CN107392968B - The image significance detection method of Fusion of Color comparison diagram and Color-spatial distribution figure - Google Patents

The image significance detection method of Fusion of Color comparison diagram and Color-spatial distribution figure Download PDF

Info

Publication number
CN107392968B
CN107392968B CN201710579455.7A CN201710579455A CN107392968B CN 107392968 B CN107392968 B CN 107392968B CN 201710579455 A CN201710579455 A CN 201710579455A CN 107392968 B CN107392968 B CN 107392968B
Authority
CN
China
Prior art keywords
color
pixel
super
class
image
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
CN201710579455.7A
Other languages
Chinese (zh)
Other versions
CN107392968A (en
Inventor
许金兰
崔玲玲
徐岗
吴卿
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.)
Hangzhou Dianzi University
Original Assignee
Hangzhou Dianzi 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 Hangzhou Dianzi University filed Critical Hangzhou Dianzi University
Priority to CN201710579455.7A priority Critical patent/CN107392968B/en
Publication of CN107392968A publication Critical patent/CN107392968A/en
Application granted granted Critical
Publication of CN107392968B publication Critical patent/CN107392968B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/90—Determination of colour characteristics
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/23—Clustering techniques
    • G06F18/232—Non-hierarchical techniques
    • G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
    • G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • General Physics & Mathematics (AREA)
  • Artificial Intelligence (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Probability & Statistics with Applications (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Evolutionary Biology (AREA)
  • Evolutionary Computation (AREA)
  • General Engineering & Computer Science (AREA)
  • Image Analysis (AREA)
  • Image Processing (AREA)

Abstract

The invention discloses the image significance detection methods of a kind of Fusion of Color comparison diagram and Color-spatial distribution figure.Common image significance detection method from bottom to top is calculated using low-level image features such as the color of image, brightness, edges.Synthetic image primary colour contrast characteristic figure and the Saliency maps of color of image spatial distribution characteristic Tu Lai get to the end of the present invention.The color contrast characteristic pattern of image is sought according to the image after SLIC super-pixel segmentation first, then, image after being clustered using K-Means obtains preliminary Color-spatial distribution characteristic pattern by calculating, it re-maps on super-pixel segmentation figure, and then advanced optimizes Color Distribution Features figure according to the similarity of color of image.Color of image spatial distribution characteristic figure after last Fusion of Color contrast characteristic figure and optimization obtains final notable figure.The present invention can obtain more accurate, complete notable figure in lower time complexity.

Description

The image significance detection method of Fusion of Color comparison diagram and Color-spatial distribution figure
Technical field
The invention belongs to Computer Image Processing fields, are related to the salient region of detection image, and in particular to fusion face The image significance detection method of colour contrast figure and Color-spatial distribution figure.
Background technique
The conspicuousness detection of image is to come out zone marker most important in image, comprising abundant information.Conspicuousness Detection has an important application in fields such as image segmentation, compression of images, image retrieval, target identifications, understanding for image and Processing has important application value.How letter that people want concern is rapidly and accurately retrieved from a large amount of image information Breath is an extremely important project.Research finds that the vision system of the mankind has visual selective ability, and the view of the mankind Feel system is broadly divided into two kinds, i.e. bottom-up strategy attention mechanism and top-down strategy attention mechanism.The former is to utilize figure The features such as color, brightness, the edge of picture calculate, and the latter calculates the conspicuousness of image mainly for the special characteristic of image Region.Since current image detected is uncertain, aimless region mostly, most of algorithm is the bottom of from Upward model.
It is introduced below to both domestic and external based on bottom-up image significance detection method.Wherein, earliest by Itti et al. (" A Model of Saliency-Based Visual Attention for Rapid Scene Analysis ") the famous biological heuristic models that were proposed in 1998, according to the behavior of vision system and neural network structure, Brightness, color and the direction character for extracting image, feature of the image under different scale is obtained using Core-Periphery difference. Hofmann et al. (" Graph-Based Visual Saliency ") proposed base in 2006 on the basis of Itti model In the GBVS algorithm of graph theory, which uses the feature extracting method of Itti, using pixel (or image block) as node simultaneously Difference between calculate node obtains a weighted-graph, finally calculates final notable figure using Markov chain.Zhai et al. (" attention detection in video sequences using spatiotemporal cues ") was in 2006 The LC algorithm of proposition can be calculated notable figure by calculating the difference of each pixel and rest of pixels grayscale information, but lack Few color information.Hou et al. (" A Spectral Residual Approach ") proposed frequency from frequency domain angle in 2007 Spectral difference method SR does inversefouriertransform using the Fourier spectrum of image and the difference of average frequency spectrum and obtains notable figure.This method is suitable For the lesser well-marked target of size, but notable figure often only has blinkpunkt region, without clear boundary.Achanta et al. In the AC algorithm that (" Salient region detection and segmentation ") was proposed in 2008, conspicuousness quilt It is defined as local contrast of the image-region relative to its neighborhood under multiple dimensioned, is a kind of full resolution algorithm, can be obtained clear Clear boundary information.Achanta et al. (" Frequency-tuned Salient Region Detection ") was in 2009 It is proposed that a kind of frequency domain modulation algorithm FT based on DOG operator, the algorithm utilize image each channel and color in lab color space The notable figure of global contrast can be obtained in the difference of color mean value.Cheng et al. (" Global Contrast based Salient Region Detection ") in 2011 proposed the detection method based on global contrast.This method is to the image after quantization Color histogram is established, obtains histogram contrast (HC) by calculating the diversity factor between each color and other colors. It using above-mentioned histogram image segmentation is different color blocks, recombinant spatial relationship calculates the saliency value in each region (RC), the notable figure based on region contrast is finally obtained.Hornung et al. (" Contrast based filtering for Salient region ") a kind of method that computational efficiency is improved based on filtering, the Computing Meta of notable figure were proposed in 2012 Element is that each super-pixel block after SLIC super-pixel segmentation, Color contrast and distribution of color variance are significant to calculate Figure.Guo Yingchun et al. (" being detected based on the saliency of Local feature and Regional feature ") utilized image in 2013 The local feature and provincial characteristics for the sub-block being calculated at multiple scales carry out the conspicuousness detection of natural image.Zhang Xudong Et al. (saliency of the bond area analysis of covariance " detect ") proposed in 2016 it is a kind of based on covariance matrix Detection method, but part well-marked target is imperfect.
Summary of the invention
The purpose of the present invention is in view of the deficiencies of the prior art, propose it is a kind of using super-pixel segmentation in conjunction with cluster segmentation Calculate the conspicuousness detection method of color of image contrast and Color-spatial distribution.This method can effectively inhibit background area The conspicuousness in domain detects the accurate salient region of image, the more previous only detection method characterized by color It can only detect that the particular image of zone errors, this method can correct the color of transient error according to Color-spatial distribution information Comparison diagram and obtain final accurate notable figure.
The specific steps of the present invention are as follows:
Input picture is smoothed by step 1, obtains smoothed image.
Smoothed image is divided into super-pixel figure using SLIC super-pixel segmentation algorithm, and calculates each super-pixel by step 2 Average color and mean place:
Wherein, RiIndicate i-th of super-pixel, pixel Ii∈Ri,For pixel IiColor vector,For pixel IiPosition Set vector, ciFor RiThe average color vector of middle all pixels, piFor RiThe mean place vector of middle all pixels, | Ri| indicate Ri The number of middle pixel.
Step 3, the color contrast value F that each super-pixel in super-pixel figure is calculated using center-surrounding principlei, obtain face Colour contrast figure.
Wherein, C (ci,cj)=| | ci-cj| | indicate ciWith cjEuclidean distance, Wp(pi,pj) it is the space for adjusting reduced value Weight, P (pi,pj)=| | pi-pj| | indicate piWith pjEuclidean distance.1/ZiBe so thatNormalizing Change the factor, σpValue 0.5, exp are exponent arithmetic, N=| Ri| -1, cjFor RjThe average color vector of middle all pixels, pjFor Rj The mean place vector of middle all pixels, RjIndicate j-th of super-pixel.
Smoothed image is divided into M class according to color cluster using K-Means clustering algorithm by step 4, and M≤10 obtain Dendrogram, class are also known as color component.The average color and mean place of each color component is calculated as follows.
Wherein, GiIndicate i-th of color component, pixel Pi∈Gi,For pixel PiColor vector, color vector use L component, a component and the expression of b component,For pixel PiPosition vector；cg iFor color component GiAverage color vector, i.e., Color component GiCluster centre in color；pg iFor color component GiMean place vector, i.e. color component GiSpatially Cluster centre, | Gi| indicate color component GiThe number of middle pixel.
Step 5 calculates space variance between class in dendrogram, color variance in space variance and class in class.
Wherein, Vi inAnd Vi outRespectively indicate color component GiClass in space variance between space variance and class,Indicate face Colouring component GiClass in color variance,Indicate color component GjCluster centre spatially.
Space variance between step 6, the class for acquiring step 5, in class non-linear group of color variance in space variance and class Synthesize preliminary Color-spatial distribution value.
Gauss weight is used to distribute weight for color variance in class, i.e.,Take 5.
Step 7 after obtaining the preliminary Color-spatial distribution value of each class by dendrogram, is mapped on super-pixel figure and carries out Optimization, the Color-spatial distribution value D optimizedi', and be normalized, the Color-spatial distribution figure optimized:
Wherein, UiRefer to super-pixel RiAdjoining super-pixel number；KiAnd KjIt is super-pixel R respectivelyiWith super-pixel RjJust The Color-spatial distribution value of step, viRefer to super-pixel RiIn the preliminary Color-spatial distribution value of each pixel.
The Color-spatial distribution figure of step 8, Fusion of Color comparison diagram and optimization, obtains notable figure.
Si=Fi·Di′ (11)
Due to FiAnd Di' all directly proportional to notable figure, so, if the gray value of certain super-pixel is zero, from formula (11) Obtain corresponding S in last notable figureiIt also is zero.
C(ci,cj) express super-pixel RiWith super-pixel RjColour-difference, colour-difference calculate when only take a component and b component The difference in two channels.
Beneficial effects of the present invention:
The notable figure that the present invention obtains combines Color-spatial distribution, is not in particular for color high-contrast area The image of marking area can efficiently detect accurate conspicuousness target area, meanwhile, reduce the significant of background area Property.
Detailed description of the invention
Fig. 1 (a), 1 (b), 1 (c) are respectively input picture, smoothed image and super-pixel figure；
Fig. 2 (a), 2 (b), 2 (c) are respectively color contrast figure, dendrogram and preliminary Color-spatial distribution figure；
Fig. 3 (a), 3 (b), 3 (c) are respectively the color space point of preliminary color spatial distribution map based on super-pixel, optimization Butut and notable figure；
Result figure, benchmark notable figure after the notable figure progress binaryzation that the respectively present invention of Fig. 4 (a), 4 (b) obtains；
Fig. 5 is algorithm flow chart of the invention.
Specific embodiment
There is no the part being described in detail to please refer to the description of summary of the invention in the present embodiment.
As shown in figure 5, the image significance detection method of Fusion of Color comparison diagram and Color-spatial distribution figure, specific steps It is as follows:
Step 1, the input picture to be processed for one, in order to which efficient process equal proportion adjustment length and width maximum resolution is 400.In order to avoid the influence of noise, convolution kernel size is used to be smoothed image for the gauss low frequency filter of 3*3, Smoothed image is obtained after filtering.Input picture and smoothed image are shown in Fig. 1 (a) and Fig. 1 (b) respectively.
Step 2 utilizes SLIC super-pixel segmentation algorithm (paper " the SLIC superpixels that Achanta et al. is delivered The super-pixel segmentation algorithm proposed in compared to state-of-the-art superpixel methods ") it will be smooth Image segmentation surpasses at about 400 super-pixel, and according to the average color and mean place of each super-pixel of formula (1) calculating Pixel map is shown in Fig. 1 (c).
Step 3, the color contrast value F for calculating each super-pixel according to formula (2), (3) using center-surrounding principlei, obtain To color contrast figure, and [0,1] is normalized to, color contrast figure is shown in Fig. 2 (a).
Step 4 utilizes K-Means clustering algorithm (paper " the Some Methods for that Macqueen et al. is delivered Clustering algorithm in Classification and Analysis of MultiVariate Observations ") smooth Image carries out cluster segmentation according to color vector, and result is that the pixel in 10 classes and each class is characterized in similar, class Also known as color component.The average color and mean place of each color component are calculated according to formula (4), dendrogram is shown in Fig. 2 (b)。
Step 5 calculates separately to obtain space variance between the class in dendrogram, space in class according to formula (5), (6), (7) Color variance in variance and class.
Step 6, according to color variance in space variance in space variance between the class in step 5, class and class according to formula (8) nonlinear combination is calculated preliminary Color-spatial distribution figure and sees Fig. 2 (c).In calculating process,WithRespectively between space space variance, class in class The normalized of three factors of color variance in variance and class, so that numberical range becomes [0,1].Wherein, space side in class Space variance is inversely proportional with conspicuousness between difference and class, so after normalizing by exponential function, in class between space variance and class Space variance is bigger closer to 0；Color variance is directly proportional to conspicuousness in class, so after being normalized by exponential function, class Interior color variance is bigger closer to 1.
Step 7, in the super-pixel figure that step 2 obtains, all pixels point for forming each super-pixel is searched, in preliminary face The mean value of these pixels of corresponding position is exactly the new value of super-pixel in colorspace distribution figure, thus first in dendrogram Step Color-spatial distribution figure has been mapped in super-pixel figure, is seen to obtain the preliminary Color-spatial distribution figure based on super-pixel Fig. 3 (a) utilizes formula (9) further according to the difference between the preliminary Color-spatial distribution figure and super-pixel based on super-pixel It optimizes, the Color-spatial distribution figure optimized is shown in Fig. 3 (b), and normalizes to [0,1].
Step 8, according to the color space of formula (11) fusion steps 3 and step 7 obtained color contrast figure and optimization Distribution map obtains notable figure and sees Fig. 3 (c).
Step 9, in order to obtain binary picture, 2 times for seeking notable figure average gray value are used as threshold value, are greater than or equal to threshold The pixel of value, is assigned a value of 1, less than the pixel of threshold value, is assigned a value of 0, obtains binary picture and sees Fig. 4 (a).Benchmark notable figure is shown in Fig. 4 (b).

Claims (2)

1. the image significance detection method of Fusion of Color comparison diagram and Color-spatial distribution figure, it is characterised in that: this method It comprises the concrete steps that:
Input picture is smoothed by step 1, obtains smoothed image；
Smoothed image is divided into super-pixel figure using SLIC super-pixel segmentation algorithm by step 2, and calculates the flat of each super-pixel Equal color and mean place:
Wherein, RiIndicate i-th of super-pixel, pixel Ii∈Ri,For pixel IiColor vector,For pixel IiPosition to Amount, ciFor RiThe average color vector of middle all pixels, piFor RiThe mean place vector of middle all pixels, | Ri| indicate RiMiddle picture The number of element；
Step 3, the color contrast value F that each super-pixel in super-pixel figure is calculated using center-surrounding principlei, obtain color contrast Figure；
Wherein, C (ci,cj)=| | ci-cj| | indicate ciWith cjEuclidean distance, Wp(pi,pj) it is the space right for adjusting reduced value Weight, P (pi,pj)=| | pi-pj| | indicate piWith pjEuclidean distance；1/ZiBe so thatNormalization The factor, σpValue 0.5, exp are exponent arithmetic, N=| Ri| -1, cjFor RjThe average color vector of middle all pixels, pjFor RjIn The mean place vector of all pixels, RjIndicate j-th of super-pixel；
Smoothed image is divided into M class according to color cluster using K-Means clustering algorithm by step 4, and M≤10 are clustered Figure, class are also known as color component；The average color and mean place of each color component is calculated as follows；
Wherein, GiIndicate i-th of color component, pixel Pi∈Gi,For pixel PiColor vector, color vector is using l point Amount, a component and the expression of b component, Pi pFor pixel PiPosition vector；cg iFor color component GiAverage color vector, i.e. color Component GiCluster centre in color；pg iFor color component GiMean place vector, i.e. color component GiSpatially poly- Class center, | Gi| indicate color component GiThe number of middle pixel；
Step 5 calculates in class in dendrogram between space variance, class color variance in space variance and class；
Wherein, Vi inAnd Vi outRespectively indicate color component GiClass in space variance between space variance and class,Indicate color point Measure GiClass in color variance,Indicate color component GjCluster centre spatially；
Space variance between step 6, the class for acquiring step 5, in class in space variance and class color variance nonlinear combination at Preliminary Color-spatial distribution value；
Gauss weight is used to distribute weight for color variance in class, i.e.,Take 5；
Step 7, after obtaining the preliminary Color-spatial distribution value of each class by dendrogram, be mapped on super-pixel figure carry out it is excellent Change, the Color-spatial distribution value D optimizedi', and be normalized, the Color-spatial distribution figure optimized:
Wherein, UiRefer to super-pixel RiAdjoining super-pixel number；KiAnd KjIt is super-pixel R respectivelyiWith super-pixel RjIt is preliminary Color-spatial distribution value, viRefer to super-pixel RiIn the preliminary Color-spatial distribution value of each pixel；
The Color-spatial distribution figure of step 8, Fusion of Color comparison diagram and optimization, obtains notable figure；
Si=Fi·Di′ (11)
Due to FiAnd Di' all directly proportional to notable figure, so, if the gray value of certain super-pixel is zero, obtained from formula (11) Corresponding S in last notable figureiIt also is zero.
2. the image significance detection method of Fusion of Color comparison diagram and Color-spatial distribution figure according to claim 1, It is characterized by: C (ci,cj) express super-pixel RiWith super-pixel RjColour-difference, colour-difference only takes a component and b points when calculating Measure the difference in two channels.
CN201710579455.7A 2017-07-17 2017-07-17 The image significance detection method of Fusion of Color comparison diagram and Color-spatial distribution figure Active CN107392968B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201710579455.7A CN107392968B (en) 2017-07-17 2017-07-17 The image significance detection method of Fusion of Color comparison diagram and Color-spatial distribution figure

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201710579455.7A CN107392968B (en) 2017-07-17 2017-07-17 The image significance detection method of Fusion of Color comparison diagram and Color-spatial distribution figure

Publications (2)

Publication Number Publication Date
CN107392968A CN107392968A (en) 2017-11-24
CN107392968B true CN107392968B (en) 2019-07-09

Family

ID=60340235

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201710579455.7A Active CN107392968B (en) 2017-07-17 2017-07-17 The image significance detection method of Fusion of Color comparison diagram and Color-spatial distribution figure

Country Status (1)

Country Link
CN (1) CN107392968B (en)

Families Citing this family (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109145994B (en) * 2018-08-28 2021-01-05 昆明理工大学 Image similarity judgment method based on improved color variance algorithm
CN109711432A (en) * 2018-11-29 2019-05-03 昆明理工大学 A kind of similar determination method of image based on color variance
CN110188763B (en) * 2019-05-28 2021-04-30 江南大学 Image significance detection method based on improved graph model
CN110765948A (en) * 2019-10-24 2020-02-07 长沙品先信息技术有限公司 Target detection and identification method and system based on unmanned aerial vehicle
CN111209918B (en) * 2020-01-06 2022-04-05 河北工业大学 Image saliency target detection method
CN112883827B (en) * 2021-01-28 2024-03-29 腾讯科技（深圳）有限公司 Method and device for identifying specified target in image, electronic equipment and storage medium
CN118212635A (en) * 2021-07-28 2024-06-18 中国科学院微小卫星创新研究院 Star sensor
CN115936540B (en) * 2023-02-16 2023-05-23 南京方园建设工程材料检测中心有限公司 Digital witness sampling and detecting system for construction engineering
CN117409000B (en) * 2023-12-14 2024-04-05 华能澜沧江水电股份有限公司 Radar image processing method for slope

Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105488812A (en) * 2015-11-24 2016-04-13 江南大学 Motion-feature-fused space-time significance detection method

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102779338B (en) * 2011-05-13 2017-05-17 欧姆龙株式会社 Image processing method and image processing device

Patent Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105488812A (en) * 2015-11-24 2016-04-13 江南大学 Motion-feature-fused space-time significance detection method

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
Saliency detection by selective color features;Yanbang Zhang et al;《Neurocomputing》;20160506;34-40 *
基于颜色通道比较的显著性检测;户卫东 等;《计算机系统应用》;20160831;第25卷(第8期);35-40 *

Also Published As

Publication number Publication date
CN107392968A (en) 2017-11-24

Similar Documents

Publication Publication Date Title
CN107392968B (en) The image significance detection method of Fusion of Color comparison diagram and Color-spatial distribution figure
CN104408429B (en) A kind of video represents frame extracting method and device
CN107610114B (en) optical satellite remote sensing image cloud and snow fog detection method based on support vector machine
Chen et al. Efficient hierarchical method for background subtraction
Yuan et al. Robust traffic sign recognition based on color global and local oriented edge magnitude patterns
CN103971126B (en) A kind of traffic sign recognition method and device
CN107909039B (en) High-resolution remote sensing image earth surface coverage classification method based on parallel algorithm
CN106909902B (en) Remote sensing target detection method based on improved hierarchical significant model
CN104850850B (en) A kind of binocular stereo vision image characteristic extracting method of combination shape and color
CN111738064B (en) Haze concentration identification method for haze image
CN104835175B (en) Object detection method in a kind of nuclear environment of view-based access control model attention mechanism
CN103473551A (en) Station logo recognition method and system based on SIFT operators
CN109918971B (en) Method and device for detecting number of people in monitoring video
CN107886507B (en) A kind of salient region detecting method based on image background and spatial position
CN106960182B (en) A kind of pedestrian's recognition methods again integrated based on multiple features
WO2017181892A1 (en) Foreground segmentation method and device
CN106485253B (en) A kind of pedestrian of maximum particle size structured descriptor discrimination method again
CN104282008B (en) The method and apparatus that Texture Segmentation is carried out to image
CN109063619A (en) A kind of traffic lights detection method and system based on adaptive background suppression filter and combinations of directions histogram of gradients
CN108629783A (en) Image partition method, system and medium based on the search of characteristics of image density peaks
CN104636754A (en) Intelligent image classifying method based on tongue body partition color feature
CN107590500A (en) A kind of color recognizing for vehicle id method and device based on color projection classification
CN106203448B (en) A kind of scene classification method based on Nonlinear Scale Space Theory
CN108629297A (en) A kind of remote sensing images cloud detection method of optic based on spatial domain natural scene statistics
CN104899559B (en) A kind of rapid pedestrian detection method based on video monitoring

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