equipment Infrared Image Segmentations based on particle group optimizing, comprise the following steps:
1) infrared image is obtained, the infrared image is divided into q nonoverlapping Continuous Rectangular neighborhoods, each square is calculated
The gray average and grey scale variance of shape neighborhood;
2) each rectangular neighborhood is directed to, a threshold value optimizing interval for corresponding to the rectangular neighborhood is obtained according to setting step-length
{T1,T2,...,Ti,...,Tn, q dimension population solution spaces are formed, and particle cluster algorithm fitness letter is used as using inter-class variance
Number, automatic searching corresponds to the optimum segmentation threshold value T of each rectangular neighborhood in the q ties up population solution space*, it is described optimal
Segmentation threshold T*So that inter-class variance is maximum, wherein, Ti=m+kiS, i=1,2 ..., n, m are equal for the gray scale of current rectangle neighborhood
Value, s is the grey scale variance of current rectangle neighborhood, kiTo set i-th of value on interval according to setting step-length is equidistant,
N is value number;
3) according to step 2) obtain each rectangular neighborhood optimum segmentation threshold value to each rectangular neighborhood carry out binary conversion treatment.
The step 1) in, the infrared image is divided into before some nonoverlapping rectangular neighborhoods, infrared image is carried out
Continuation is handled.
The step 2) in, step-length is set as 0.05, and interval setting is [- 1,1].
The step 2) in, the pixel grey scale of rectangular neighborhood is divided into D1=[0 ..., T], D2=[T+1 ..., L-1] two
Class, inter-class variance formula is defined as:
Wherein, σ2(T) it is inter-class variance,D is represented respectively1Picture in class
Probability and D that plain gray scale occurs1The gray average of class, pjRepresent probability of the pixel grey scale for j pixel;
Threshold value of the threshold value optimizing of each rectangular neighborhood in interval is substituted into the inter-class variance formula successively, passes through population
Algorithm search obtains the optimum segmentation threshold value of each rectangular neighborhood.
The step 2) in, during searching optimum segmentation threshold value using particle cluster algorithm,
Particle i is { T in the q position marks for tieing up population solution spacei,1,Ti,2,…,Ti,q, each particle according to
Lower formula updates position and the speed of oneself, and particle is with speed Vi(t+1) from current location Ti(t) it is moved to the next position Ti(t+
1):
Vi(t+1)=ω × Vi(t)+c1×r1[Pbesti-Ti(t)]+c2×r2[Gbesti-Ti(t)]
Ti(t+1)=Ti(t)+Vi(t+1)
Wherein, ViAnd TiSpeed and position of i-th of particle in solution space are represented respectively, and it is optimal that t represents that population is searched
Current iteration number of times in threshold process, c1、c2For aceleration pulse, r1、r2For the random number between [0,1], ω represents particle
Inertia weight, PbestiFor current optimal value, GbestiFor global optimum.
The inertia weight ω passes through below equation adaptive change:
Wherein, ωmax、ωminThe maximum and minimum value of inertia weight are represented respectively, and G represents maximum iteration.
Compared with prior art, the present invention has advantages below:
1) scheme in the fitness function of the invention for using inter-class variance as particle cluster algorithm, automatic searching Niblack methods
As the optimum segmentation threshold value of not overlapping rectangles neighborhood, and the binarization segmentation of current neighborhood is used it for, solved using tradition
Global threshold dividing method causes infrared image over-segmentation problem;
2) present invention carries out segmentation threshold optimizing based on pixel grey scale, greatly reduces non-homogeneous background to each equipment
The influence of infrared thermal imaging figure segmentation effect, the integrality for improving target area.
Brief description of the drawings
Fig. 1 is schematic flow sheet of the invention;
Fig. 2 is the schematic diagram of five infrared artworks of power equipment;
Fig. 3 is segmentation result schematic diagram corresponding with Fig. 2.
Embodiment
The present invention is described in detail with specific embodiment below in conjunction with the accompanying drawings.The present embodiment is with technical solution of the present invention
Premised on implemented, give detailed embodiment and specific operating process, but protection scope of the present invention is not limited to
Following embodiments.
As shown in figure 1, the present embodiment provides a kind of Niblack power equipment infrared Image Segmentations based on particle group optimizing
Method, comprises the following steps:
1) infrared image is obtained, the infrared image is divided into q nonoverlapping Continuous Rectangular neighborhoods, each square is calculated
The gray average and grey scale variance of shape neighborhood.
In the present embodiment, importing resolution ratio is the infrared artwork g (x, y) of 320 × 240 power equipments, as shown in Fig. 2 as neighbour
Domain pixel takes 90 × 80, Neighborhood Number measure for the row of 4 row 3 totally 12 pieces not overlapping rectangles neighborhood when (to infrared original image horizontal boundary
Each 20 pixel symmetric extensions around being done, so that original image turns into neighborhood horizontal pixel integral multiple) segmentation effect is preferable.
2) each rectangular neighborhood is directed to, a threshold value optimizing interval for corresponding to the rectangular neighborhood is obtained according to setting step-length
{T1,T2,...,Ti,...,Tn, q dimension population solution spaces are formed, and particle cluster algorithm fitness letter is used as using inter-class variance
Number, automatic searching corresponds to the optimum segmentation threshold value T of each rectangular neighborhood in the q ties up population solution space*, it is described optimal
Segmentation threshold T*So that inter-class variance is maximum, wherein, Ti=m+kiS, i=1,2 ..., n, m are equal for the gray scale of current rectangle neighborhood
Value, s is the grey scale variance of current rectangle neighborhood, kiTo set i-th of value on interval according to setting step-length is equidistant,
N is value number.
In the present embodiment, interval setting is [- 1,1], sets step-length as 0.05, then can obtain 41 k values, can be calculated with this
Go out each rectangular neighborhood and be respectively provided with one 41 dimension threshold value optimizing interval { T1,T2,...,Ti,...,T41}.Therefore, population is at 12 pieces
Threshold value optimizing on rectangular partition neighborhood is interval, i.e., the dimension of population 12 solution space is expressed as:
Inter-class variance formula (2) is selected to search the maximum kind of 12 neighborhoods by PSO methods as population fitness function
Between varianceAnd its corresponding optimum segmentation threshold value
The pixel grey scale of rectangular neighborhood is divided into D1=[0 ..., T], D2=[T+1 ..., L-1] two classes, inter-class variance is public
Formula is defined as:
Wherein, σ2(T) it is inter-class variance,D is represented respectively1Picture in class
Probability and D that plain gray scale occurs1The gray average of class, pjRepresent probability of the pixel grey scale for j pixel;
Threshold value of the threshold value optimizing of each rectangular neighborhood in interval is substituted into the inter-class variance formula successively, passes through population
Algorithm search obtains the optimum segmentation threshold value of each rectangular neighborhood.
In particle cluster algorithm, interval for this small dimension threshold value optimizing of 41 dimensions, group's population is set to 10.Particle i exists
The position mark of solution space is { Ti,1,Ti,2,…,Ti,12}.Each particle updates position and the speed of oneself according to formula (3), (4)
Degree, particle is with speed Vi(t+1) from current location Ti(t) it is moved to the next position Ti(t+1)。
Vi(t+1)=ω × Vi(t)+c1×r1[Pbesti-Ti(t)]+c2×r2[Gbesti-Ti(t)] (3)
Ti(t+1)=Ti(t)+Vi(t+1) (4)
Wherein, ViAnd TiSpeed and position of i-th of particle in solution space are represented respectively, and it is optimal that t represents that population is searched
Current iteration number of times in threshold process, c1、c2For aceleration pulse, in the present embodiment, c1=c2=2, r1、r2For between [0,1]
Random number, ω represents the inertia weight of particle,ωmax、ωminInertia weight is represented respectively
Maximum and minimum value, in the present embodiment, ωmax=0.95, ωmin=0.4, G represent maximum iteration, the present embodiment
In, G=25, PbestiFor current optimal value, GbestiFor global optimum.
The optimum segmentation threshold value for obtaining each neighborhood by above-mentioned particle cluster algorithm is designated as:
3) according to step 2) obtain each rectangular neighborhood optimum segmentation threshold value to each rectangular neighborhood carry out binary conversion treatment,
As a result it is as shown in Figure 3.
Preferred embodiment of the invention described in detail above.It should be appreciated that one of ordinary skill in the art without
Need creative work just can make many modifications and variations according to the design of the present invention.Therefore, all technologies in the art
Personnel are available by logical analysis, reasoning, or a limited experiment on the basis of existing technology under this invention's idea
Technical scheme, all should be in the protection domain being defined in the patent claims.
Claims (6)
1. a kind of Niblack power equipment Infrared Image Segmentations based on particle group optimizing, it is characterised in that including following
Step:
1) infrared image is obtained, the infrared image is divided into q nonoverlapping Continuous Rectangular neighborhoods, each rectangle is calculated adjacent
The gray average and grey scale variance in domain;
2) each rectangular neighborhood is directed to, a threshold value optimizing interval { T for corresponding to the rectangular neighborhood is obtained according to setting step-length1,
T2,...,Ti,...,Tn, q dimension population solution spaces are formed, and using inter-class variance as particle cluster algorithm fitness function,
Automatic searching corresponds to the optimum segmentation threshold value T of each rectangular neighborhood in the q ties up population solution space*, the most optimal sorting
Cut threshold value T*So that inter-class variance is maximum, wherein, Ti=m+kiS, i=1,2 ..., n, m are equal for the gray scale of current rectangle neighborhood
Value, s is the grey scale variance of current rectangle neighborhood, kiTo set i-th of value on interval according to setting step-length is equidistant,
N is value number;
3) according to step 2) obtain each rectangular neighborhood optimum segmentation threshold value to each rectangular neighborhood carry out binary conversion treatment.
2. the Niblack power equipment Infrared Image Segmentations according to claim 1 based on particle group optimizing, it is special
Levy and be, the step 1) in, the infrared image is divided into before some nonoverlapping rectangular neighborhoods, infrared image is prolonged
Open up processing.
3. the Niblack power equipment Infrared Image Segmentations according to claim 1 based on particle group optimizing, it is special
Levy and be, the step 2) in, step-length is set as 0.05, and interval setting is [- 1,1].
4. the Niblack power equipment Infrared Image Segmentations according to claim 1 based on particle group optimizing, it is special
Levy and be, the step 2) in, the pixel grey scale of rectangular neighborhood is divided into D1=[0 ..., T], D2=[T+1 ..., L-1] two classes,
Inter-class variance formula is defined as:
<mrow>
<msup>
<mi>&sigma;</mi>
<mn>2</mn>
</msup>
<mrow>
<mo>(</mo>
<mi>T</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mfrac>
<msup>
<mrow>
<mo>&lsqb;</mo>
<mi>m</mi>
<mo>&times;</mo>
<msub>
<mi>p</mi>
<mrow>
<mi>D</mi>
<mn>1</mn>
</mrow>
</msub>
<mrow>
<mo>(</mo>
<mi>T</mi>
<mo>)</mo>
</mrow>
<mo>-</mo>
<msub>
<mi>m</mi>
<mrow>
<mi>D</mi>
<mn>1</mn>
</mrow>
</msub>
<mrow>
<mo>(</mo>
<mi>T</mi>
<mo>)</mo>
</mrow>
<mo>&rsqb;</mo>
</mrow>
<mn>2</mn>
</msup>
<mrow>
<msub>
<mi>p</mi>
<mrow>
<mi>D</mi>
<mn>1</mn>
</mrow>
</msub>
<mrow>
<mo>(</mo>
<mi>T</mi>
<mo>)</mo>
</mrow>
<mo>&lsqb;</mo>
<mn>1</mn>
<mo>-</mo>
<msub>
<mi>p</mi>
<mrow>
<mi>D</mi>
<mn>1</mn>
</mrow>
</msub>
<mrow>
<mo>(</mo>
<mi>T</mi>
<mo>)</mo>
</mrow>
<mo>&rsqb;</mo>
</mrow>
</mfrac>
</mrow>
Wherein, σ2(T) it is inter-class variance,D is represented respectively1Pixel grey scale in class
The probability and D of appearance1The gray average of class, pjRepresent probability of the pixel grey scale for j pixel;
Threshold value of the threshold value optimizing of each rectangular neighborhood in interval is substituted into the inter-class variance formula successively, passes through particle cluster algorithm
Search the optimum segmentation threshold value for obtaining each rectangular neighborhood.
5. the Niblack power equipment Infrared Image Segmentations according to claim 1 based on particle group optimizing, it is special
Levy and be, the step 2) in, during searching optimum segmentation threshold value using particle cluster algorithm,
Particle i is { T in the q position marks for tieing up population solution spacei,1,Ti,2,…,Ti,q, each particle is according to following public affairs
Formula updates position and the speed of oneself, and particle is with speed Vi(t+1) from current location Ti(t) it is moved to the next position Ti(t+1):
Vi(t+1)=ω × Vi(t)+c1×r1[Pbesti-Ti(t)]+c2×r2[Gbesti-Ti(t)]
Ti(t+1)=Ti(t)+Vi(t+1)
Wherein, ViAnd TiSpeed and position of i-th of particle in solution space are represented respectively, and t represents that population searches optimal threshold
During current iteration number of times, c1、c2For aceleration pulse, r1、r2For the random number between [0,1], ω represents the inertia of particle
Weight, PbestiFor current optimal value, GbestiFor global optimum.
6. the Niblack power equipment Infrared Image Segmentations according to claim 5 based on particle group optimizing, it is special
Levy and be, the inertia weight ω passes through below equation adaptive change:
<mrow>
<mi>&omega;</mi>
<mo>=</mo>
<msub>
<mi>&omega;</mi>
<mrow>
<mi>m</mi>
<mi>a</mi>
<mi>x</mi>
</mrow>
</msub>
<mo>-</mo>
<mrow>
<mo>(</mo>
<msub>
<mi>&omega;</mi>
<mrow>
<mi>m</mi>
<mi>a</mi>
<mi>x</mi>
</mrow>
</msub>
<mo>-</mo>
<msub>
<mi>&omega;</mi>
<mrow>
<mi>m</mi>
<mi>i</mi>
<mi>n</mi>
</mrow>
</msub>
<mo>)</mo>
</mrow>
<mfrac>
<mi>t</mi>
<mi>G</mi>
</mfrac>
</mrow>
1
Wherein, ωmax、ωminThe maximum and minimum value of inertia weight are represented respectively, and G represents maximum iteration.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710249006.6A CN107103609B (en) | 2017-04-17 | 2017-04-17 | Niblack power equipment Infrared Image Segmentation based on particle group optimizing |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710249006.6A CN107103609B (en) | 2017-04-17 | 2017-04-17 | Niblack power equipment Infrared Image Segmentation based on particle group optimizing |
Publications (2)
Publication Number | Publication Date |
---|---|
CN107103609A true CN107103609A (en) | 2017-08-29 |
CN107103609B CN107103609B (en) | 2019-06-11 |
Family
ID=59675836
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710249006.6A Active CN107103609B (en) | 2017-04-17 | 2017-04-17 | Niblack power equipment Infrared Image Segmentation based on particle group optimizing |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107103609B (en) |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111161300A (en) * | 2019-12-05 | 2020-05-15 | 西安工程大学 | Nilback image segmentation method based on improved Otsu method |
CN111489317A (en) * | 2020-05-12 | 2020-08-04 | 江西天境精藏科技有限公司 | Intelligent cinerary casket storage system |
CN111561771A (en) * | 2020-06-16 | 2020-08-21 | 重庆大学 | Intelligent air conditioner temperature adjusting method |
CN111583272A (en) * | 2020-04-17 | 2020-08-25 | 西安工程大学 | Improved Niblack infrared image segmentation method combined with maximum entropy |
CN111899250A (en) * | 2020-08-06 | 2020-11-06 | 罗春华 | Remote disease intelligent diagnosis system based on block chain and medical image |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102496156A (en) * | 2011-11-17 | 2012-06-13 | 西安电子科技大学 | Medical image segmentation method based on quantum-behaved particle swarm cooperative optimization |
CN103914831A (en) * | 2014-03-10 | 2014-07-09 | 西安电子科技大学 | Two-dimensional dual-threshold SAR image segmentation method based on quantum particle swarm optimization |
-
2017
- 2017-04-17 CN CN201710249006.6A patent/CN107103609B/en active Active
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102496156A (en) * | 2011-11-17 | 2012-06-13 | 西安电子科技大学 | Medical image segmentation method based on quantum-behaved particle swarm cooperative optimization |
CN103914831A (en) * | 2014-03-10 | 2014-07-09 | 西安电子科技大学 | Two-dimensional dual-threshold SAR image segmentation method based on quantum particle swarm optimization |
Non-Patent Citations (3)
Title |
---|
XIAOPING SU ET AL.: "An Image Enhancement Method Using the Quantum-Behaved Particle Swarm Optimization with an Adaptive Strategy", 《MATHEMATICAL PROBLEMS IN ENGINEERIN》 * |
王菊 等: "基于PSO和LM的信号稀疏分解快速算法", 《激光与红外》 * |
蔡超峰 等: "手背静脉图像分割及细化算法研究", 《图形图像》 * |
Cited By (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111161300A (en) * | 2019-12-05 | 2020-05-15 | 西安工程大学 | Nilback image segmentation method based on improved Otsu method |
CN111161300B (en) * | 2019-12-05 | 2023-03-21 | 西安工程大学 | Niblack image segmentation method based on improved Otsu method |
CN111583272A (en) * | 2020-04-17 | 2020-08-25 | 西安工程大学 | Improved Niblack infrared image segmentation method combined with maximum entropy |
CN111489317A (en) * | 2020-05-12 | 2020-08-04 | 江西天境精藏科技有限公司 | Intelligent cinerary casket storage system |
CN111489317B (en) * | 2020-05-12 | 2021-09-14 | 江西天境精藏科技有限公司 | Intelligent cinerary casket storage system |
CN111561771A (en) * | 2020-06-16 | 2020-08-21 | 重庆大学 | Intelligent air conditioner temperature adjusting method |
CN111899250A (en) * | 2020-08-06 | 2020-11-06 | 罗春华 | Remote disease intelligent diagnosis system based on block chain and medical image |
CN111899250B (en) * | 2020-08-06 | 2021-04-02 | 朗森特科技有限公司 | Remote disease intelligent diagnosis system based on block chain and medical image |
Also Published As
Publication number | Publication date |
---|---|
CN107103609B (en) | 2019-06-11 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107103609A (en) | Niblack power equipment Infrared Image Segmentations based on particle group optimizing | |
CN108764085A (en) | Based on the people counting method for generating confrontation network | |
CN110335270A (en) | Transmission line of electricity defect inspection method based on the study of hierarchical regions Fusion Features | |
Song et al. | MSFYOLO: Feature fusion-based detection for small objects | |
CN104992452B (en) | Airbound target automatic tracking method based on thermal imaging video | |
CN104992223A (en) | Intensive population estimation method based on deep learning | |
CN108345850A (en) | The scene text detection method of the territorial classification of stroke feature transformation and deep learning based on super-pixel | |
CN109376637A (en) | Passenger number statistical system based on video monitoring image processing | |
CN112347930B (en) | High-resolution image scene classification method based on self-learning semi-supervised deep neural network | |
CN109784375A (en) | Adaptive transformer part detection recognition method based on Faster RCNN | |
CN106874825A (en) | The training method of Face datection, detection method and device | |
CN109948457A (en) | The real time target recognitio algorithm accelerated based on convolutional neural networks and CUDA | |
CN116168240A (en) | Arbitrary-direction dense ship target detection method based on attention enhancement | |
Wang et al. | The obtainment and recognition of raw silk defects based on machine vision and image analysis | |
Mao et al. | Semantic segmentation for point cloud scenes via dilated graph feature aggregation and pyramid decoders | |
CN110348329A (en) | Pedestrian detection method based on video sequence interframe information | |
Zhang | [Retracted] Sports Action Recognition Based on Particle Swarm Optimization Neural Networks | |
Huang et al. | Drone-based car counting via density map learning | |
Heng et al. | Anti-vibration hammer detection in UAV image | |
CN113920498B (en) | Point cloud 3D object detection method based on multilayer feature pyramid | |
Zhang et al. | Video anomaly detection using open data filter and domain adaptation | |
Zhao et al. | Elite Loss for scene text detection | |
Cheng et al. | Text Detection of Transformer Based on Deep Learning Algorithm | |
CN113139423A (en) | Relation graph learning method for scene graph detection | |
Song et al. | Infrared passenger flow collection system based on rbf neural net |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |