CN104751170A - Method of estimating adaptability of heterologous radar image based on learning monitoring strategy - Google Patents
Method of estimating adaptability of heterologous radar image based on learning monitoring strategy Download PDFInfo
- Publication number
- CN104751170A CN104751170A CN201510097972.1A CN201510097972A CN104751170A CN 104751170 A CN104751170 A CN 104751170A CN 201510097972 A CN201510097972 A CN 201510097972A CN 104751170 A CN104751170 A CN 104751170A
- Authority
- CN
- China
- Prior art keywords
- image
- middle level
- training set
- level features
- suitability
- 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
Abstract
The invention relates to a method of estimating adaptability of a heterologous radar image based on a learning monitoring strategy. The method comprises the following steps: (1) dividing an image of a known matching result into two parts, which are respectively named as a training set I and a training set II; (2) training a middle level characteristic classifier by using the training set I; (3) using the training set II to train an image adaptability estimation model by using the training set II; (4) obtaining a to-be-estimated prediction set, and obtaining a matching result of an existing image of the prediction set according to the middle level characteristic classifier and the adaptability estimation model. Compared with the prior art, according to the method provided by the invention, the middle level characteristics can be adopted to describe the image characteristics, and the recognition on common features among the heterologous radar images by a user can be added into a training process of the classifier, so as to effectively distinguish the common features among the heterologous radar images, and solve the problem of estimating the adaptability among the heterologous radar image.
Description
Technical field
The present invention relates to a kind of allos radar image suitability appraisal procedure, especially relate to a kind of allos radar image suitability appraisal procedure based on supervised learning strategy.
Background technology
Utilizing Scene Matching Techniques in the positioning system realizing aircraft navigation, in order to obtain high-precision positioning result, except matching algorithm performance is had higher requirements, also need to consider the Adapter Property between coupling image, high performance matching algorithm so just can be made to play its due usefulness.Contain much information according to specified criteria selected characteristic on flight track, information stability, the suitability technology in scene region can be called as the technology of Matching band in the scene region that meets the demands of high, the size of matching.Suitability technology domestic be at present all mostly for allos optical image between matching problem, cannot effectively solve between allos radar image suitability assessment.And assist guidance technology to be key link in high-speed aircraft terminal guidance link based on the scene matching aided navigation of allos radar image.Therefore research and development for the suitability assessment technology of allos radar image be China's space guidance technical development in the urgent need to.
Summary of the invention
Object of the present invention is exactly that a kind of suitability test accuracy is high, the fireballing allos radar image suitability appraisal procedure based on supervised learning strategy of feature extraction in order to the suitability assessment solved between allos radar image provides.
Object of the present invention can be achieved through the following technical solutions:
Based on an allos radar image suitability appraisal procedure for supervised learning strategy, the method comprises the following steps:
(1) by the image of known matching result to being divided into two parts, respectively called after training set I and training set II;
(2) training set I is utilized to train middle level features sorter;
(3) according to the middle level features sorter that step (2) obtains, training set II is utilized to train image suitability assessment models;
(4) obtain forecast set to be assessed, obtain the matching result of the current image of described forecast set according to described middle level features sorter and suitability assessment models.
Described training set I and training set II comprise real-time diagram data and benchmark graph data respectively.
In described training set I and training set II, the difference of positive sample size and negative sample quantity is ± 5%.
Described step (2) concrete steps are as follows:
(2-1) the common feature binary map of data in training set I is drawn;
(2-2) common feature binary map is resolved into multiple image block, retain the image block that center exists effective pixel points, reject residual image block, obtain effective block collection S;
(2-3) corresponding to all blocks in S SAR image region calculates Daisy descriptor;
(2-4) using K-means algorithm as clustering algorithm, the Daisy descriptor of each block, as cluster amount, carries out cluster to all blocks in S;
(2-5) corresponding to block each in S SAR imagery zone calculates multiple gradient respectively and auto-correlation statistic composition middle level features inputs as X, and the category feature label in clustering algorithm exports as Y, trains a decision forest;
(2-6) utilize decision forest training classifier, obtain middle level features sorter.
In described clustering algorithm, cluster numbers is set as 150 classes.
Described step (3) concrete steps are as follows:
(3-1) utilize middle level features sorter to carry out feature extraction to the image in training set II, obtain middle level features figure;
(3-2) according to middle level features figure, binary conversion treatment is carried out to image, obtain binary image;
(3-3) add up foreground pixel in binary image and account for the ratio of view picture image and the connected region number of foreground pixel;
(3-4) two statistics step (3-3) obtained are as proper vector, and the matching result of corresponding image, as output, utilizes SVM Algorithm for Training image suitability assessment models.
In described step (4), the matching result obtaining the current image of described forecast set according to described middle level features sorter and suitability assessment models is specially:
(4-1) the middle level features sorter utilizing step (2) to obtain carries out feature extraction to the image in forecast set, obtains middle level features figure;
(4-2) according to middle level features figure, binary conversion treatment is carried out to image, obtain binary image;
(4-3) add up foreground pixel in binary image and account for the ratio of view picture image and the connected region number of foreground pixel;
(4-4) two statistics step (4-3) obtained, as in proper vector input adaptation assessment models, obtain the matching result of current image.
Describedly according to middle level features figure, binary conversion treatment is carried out to image and is specially:
Draw the grey level histogram of middle level features figure, travel through to low gray level from the high grade grey level grey level histogram, when the pixel count of a rear gray level is 3 times of last gray-level pixels number, get last gray level as threshold value, binary conversion treatment is carried out to image, obtains binary image.
Compared with prior art, the present invention has the following advantages:
1, adopt middle level features to describe image feature, effectively distinguish the common feature between allos radar image.
2, auto-correlation texture statistics amount is introduced as Partial Feature description amount, improve the accuracy of middle level features sorter.
3, introduce UNICOM's areal of ratio and foreground pixel that foreground pixel accounts for view picture image in binary image as feature interpretation amount, improve the accuracy of suitability assessment models.
4, people is added in the training process of sorter to the understanding of common feature between allos radar image, adopt supervised learning Strategies Training middle level sorter, improve the speed of feature extraction.
Accompanying drawing explanation
Fig. 1 is schematic flow sheet of the present invention.
Embodiment
Below in conjunction with the drawings and specific embodiments, the present invention is described in detail.The present embodiment is implemented premised on technical solution of the present invention, give detailed embodiment and concrete operating process, but protection scope of the present invention is not limited to following embodiment.
As shown in Figure 1, the present embodiment provides a kind of allos radar image suitability appraisal procedure based on supervised learning strategy, and the method comprises the following steps:
(1) by the image of known matching result to being divided into two parts, respectively called after training set I and training set II.Training set I and training set II comprise real-time diagram data (RAR respectively, Real Aperture Radar image) and benchmark graph data (SAR, Synthetic Aperture Radar image), in two training sets, positive sample size and negative sample quantity roughly equal, difference is for ± 5%.
(2) utilize training set I to train middle level features sorter, concrete steps are as follows:
(2-1) according to common feature priori, the common feature binary map of data in training set I is drawn by visual identification;
(2-2) common feature binary map is resolved into multiple image block, in the present embodiment, the size of each image block is 35 × 35, retains the image block that center exists effective pixel points, rejects residual image block, obtain effective block collection S;
(2-3) corresponding to all blocks in S SAR image region calculates Daisy descriptor;
(2-4) using K-means algorithm as clustering algorithm, the Daisy descriptor of each block, as cluster amount, carries out cluster to all blocks in S, and in the clustering algorithm of the present embodiment, cluster numbers is set as 150 classes;
(2-5) corresponding to block each in S SAR imagery zone calculates multiple gradient and auto-correlation statistic (Sketch-token statistic) respectively and forms middle level features and input as X, category feature label in clustering algorithm exports as Y, trains a decision forest;
(2-6) utilize decision forest training classifier, obtain middle level features sorter.
The final purpose of this method is the Adapter Property in order to assess image, therefore although the training essential of middle level features sorter is classification problem more than, but two classification problems can be reduced to, also namely judge whether current pixel is feature, and it belongs in 150 category features on earth which kind of is not paid close attention to.
(3) according to the middle level features sorter that step (2) obtains, utilize training set II to train image suitability assessment models, concrete steps are as follows:
(3-1) utilize middle level features sorter to carry out feature extraction to the image in training set II, obtain middle level features figure;
(3-2) grey level histogram of middle level features figure is drawn, travel through from the high grade grey level grey level histogram to low gray level, when the pixel count of a rear gray level is 3 times of last gray-level pixels number, get last gray level as threshold value, binary conversion treatment is carried out to image, obtains binary image
tb;
(3-3) binary image is added up
tbmiddle foreground pixel accounts for the ratio of view picture image and the connected region number of foreground pixel;
(3-4) two statistics step (3-3) obtained are as proper vector, and the matching result of corresponding image, as output, utilizes SVM Algorithm for Training image suitability assessment models.
(4) obtain forecast set to be assessed, obtain the matching result of the current image of described forecast set according to described middle level features sorter and suitability assessment models, be specially:
(4-1) the middle level features sorter utilizing step (2) to obtain carries out feature extraction to the image in forecast set, obtains middle level features figure;
(4-2) adopt step (3-2) described step to carry out binary conversion treatment to image, obtain binary image;
(4-3) add up foreground pixel in binary image and account for the ratio of view picture image and the connected region number of foreground pixel;
(4-4) two statistics step (4-3) obtained are as in proper vector input adaptation assessment models, obtain the matching result of current image, whether mate correctly, obtain the Adapter Property of allos radar image if can dope current image.
Claims (8)
1., based on an allos radar image suitability appraisal procedure for supervised learning strategy, it is characterized in that, the method comprises the following steps:
(1) by the image of known matching result to being divided into two parts, respectively called after training set I and training set II;
(2) training set I is utilized to train middle level features sorter;
(3) according to the middle level features sorter that step (2) obtains, training set II is utilized to train image suitability assessment models;
(4) obtain forecast set to be assessed, obtain the matching result of the current image of described forecast set according to described middle level features sorter and suitability assessment models.
2. the allos radar image suitability appraisal procedure based on supervised learning strategy according to claim 1, it is characterized in that, described training set I and training set II comprise real-time diagram data and benchmark graph data respectively.
3. the allos radar image suitability appraisal procedure based on supervised learning strategy according to claim 1, is characterized in that, in described training set I and training set II, the difference of positive sample size and negative sample quantity is ± 5%.
4. the allos radar image suitability appraisal procedure based on supervised learning strategy according to claim 1, it is characterized in that, described step (2) concrete steps are as follows:
(2-1) the common feature binary map of data in training set I is drawn;
(2-2) common feature binary map is resolved into multiple image block, retain the image block that center exists effective pixel points, reject residual image block, obtain effective block collection S;
(2-3) corresponding to all blocks in S SAR image region calculates Daisy descriptor;
(2-4) using K-means algorithm as clustering algorithm, the Daisy descriptor of each block, as cluster amount, carries out cluster to all blocks in S;
(2-5) corresponding to block each in S SAR imagery zone calculates multiple gradient respectively and auto-correlation statistic composition middle level features inputs as X, and the category feature label in clustering algorithm exports as Y, trains a decision forest;
(2-6) utilize decision forest training classifier, obtain middle level features sorter.
5. the allos radar image suitability appraisal procedure based on supervised learning strategy according to claim 4, it is characterized in that, in described clustering algorithm, cluster numbers is set as 150 classes.
6. the allos radar image suitability appraisal procedure based on supervised learning strategy according to claim 1, it is characterized in that, described step (3) concrete steps are as follows:
(3-1) utilize middle level features sorter to carry out feature extraction to the image in training set II, obtain middle level features figure;
(3-2) according to middle level features figure, binary conversion treatment is carried out to image, obtain binary image;
(3-3) add up foreground pixel in binary image and account for the ratio of view picture image and the connected region number of foreground pixel;
(3-4) two statistics step (3-3) obtained are as proper vector, and the matching result of corresponding image, as output, utilizes SVM Algorithm for Training image suitability assessment models.
7. the allos radar image suitability appraisal procedure based on supervised learning strategy according to claim 1, it is characterized in that, in described step (4), the matching result obtaining the current image of described forecast set according to described middle level features sorter and suitability assessment models is specially:
(4-1) the middle level features sorter utilizing step (2) to obtain carries out feature extraction to the image in forecast set, obtains middle level features figure;
(4-2) according to middle level features figure, binary conversion treatment is carried out to image, obtain binary image;
(4-3) add up foreground pixel in binary image and account for the ratio of view picture image and the connected region number of foreground pixel;
(4-4) two statistics step (4-3) obtained, as in proper vector input adaptation assessment models, obtain the matching result of current image.
8. a kind of allos radar image suitability appraisal procedure based on supervised learning strategy according to claim 6 or 7, is characterized in that, describedly carries out binary conversion treatment according to middle level features figure to image and is specially:
Draw the grey level histogram of middle level features figure, travel through to low gray level from the high grade grey level grey level histogram, when the pixel count of a rear gray level is 3 times of last gray-level pixels number, get last gray level as threshold value, binary conversion treatment is carried out to image, obtains binary image.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510097972.1A CN104751170B (en) | 2015-03-05 | 2015-03-05 | Heterologous radar image suitability appraisal procedure based on supervised learning strategy |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510097972.1A CN104751170B (en) | 2015-03-05 | 2015-03-05 | Heterologous radar image suitability appraisal procedure based on supervised learning strategy |
Publications (2)
Publication Number | Publication Date |
---|---|
CN104751170A true CN104751170A (en) | 2015-07-01 |
CN104751170B CN104751170B (en) | 2018-08-24 |
Family
ID=53590823
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201510097972.1A Expired - Fee Related CN104751170B (en) | 2015-03-05 | 2015-03-05 | Heterologous radar image suitability appraisal procedure based on supervised learning strategy |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN104751170B (en) |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102622607A (en) * | 2012-02-24 | 2012-08-01 | 河海大学 | Remote sensing image classification method based on multi-feature fusion |
US20140270347A1 (en) * | 2013-03-13 | 2014-09-18 | Sharp Laboratories Of America, Inc. | Hierarchical image classification system |
CN104102928A (en) * | 2014-06-26 | 2014-10-15 | 华中科技大学 | Remote sensing image classification method based on texton |
-
2015
- 2015-03-05 CN CN201510097972.1A patent/CN104751170B/en not_active Expired - Fee Related
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102622607A (en) * | 2012-02-24 | 2012-08-01 | 河海大学 | Remote sensing image classification method based on multi-feature fusion |
US20140270347A1 (en) * | 2013-03-13 | 2014-09-18 | Sharp Laboratories Of America, Inc. | Hierarchical image classification system |
CN104102928A (en) * | 2014-06-26 | 2014-10-15 | 华中科技大学 | Remote sensing image classification method based on texton |
Non-Patent Citations (1)
Title |
---|
杨朝辉: "异源合成孔径雷达景象匹配的评估技术研究", 《同济大学2009年度博士论文》 * |
Also Published As
Publication number | Publication date |
---|---|
CN104751170B (en) | 2018-08-24 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN102968637B (en) | Complicated background image and character division method | |
CN111832655B (en) | Multi-scale three-dimensional target detection method based on characteristic pyramid network | |
CN105956560B (en) | A kind of model recognizing method based on the multiple dimensioned depth convolution feature of pondization | |
CN104599275B (en) | The RGB-D scene understanding methods of imparametrization based on probability graph model | |
CN110276264B (en) | Crowd density estimation method based on foreground segmentation graph | |
CN103971097B (en) | Vehicle license plate recognition method and system based on multiscale stroke models | |
CN101329734B (en) | License plate character recognition method based on K-L transform and LS-SVM | |
CN103366190B (en) | A kind of method of identification traffic signss | |
CN110472467A (en) | The detection method for transport hub critical object based on YOLO v3 | |
CN108009518A (en) | A kind of stratification traffic mark recognition methods based on quick two points of convolutional neural networks | |
CN103198705B (en) | Parking place state automatic detection method | |
CN107491720A (en) | A kind of model recognizing method based on modified convolutional neural networks | |
CN103258434A (en) | Image border detecting system based on video traffic flow detection and vehicle identification | |
CN110287879A (en) | A kind of video behavior recognition methods based on attention mechanism | |
CN103065325A (en) | Target tracking method based on color distance of multicolors and image dividing and aggregating | |
CN108491828B (en) | Parking space detection system and method based on level pairwise similarity PVAnet | |
CN103679214A (en) | Vehicle detection method based on online area estimation and multi-feature decision fusion | |
CN105389561A (en) | Method for detecting bus lane based on video | |
CN113327248A (en) | Tunnel traffic flow statistical method based on video | |
CN109063630A (en) | A kind of fast vehicle detection method based on separable convolution technique and frame difference compensation policy | |
CN104573701B (en) | A kind of automatic testing method of Tassel of Corn | |
CN113538585B (en) | High-precision multi-target intelligent identification, positioning and tracking method and system based on unmanned aerial vehicle | |
CN105069419A (en) | Traffic sign detection method based on edge color pair and characteristic filters | |
Zhan et al. | Pedestrian detection and behavior recognition based on vision | |
CN105023231A (en) | Bus data acquisition method based on video recognition and cell phone GPS |
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 | ||
CF01 | Termination of patent right due to non-payment of annual fee | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20180824 Termination date: 20210305 |