CN104636753B - A kind of Region Feature Extraction method based on PCNN neuronal activations rate and group's dispersion - Google Patents

A kind of Region Feature Extraction method based on PCNN neuronal activations rate and group's dispersion Download PDF

Info

Publication number
CN104636753B
CN104636753B CN201510056549.7A CN201510056549A CN104636753B CN 104636753 B CN104636753 B CN 104636753B CN 201510056549 A CN201510056549 A CN 201510056549A CN 104636753 B CN104636753 B CN 104636753B
Authority
CN
China
Prior art keywords
neuron
group
pcnn
activation
activated
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.)
Active
Application number
CN201510056549.7A
Other languages
Chinese (zh)
Other versions
CN104636753A (en
Inventor
卞红雨
李曙光
张志刚
张健
陈奕名
韩冷
刘珈麟
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.)
Harbin Engineering University
Original Assignee
Harbin Engineering University
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 Harbin Engineering University filed Critical Harbin Engineering University
Priority to CN201510056549.7A priority Critical patent/CN104636753B/en
Publication of CN104636753A publication Critical patent/CN104636753A/en
Application granted granted Critical
Publication of CN104636753B publication Critical patent/CN104636753B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/08—Feature extraction
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/12—Classification; Matching

Landscapes

  • Engineering & Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Theoretical Computer Science (AREA)
  • Artificial Intelligence (AREA)
  • Signal Processing (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Image Analysis (AREA)
  • Image Processing (AREA)

Abstract

The invention discloses a kind of Region Feature Extraction method based on PCNN neuronal activations rate and group's dispersion.Comprise the following steps：Step 1：Original image is pre-processed, neutral net PCNN is corresponding with image；Step 2：0~255 tonal range is divided into N number of gray scale interval from big to small according to target area；Step 3：Obtain occurring in current gray level section the neuron of group's activation；Step 4：Statistics occurs group and activates excited target neuron number in the field of neuron, judges the neuron for lifting activation；Step 5：Activated in advance neuron number is counted, obtains group activity ratio and group's dispersion；Step 6：Next gray scale interval is read, repeat step three arrives step 6, until n-th section.The present invention has the advantages of computation complexity is small, good classification effect.

Description

A kind of Region Feature Extraction method based on PCNN neuronal activations rate and group's dispersion
Technical field
The invention belongs to a kind of feature extracting method of sonar image target area, more particularly to one kind can extract target Area pixel Distribution value feature, the Region Feature Extraction method based on PCNN neuronal activations rate and group's dispersion.
Background technology
With the development of science and technology, it is still that machine hopes dirt for the recognition capability of image by the mammal of representative of the mankind Not and, therefore a large amount of researcher both at home and abroad is constantly explored and the visual system of simulation animal, although neutral net Research experienced one and another low tide, but the step of scientists never stops.With third generation neutral net PCNN Proposition, the research for neutral net enters a new chapter, and Lindblad and Kinser is ground in its PCNN within 2005 The second edition for studying carefully monograph elaborates the artificial neural network based on mammalian visual cortical neuron, makes people in machine vision Be a a progressive step in terms of research, in terms of the research how handled for animal brain image and have it is many it is new into Fruit.For example Johnson proposes that two dimensional image is converted into the thought of one-dimensional pulse sequence using PCNN networks, some researchers carry Go out with PCNN image layered method.Document 1：Bo Yu；Liming Zhang.Pulse-Coupled Neural Networks for Contour and Motion Matchings.2004IEEE TRANSACTIONS ON NEURAL NETWORKS, p 1186-1201,2004, this article is using PCNN chain type conduction Activiation method to the edge of Ideal graph Carry out match cognization.Document 2：Liu's Qing, PCNN images steganalysis new method electricity of the Ma Yide based on histogram vectors center of gravity Sub- technology application, No.10,2006.Document 3：Liu's Qing, Xu Luping, Ma Yide, the small mesh of PCNN of a magnificent combinations gray level entropy conversion Logo image new detecting method Beijing Institute of Technology journal, Vol.29, No.12,2009.This two documents using matrix when assigning come Describe characteristics of image and be used for identification etc..
Generally speaking, the paper for image recognition being carried out using PCNN can be divided into two major classes：One kind be build it is more complicated PCNN structures, innovated using the complexity of increase network as one kind；The nerve that second class occurs group and activated by counting PCNN First number is analyzed, and segmentation, the denoising of image are carried out with this.These ways deposit deficiency both ways, first, have ignored PCNN Neighborhood neuron does not embody the dynamic movement of the main and auxiliary neuron of neighborhood to the dynamic effects of Master neuron；On the other hand ignore Simultaneously previously activated neuron characteristic distributions occur for the how individual neurons of PCNN, do not utilize the distribution of activated in advance neuron Characteristic.
The content of the invention
It is an object of the invention to provide a kind of computation complexity is small, good classification effect, based on PCNN neuronal activation rates With the Region Feature Extraction method of group's dispersion.
The present invention is achieved by the following technical solutions：
A kind of Region Feature Extraction method based on PCNN neuronal activations rate and group's dispersion, including following step Suddenly：
Step 1：Original image is gathered, original image is pre-processed, is partitioned into the profile of target area, by nerve Network PCNN is corresponding with image, and central nervous member is corresponding with the pixel of image, the neighborhood and neighborhood territory pixel of central nervous member Point is corresponding, and the input of neuron is the gray value of pixel；
Step 2：By setting the threshold parameter of PCNN neurons, 0~255 tonal range is drawn according to target area It is divided into N number of gray scale interval, gray scale interval is arranged according to the order of gray value from big to small；
Step 3：Current k-th of gray scale interval is read, obtains occurring in current gray level section the neuron of group's activation, hair Pixel value corresponding to the neuron of raw group's activation is corresponding to current gray level section in tonal range；
Step 4：Statistics occurs group and activates excited target neuron number λ in the neighborhood of neuron, recalculates generation group and swashs The value of the internal activity item of excited target neuron in the neighborhood of neuron living, if the value of the internal activity item of excited target neuron Positioned in tonal range, then current excited target neuron is activated in advance neuron, records and carries corresponding to current gray level section The coordinate of preceding activation neuron corresponding pixel points；
Step 5：Activated in advance neuron number λ ' is counted, obtains group activity ratio and group's dispersion：
Wherein x, y are the coordinates of each gray scale interval activated in advance neuron corresponding pixel points,
Step 6：It is 0 to make pixel point value corresponding to generation group activation and the activated in advance neuron in current gray level section, K=k+1 is made, repeat step three arrives step 6, until k=N.
Beneficial effect：
The present invention has refined PCNN neuron exciting phenomena, and group's activation and the previously activated god of excited target will successfully occur Separated through member, on the one hand characterize the grey value profile characteristic of target area using the activity ratio of activated in advance neuron, it is another Aspect counts pixel point coordinates corresponding to each gray scale interval activated in advance neuron, and having in target area, similar activation is special The pixel of property connects analysis.
The present invention utilizes the distributed architecture of PCNN neighborhood neurons, and target area flexibly is divided into small range pixel region Domain, the pixel value distribution characteristics of these small range pixel regions is characterized with previously activated neuron；There to be similar activity The region of matter considers jointly, has taken into account the pixel value gamma characteristic of neighborhood gradient characteristics and bigger regional extent, can count PCNN The distribution discreteness of activated in advance neuron in a wider context, therefore target area is described in microcosmic and macroscopical two kinds of different angles The feature in domain.Compared to other texture characteristic extracting methods, extract frogman using context of methods and fish target signature is used for During classification, computation complexity is smaller, classifying quality is more preferable.
Brief description of the drawings
Fig. 1 is the corresponding method of PCNN neurons and sonar image；
Fig. 2 is the gray level image of target region；
Fig. 3 is the target image being partitioned into；
The pixel that Fig. 4 activates for first time, the maximum gray scale interval of respective pixel value；
Fig. 5 is the pixel region influenceed by activation neuron；
Fig. 6 is the pixel that activated in advance phenomenon occurs in impacted pixel region；
Fig. 7 is the suppressed region no longer activated afterwards；
Fig. 8 is that un-activation neuron Q is influenceed by neuron P has been activated in neighborhood, and previously activated schematic diagram occurs；
Fig. 9 is that group activates neuron statistical chart；
Figure 10 is group's activity ratio statistical chart；
Figure 11 is group's dispersion statistical chart；
Figure 12 is the sonar image of fish target；
Figure 13 is the sonar image of frogman's target；
Figure 14 is the disposed of in its entirety flow chart of the inventive method.
Embodiment
The present invention is described in further details below in conjunction with accompanying drawing.
What the present invention was realized in：
1. a pair original image pre-processes, split, calibrate target area, PCNN is corresponding with sonar image, center Neuron is corresponding with the pixel of image, and the input of neuron is the gray value of pixel, the neighborhood and neighborhood of central nervous member Pixel is corresponding；
2. setting PCNN neuron threshold value relevant parameters, it is divided into N number of section by 0~255 gray level is autotelic；
3. the grey level range divided in statistic procedure 2, according to relatively small to gray value from the relatively large section of gray value Interval statistics (every time statistics one gray scale interval), judge whether the neuron in current gray level section occurs group and activate phenomenon；
4. according to the property of PCNN neurons, having activated neuron can encourage un-activation neuron in neighborhood to activate. Therefore the U values of impacted neuron to be recalculated；
5. judge impacted un-activation neuron whether to meet activation condition again and activate that (this is a kind of activated in advance Phenomenon).Activated in advance occurs for neuron corresponding to the pixel if meeting, records previously activated neuron and corresponds to picture The coordinate and quantity of vegetarian refreshments；
6. it is 0 to make pixel point value corresponding to all generation groups activation and the previously activated neuron of excited target, no longer occur Activation；
7. returning to step 3, statistics is lower as group's activating property of gray scale interval, is united according to the order from step 3 to step 6 The previously activated property of neuron is counted, is finished until all sections of the gray value in the range of 0~255 all count；
8. the neuron number of group's activation and previously activated neuron number occur for each gray scale interval of statistics, group is obtained Activity ratio.Pixel point coordinates corresponding to counting each gray scale interval activated in advance neuron, obtains group dispersion.By group's activity ratio With characteristic value of group's dispersion as description provincial characteristics.
The implementation process of the present invention is described below, and the present invention is described in more detail mainly in combination with Figure 14：
PCNN fundamental formular is as follows：
Uij[n]=Fij[n](1+βLij[n]) (3)
Tij[n]=exp (- t αT)Tij[n-1]+VTYij[n-1] (4)
In above formula：
Fij[n] represents neuron primary input.
SijRepresent image pixel gray level value corresponding to primary input central point.
Lij[n] represents auxiliary input, and its value represents the auxiliary input of primary input neighborhood.
Uij[n] is internal activity item.
Tij[n] is dynamic threshold.
MijklAnd WijklIt is the link weight matrix of neuron, represents the impression visual field of neuron.
Ykl[n-1] is the pulse output result of pulse-coupled neural networks, is worth for 1 (activation) and 0 (un-activation).
Ij represents the transverse and longitudinal coordinate of primary input region inconocenter vegetarian refreshments.Kl represents the transverse and longitudinal coordinate of auxiliary neighborhood territory pixel point.
VF、VLAnd VTIt is the affecting parameters of neighborhood, β is the interaction strength between primary input F roads and auxiliary input L roads.
N=1,2 ..., n are the multiples in minimum sampling period, represent which time dynamic change current formula is in, below Abbreviation iterations.
The implementation process of the present invention is described in detail below in conjunction with the accompanying drawings：
1. as shown in figure 1, PCNN is corresponding with sonar image, i.e. first, the neighborhood corresponding with the pixel of image of central nervous Neuron is corresponding with neighborhood territory pixel point, and neuron input is the gray value of pixel.Original image is pre-processed, main mesh Be smoothing denoising, be partitioned into target area, obtain Fig. 2, the image that PCNN corresponds to after processing；Here can be to objective contour Split, by the gray value zero setting of background, eliminate the influence of ambient noise, obtain Fig. 3.
2. by PCNN fundamental formular abbreviations：
F when iteration startsij[n] is equal to PCNN central nervous member corresponding grey scale values, i.e. Fij[n-1]=Sij, all neurons All in holddown, i.e. Yij[n-1]=0, now the activation characteristic of central nervous member is by SijWith attenuation parameter αFInfluence, if Remove the constant biasing S in formula (1)ij, formula (1) is deformed into：
Fij *[n]=Sijexp(-tαF) (6)
When judging whether un-activation neuron occurs activated in advance first, the neuron activated is mainly considered Influence to central nervous member, therefore formula (2) abbreviation is：
Influence of the PCNN neighborhoods neuron to central nervous member is a dynamic process, after neighborhood neuron initial activation, YijThe change of [n-1] can significantly affect its peripheral neurons and activated in advance occurs.Therefore formula (3) is deformed into two by the present invention Point：
Part I：Uij *[n]=Fij *[n] (8)
Part II：Uij[n]=Uij *[n]+βFij *[n]Lij[n] (9)
When judging whether neuron occurs group's activation first, all neurons are all in holddown, i.e. Yij[n-1]=0. Therefore formula (4) abbreviation is：
Tij[n]=exp (- t αT)T0 (10)
Wherein T0For initial threshold.
PCNN fundamental formulars (9) and formula (10) relevant parameter is set, makes similar gray value in identical sampling instant root It can be activated simultaneously according to formula (5)；0~255 gray level is divided into several sections, each gray scale interval includes can be simultaneously The neuron of activation；
3. according to the gray scale interval divided in step 2, according to the order that gray value corresponding to gray scale interval is descending, often Secondary one gray scale interval of Iterative statistical, judge PCNN neuronal activations situation corresponding to sonar image pixel according to formula (5). Fig. 4 show first time iteration, i.e., pixel corresponding to the neuron of group's activation occurs in maximum gray scale interval；
4. according to the exciting phenomena of PCNN neurons, having activated neuron can encourage un-activation neuron in neighborhood to occur to swash It is living, as shown in Figure 8.Therefore the U of impacted neuron is recalculated using formula (9)ij[n] value, Fig. 5 show excited target nerve Pixel corresponding to member；
5. judge impacted un-activation neuron whether to meet activation condition again and activate that (this is a kind of activated in advance Phenomenon), activated in advance occurs for neuron corresponding to the pixel if meeting, records previously activated neuron and corresponds to picture The coordinate and quantity of vegetarian refreshments, Fig. 6 show pixel corresponding to activated in advance neuron；
6. it is 0 to make pixel point value corresponding to all generation groups activation and the previously activated neuron of excited target, no longer occur Activation, as shown in Figure 7；
7. return to step 3 enters next iteration, according to the lower gray scale interval of order statistics from step 3 to step 6 The situation of group's activation neuron, excited target neuron and activated in advance neuron, until institute of the gray value in the range of 0~255 There is section all to count to finish, obtain the time pulse sequence figure shown in Fig. 9, abscissa is iterations in figure, and ordinate is every The neuron number of group's activation occurs in gray scale interval corresponding to secondary iteration；
8. the present invention successfully occurs previously activated in the neuron number λ and these neurons by counting excited target Number λ ', obtain activating neuron to group's activity ratio (hereinafter referred to as " group's activity ratio ") of neighborhood neuron, it is defined as follows：
Group's activity ratio corresponding to counting each gray scale interval activated in advance neuron, it can obtain group's activation as shown in Figure 10 Rate is vectorial, and abscissa is iterations in figure, and ordinate is group's activity ratio of gray scale interval corresponding to each iteration.
In addition, the present invention proposes a kind of PCNN " activation group dispersion " concept (hereinafter referred to as " group's dispersion "), it is used for The distribution dispersion of previously activated neuron spatially is described.It is proposed be the reason for it：Group's activity ratio characterizes neighborhood god Fillip and texture properties through member, but without embodying previously activated neuron in larger range of spatial distribution characteristic, Therefore propose that the concept of " group's dispersion " characterizes this feature, formula (12) is the definition of group dispersion：
Wherein x, y are the coordinates of each gray scale interval activated in advance pixel.
Group's dispersion corresponding to counting each gray scale interval activated in advance neuron, obtain group's dispersion shown in Figure 11 to Measure, abscissa is iterations in figure, and ordinate is group's dispersion of gray scale interval corresponding to each iteration.Group dispersion it is general The neighborhood limitation that compensate for activity ratio sign is read, embodies spatial distribution characteristic of the activation neuron in view picture target zone.
Using context of methods, the sonar image (and other images) shown in Figure 12 and Figure 13 can be classified.Method It is：Group's activity ratio vector of two targets to be sorted is matched into (matching process of group's dispersion vector is identical), matching Degree is defined as follows：
X in formula1iAnd x2iThe value of the features described above vector correspondence position of two targets is represented respectively, and N represents gray scale interval Number.
Tests prove that more than matching degree an order of magnitude higher than the matching degree of inhomogeneity target of the similar target of the present invention, With preferable classifying quality.

Claims (1)

  1. A kind of 1. Region Feature Extraction method based on PCNN neuronal activations rate and group's dispersion, it is characterised in that including with Under several steps：
    Step 1：Original image is gathered, original image is pre-processed, is partitioned into the profile of target area, by neutral net PCNN is corresponding with image, and central nervous member is corresponding with the pixel of image, the neighborhood and neighborhood territory pixel point pair of central nervous member Should, the input of neuron is the gray value of pixel；
    Step 2：By setting the threshold parameter of PCNN neurons, 0~255 tonal range is divided into N according to target area Individual gray scale interval, gray scale interval is arranged according to the order of gray value from big to small；
    Step 3：Current k-th of gray scale interval is read, obtains occurring in current gray level section the neuron of group's activation, group occurs Pixel value corresponding to the neuron of activation is corresponding to current gray level section in tonal range；
    Step 4：Statistics occurs group and activates excited target neuron number λ in the neighborhood of neuron, recalculates and group activation god occurs The value of the internal activity item of excited target neuron in neighborhood through member, if the value of the internal activity item of excited target neuron is located at Corresponding to current gray level section in tonal range, then current excited target neuron is activated in advance neuron, records and swashs in advance The coordinate of neuron corresponding pixel points living；
    Step 5：Activated in advance neuron number λ ' is counted, obtains group activity ratio and group's dispersion：
    Wherein xi,yiIt is the coordinate of each gray scale interval activated in advance neuron corresponding pixel points, Step 6：It is 0 to make pixel point value corresponding to generation group activation and the activated in advance neuron in current gray level section, makes k=k+ 1, repeat step three arrives step 6, until k=N.
CN201510056549.7A 2015-02-04 2015-02-04 A kind of Region Feature Extraction method based on PCNN neuronal activations rate and group's dispersion Active CN104636753B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201510056549.7A CN104636753B (en) 2015-02-04 2015-02-04 A kind of Region Feature Extraction method based on PCNN neuronal activations rate and group's dispersion

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201510056549.7A CN104636753B (en) 2015-02-04 2015-02-04 A kind of Region Feature Extraction method based on PCNN neuronal activations rate and group's dispersion

Publications (2)

Publication Number Publication Date
CN104636753A CN104636753A (en) 2015-05-20
CN104636753B true CN104636753B (en) 2017-11-21

Family

ID=53215481

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201510056549.7A Active CN104636753B (en) 2015-02-04 2015-02-04 A kind of Region Feature Extraction method based on PCNN neuronal activations rate and group's dispersion

Country Status (1)

Country Link
CN (1) CN104636753B (en)

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2017145818A1 (en) * 2016-02-24 2017-08-31 ソニー株式会社 Signal processing device, signal processing method, and program
CN108053391B (en) * 2017-11-22 2020-06-23 华中科技大学 Method for identifying neuron reconstruction errors

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101840568A (en) * 2009-09-29 2010-09-22 天津大学 Neigh Shrink image denoising method based on PCNN (Pulse Coupled Neural Network) region segmentation
CN102306289A (en) * 2011-09-16 2012-01-04 兰州大学 Method for extracting iris features based on pulse couple neural network (PCNN)
CN103605811A (en) * 2013-12-10 2014-02-26 三峡大学 Texture image retrieval method and device

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101840568A (en) * 2009-09-29 2010-09-22 天津大学 Neigh Shrink image denoising method based on PCNN (Pulse Coupled Neural Network) region segmentation
CN102306289A (en) * 2011-09-16 2012-01-04 兰州大学 Method for extracting iris features based on pulse couple neural network (PCNN)
CN103605811A (en) * 2013-12-10 2014-02-26 三峡大学 Texture image retrieval method and device

Non-Patent Citations (5)

* Cited by examiner, † Cited by third party
Title
Review of pulse-coupled neural networks;Zhaobin Wang;《Image and Vision Computing》;20100131;第28卷(第1期);第5-13页 *
一种基于PCNN的指纹特征提取算法;李双科;《万方数据》;20111230;第134-137页 *
一种基于神经网络提取图像特征的图像检索方法;宁晓菊;《西安邮电学院学报》;20060331;第11卷(第3期);第88-91页 *
基于脉冲耦合神经网络的图像处理若干问题研究;刘勍;《中国博士学位论文全文数据库 信息科技辑》;20111231;第I138-39页 *
带钢表面缺陷图像的一类特征提取及其降维方法研究;侯扬;《中国优秀硕士学位论文全文数据库 信息科技辑》;20130315;第I138-1199页 *

Also Published As

Publication number Publication date
CN104636753A (en) 2015-05-20

Similar Documents

Publication Publication Date Title
CN107103338B (en) SAR target recognition method integrating convolution features and integrated ultralimit learning machine
CN107330465B (en) A kind of images steganalysis method and device
CN106169081B (en) A kind of image classification and processing method based on different illumination
CN103247059B (en) A kind of remote sensing images region of interest detection method based on integer wavelet and visual signature
CN109684922A (en) A kind of recognition methods based on the multi-model of convolutional neural networks to finished product dish
CN106650806A (en) Cooperative type deep network model method for pedestrian detection
CN106446942A (en) Crop disease identification method based on incremental learning
CN111862093A (en) Corrosion grade information processing method and system based on image recognition
CN105913081B (en) SAR image classification method based on improved PCAnet
CN107545279A (en) Image-recognizing method based on convolutional neural networks Yu Weighted Kernel signature analysis
CN103020582A (en) Method for computer to identify vehicle type by video image
CN104636753B (en) A kind of Region Feature Extraction method based on PCNN neuronal activations rate and group's dispersion
Lavania et al. Image enhancement using filtering techniques
Fatma et al. Identification and classification of acute leukemia using neural network
CN106096660A (en) Convolutional neural networks based on independent composition analysis algorithm
CN110991554B (en) Improved PCA (principal component analysis) -based deep network image classification method
Thibault et al. Efficient statistical/morphological cell texture characterization and classification
CN110969590B (en) Image enhancement algorithm based on CA-SPCNN
CN103150573A (en) Nerve dendritic spine image classification method based on multiresolution fractal features
Sang et al. Image quality assessment based on quaternion singular value decomposition
CN108364007A (en) A kind of brain function connection features extracting method based on brain structure connection constraints
CN112836669B (en) Driver distraction driving detection method
CN109117867B (en) Multi-tested brain image prediction method based on gradient super-calibration
CN103839082B (en) SAR image target clustering method based on wavelet feature low-rank representation
Shi et al. Image edge detection method based on a simplified PCNN model with anisotropic linking mechanism

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant