CN107507211A - Remote sensing image segmentation method based on multi-Agent and MRF - Google Patents
Remote sensing image segmentation method based on multi-Agent and MRF Download PDFInfo
- Publication number
- CN107507211A CN107507211A CN201710605148.1A CN201710605148A CN107507211A CN 107507211 A CN107507211 A CN 107507211A CN 201710605148 A CN201710605148 A CN 201710605148A CN 107507211 A CN107507211 A CN 107507211A
- Authority
- CN
- China
- Prior art keywords
- remote sensing
- layer
- mrf
- texture
- sensing images
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
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
- G06T7/143—Segmentation; Edge detection involving probabilistic approaches, e.g. Markov random field [MRF] modelling
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/40—Analysis of texture
- G06T7/41—Analysis of texture based on statistical description of texture
- G06T7/45—Analysis of texture based on statistical description of texture using co-occurrence matrix computation
-
- 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
- G06T2207/10032—Satellite or aerial image; Remote sensing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20212—Image combination
- G06T2207/20221—Image fusion; Image merging
Abstract
The present invention relates to the remote sensing image segmentation method based on multi-Agent and MRF, solves the defects of being difficult to lift Remote Sensing Image Segmentation precision compared with prior art.The present invention comprises the following steps:The construction of first layer feature extraction layer;The construction of second layer dividing layer;The construction of third layer segmentation result fused layer;The input of remote sensing images, remote sensing images input feature vector extract layer;Split the acquisition of fusion results, the segmented result fused layer again after strength characteristic caused by feature extraction layer and the segmented layer of textural characteristics, produce image segmentation result.The present invention realizes the effective extraction and fusion of intensity, texture, spatial information by Agent system and markov random file technology, improves Remote Sensing Image Segmentation precision.
Description
Technical field
The present invention relates to remote sensing images technical field, the specifically Remote Sensing Image Segmentation side based on multi-Agent and MRF
Method.
Background technology
Remote sensing is that one proposed in the eighties of last century sixties can remote, the contactless skill for obtaining Ground Surface Dynamic information
Art.On the basis of the technologies such as current computer, graph image are furtherd investigate, remote sensing technology has also obtained quick hair
Exhibition, its effect in the development of the national economy and development of defense-related science and technology field are outstanding day by day.Remote sensing technology be implemented in electromagnetic wave with
On the interaction of atural object, terrestrial object information is obtained by receiving and handling clutter reflections or radiate the electromagnetic wave returned.
Remote Sensing Image Segmentation is one and establishes on remote sensing images feature, one established with reference to image distribution and feature
Door image interpretation technology.Remote sensing images are influenceed very greatly by kind of sensor and orientation, weather, wave band, resolution ratio, therefore remote sensing figure
There can be situations such as geometric distortion, noise jamming as in, design robustness is good, the algorithm of wide adaptation range is remote sensing image interpretation
The key issue in field.In image segmentation field, effective application of object edge, texture information can significantly improve testing result
Precision, but the effective extraction and fusion of these features are difficult points.Therefore, the collaboration extraction of Characteristics of The Remote Sensing Images how is realized
And the technical problem of a fast parallel urgent need technology for having become Remote Sensing Image Segmentation.
The content of the invention
The invention aims to solve the defects of being difficult to lift Remote Sensing Image Segmentation precision in the prior art, there is provided one
Kind is solved the above problems based on multi-Agent and MRF remote sensing image segmentation method.
To achieve these goals, technical scheme is as follows:
A kind of remote sensing image segmentation method based on multi-Agent and MRF, comprises the following steps:
The construction of first layer feature extraction layer;It is configured to realize remote sensing images in intensity, texture and Edge Gradient Feature
Feature extraction layer, wherein:Strength characteristic is the gray value of remote sensing images pixel, and textural characteristics are according to gray level co-occurrence matrixes mould
Type extracts;
The construction of second layer dividing layer, construct the segmentation module based on remote sensing images intensity and remote sensing texture;
The construction of third layer segmentation result fused layer, takes temporal voting strategy to construct segmentation result fused layer, and segmentation result melts
Close effective adaptive fusion that layer realizes segmentation result under multiple features;
The input of remote sensing images, remote sensing images input feature vector extract layer;
Split the acquisition of fusion results, passed through again after strength characteristic caused by feature extraction layer and the segmented layer of textural characteristics
Segmentation result fused layer, produce image segmentation result.
The construction of the second layer dividing layer comprises the following steps:
Structural strength MRF-Agent modules, intensity MRF-Agent modules are used to be split for remote sensing images intensity;
Texture MRF-Agent modules are constructed, texture MRF-Agent modules are used to be split for Remote Sensing Image Texture.
The input of the remote sensing images comprises the following steps:
Obtain remote sensing images;
Remote sensing images are defined with a step-length gray level co-occurrence matrixes T (N × N) by direction, one in units of pixel,
Wherein, M (i, j) be defined as gray level be i and j pixel and meanwhile appear in a point and along defined direction across
The frequency spent on the point of step-length, N are gray level division numbers;
Contrast, inverse difference moment entropy and auto-correlation are defined by step-length gray level co-occurrence matrixes T;
By step-length gray level co-occurrence matrixes T input feature vector extract layers, strength characteristic and textural characteristics are extracted.
The acquisition of the segmentation fusion results comprises the following steps:
By the strength characteristic input intensity MRF-Agent modules of remote sensing images, the segmentation based on remote sensing images intensity is obtained
As a result, intensity and space field information are obtained and;
The textural characteristics of remote sensing images are inputted into texture MRF-Agent modules, obtain the segmentation based on Remote Sensing Image Texture
As a result, texture and space field information are obtained and;
By remote sensing images intensity segmentation result and Texture Segmentation result input segmentation result fused layer, bond strength and space
Realm information, texture and space field information draw the segmentation result after fusion by temporal voting strategy method.
Beneficial effect
The remote sensing image segmentation method based on multi-Agent and MRF of the present invention, passes through multi-Agent compared with prior art
System and markov random file technology realize the effective extraction and fusion of intensity, texture, spatial information, improve remote sensing figure
As segmentation precision.
The feature extraction layer of the present invention, is made up of 2 Agent systems, realizes that remote sensing images intensity, texture carry respectively
Take;Dividing layer, markov random file (Markov random field are based on by 2:MRF Agent) is formed, real respectively
The now image segmentation based on intensity, texture, in this process, MRF realize strength information and space neighborhood information effectively from
Adapt to fusion, and effective adaptive fusion of texture information and space neighborhood information;Segmentation result fused layer, for merging not
With advantage of the feature in segmentation result, more preferable segmentation precision is realized.
Brief description of the drawings
Fig. 1 is the method precedence diagram of the present invention.
Embodiment
The effect of to make to architectural feature of the invention and being reached, has a better understanding and awareness, to preferable
Embodiment and accompanying drawing coordinate detailed description, are described as follows:
As shown in figure 1, a kind of remote sensing image segmentation method based on multi-Agent and MRF of the present invention, including it is following
Step:
The first step, the construction of first layer feature extraction layer.It is configured to realize that remote sensing images are special at intensity, texture and edge
The feature extraction layer of extraction is levied, wherein:Strength characteristic is the gray value of remote sensing images pixel, and textural characteristics are according to gray scale symbiosis
Matrix model extracts.By 2 Agent systems, the extraction of remote sensing images intensity, texture is realized respectively.
Second step, the construction of second layer dividing layer.Construct the segmentation module based on remote sensing images intensity and remote sensing texture.
This 2 MRF-Agent module of design, the segmentation result based on image intensity, texture is respectively obtained, and obtained while segmentation
Obtain the effective integration of intensity and space neighborhood information, texture and space neighborhood information.The characteristic of one pixel, more likely by
The influence of its surrounding pixel, with its more remote pixel of distance, the influence to its characteristic is smaller.It is comprised the following steps that:
(1) structural strength MRF-Agent modules, intensity MRF-Agent modules are used to be divided for remote sensing images intensity
Cut;
(2) texture MRF-Agent modules are constructed, texture MRF-Agent modules are used to be divided for Remote Sensing Image Texture
Cut.
3rd step, the construction of third layer segmentation result fused layer.Temporal voting strategy is taken to construct segmentation result fused layer, segmentation
As a result fused layer realizes effective adaptive fusion of segmentation result under multiple features, obtains more excellent image segmentation result.
4th step, the input of remote sensing images.Remote sensing images input feature vector extract layer, it specifically includes following steps:
(1) remote sensing images are obtained.
(2) remote sensing images are defined one by direction, one in units of pixel step-length gray level co-occurrence matrixes T (N ×
N),
Wherein, M (i, j) be defined as gray level be i and j pixel and meanwhile appear in a point and along defined direction across
The frequency spent on the point of step-length, N are gray level division numbers.
Gray level co-occurrence matrixes:The joint probability occurred simultaneously at a distance of two pixels for d.It is if right in gray level co-occurrence matrixes
Value on linea angulata is very big, illustrates that the direction has the identical pixel value at a distance of d, illustrates that change is not especially big.If gray scale is total to
Value very little in raw matrix on diagonal, might have frequently texture.The space of image can be studied according to this theory
Texture features, the texture contrast of image space is defined also according to this characteristic.
(3) contrast, inverse difference moment entropy and auto-correlation are defined by step-length gray level co-occurrence matrixes T, can be used for measuring image
Feature of the textural characteristics in brightness change, homogeney, uniformity etc..
(4) by step-length gray level co-occurrence matrixes T input feature vector extract layers, strength characteristic and textural characteristics are extracted.
5th step, split the acquisition of fusion results.It is segmented through strength characteristic caused by feature extraction layer and textural characteristics
Segmented result fused layer again after layer, produce image segmentation result.It is comprised the following steps that:
(1) by the strength characteristic input intensity MRF-Agent modules of remote sensing images, point based on remote sensing images intensity is obtained
Result is cut, and obtains intensity and space field information.
(2) textural characteristics of remote sensing images are inputted into texture MRF-Agent modules, obtains point based on Remote Sensing Image Texture
Result is cut, and obtains texture and space field information.
(3) remote sensing images intensity segmentation result and Texture Segmentation result are inputted into segmentation result fused layer, bond strength with
Space field information, texture and space field information draw the segmentation result after fusion by temporal voting strategy method.
General principle, principal character and the advantages of the present invention of the present invention has been shown and described above.The technology of the industry
For personnel it should be appreciated that the present invention is not limited to the above embodiments, that described in above-described embodiment and specification is the present invention
Principle, various changes and modifications of the present invention are possible without departing from the spirit and scope of the present invention, these change and
Improvement is both fallen within the range of claimed invention.The protection domain of application claims by appended claims and its
Equivalent defines.
Claims (4)
1. a kind of remote sensing image segmentation method based on multi-Agent and MRF, it is characterised in that comprise the following steps:
11) construction of first layer feature extraction layer;It is configured to realize remote sensing images in intensity, texture and Edge Gradient Feature
Feature extraction layer, wherein:Strength characteristic is the gray value of remote sensing images pixel, and textural characteristics are according to gray level co-occurrence matrixes model
Extraction;
12) construction of second layer dividing layer, the segmentation module based on remote sensing images intensity and remote sensing texture is constructed;
13) construction of third layer segmentation result fused layer, temporal voting strategy is taken to construct segmentation result fused layer, segmentation result fusion
Layer realizes effective adaptive fusion of segmentation result under multiple features;
14) input of remote sensing images, remote sensing images input feature vector extract layer;
15) split the acquisition of fusion results, passed through again after strength characteristic caused by feature extraction layer and the segmented layer of textural characteristics
Segmentation result fused layer, produce image segmentation result.
2. the remote sensing image segmentation method according to claim 1 based on multi-Agent and MRF, it is characterised in that described
The construction of two layers of dividing layer comprises the following steps:
21) structural strength MRF-Agent modules, intensity MRF-Agent modules are used to be split for remote sensing images intensity;
22) texture MRF-Agent modules are constructed, texture MRF-Agent modules are used to be split for Remote Sensing Image Texture.
3. the remote sensing image segmentation method according to claim 1 based on multi-Agent and MRF, it is characterised in that described distant
The input of sense image comprises the following steps:
31) remote sensing images are obtained;
32) remote sensing images are defined with a step-length gray level co-occurrence matrixes T (N × N) by direction, one in units of pixel,
Wherein, M (i, j) be defined as gray level be i and j pixel and meanwhile appear in a point and along defined direction span step
Frequency on long point, N are gray level division numbers;
33) contrast, inverse difference moment entropy and auto-correlation are defined by step-length gray level co-occurrence matrixes T;
34) by step-length gray level co-occurrence matrixes T input feature vector extract layers, strength characteristic and textural characteristics are extracted.
4. the remote sensing image segmentation method according to claim 1 based on multi-Agent and MRF, it is characterised in that described point
The acquisition for cutting fusion results comprises the following steps:
41) by the strength characteristic input intensity MRF-Agent modules of remote sensing images, the segmentation knot based on remote sensing images intensity is obtained
Fruit, and obtain intensity and space field information;
42) textural characteristics of remote sensing images are inputted into texture MRF-Agent modules, obtains the segmentation knot based on Remote Sensing Image Texture
Fruit, and obtain texture and space field information;
43) remote sensing images intensity segmentation result and Texture Segmentation result are inputted into segmentation result fused layer, bond strength and space
Realm information, texture and space field information draw the segmentation result after fusion by temporal voting strategy method.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710605148.1A CN107507211A (en) | 2017-07-24 | 2017-07-24 | Remote sensing image segmentation method based on multi-Agent and MRF |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710605148.1A CN107507211A (en) | 2017-07-24 | 2017-07-24 | Remote sensing image segmentation method based on multi-Agent and MRF |
Publications (1)
Publication Number | Publication Date |
---|---|
CN107507211A true CN107507211A (en) | 2017-12-22 |
Family
ID=60688834
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710605148.1A Pending CN107507211A (en) | 2017-07-24 | 2017-07-24 | Remote sensing image segmentation method based on multi-Agent and MRF |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107507211A (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108985181A (en) * | 2018-06-22 | 2018-12-11 | 华中科技大学 | A kind of end-to-end face mask method based on detection segmentation |
CN109816660A (en) * | 2019-02-19 | 2019-05-28 | 闽南师范大学 | A kind of image partition method, terminal device and storage medium |
CN117095299A (en) * | 2023-10-18 | 2023-11-21 | 浙江省测绘科学技术研究院 | Grain crop extraction method, system, equipment and medium for crushing cultivation area |
-
2017
- 2017-07-24 CN CN201710605148.1A patent/CN107507211A/en active Pending
Non-Patent Citations (3)
Title |
---|
HUIFANG BAO等: "A MRF-based Multi-agent system for Remote Sensing Image Segmentation", 《2017 INTERNATIONAL CONFERENCE ON COMPUTING INTELLIGENCE AND INFORMATION SYSTEM》 * |
KAMAL E.MELKEMI等: "Chaotic MultiAgent System Approach for MRF-based Image Segmentation", 《PROCEEDINGS OF THE 4TH INTERNATIONAL SYMPOSIUM ON IMAGE AND SIGNAL PROCESSING AND ANALYSIS (2005)》 * |
KAMAL E.MELKEMI等: "MRF and MultiAgent System based Approach for Image Segmentat ion", 《2004 IEEE INTERNATIONAL CONFERENCE ON INDUSTRIAL TECHNOLOGY (ICIT)》 * |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108985181A (en) * | 2018-06-22 | 2018-12-11 | 华中科技大学 | A kind of end-to-end face mask method based on detection segmentation |
CN108985181B (en) * | 2018-06-22 | 2020-07-24 | 华中科技大学 | End-to-end face labeling method based on detection segmentation |
CN109816660A (en) * | 2019-02-19 | 2019-05-28 | 闽南师范大学 | A kind of image partition method, terminal device and storage medium |
CN117095299A (en) * | 2023-10-18 | 2023-11-21 | 浙江省测绘科学技术研究院 | Grain crop extraction method, system, equipment and medium for crushing cultivation area |
CN117095299B (en) * | 2023-10-18 | 2024-01-26 | 浙江省测绘科学技术研究院 | Grain crop extraction method, system, equipment and medium for crushing cultivation area |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107239751B (en) | High-resolution SAR image classification method based on non-subsampled contourlet full convolution network | |
CN102708369B (en) | Sea ice parameter extraction method on basis of satellite image | |
CN103729842B (en) | Based on the fabric defect detection method of partial statistics characteristic and overall significance analysis | |
Huang et al. | A new synthetical method of feature enhancement and detection for SAR image targets | |
CN107507211A (en) | Remote sensing image segmentation method based on multi-Agent and MRF | |
CN101777181B (en) | Ridgelet bi-frame system-based SAR image airfield runway extraction method | |
CN103077515B (en) | Multi-spectral image building change detection method | |
CN104200452A (en) | Method and device for fusing infrared and visible light images based on spectral wavelet transformation | |
CN101441766B (en) | SAR image fusion method based on multiple-dimension geometric analysis | |
CN104094284A (en) | Image processor with edge-preserving noise suppression functionality | |
CN101587189B (en) | Texture elementary feature extraction method for synthetizing aperture radar images | |
CN108829711B (en) | Image retrieval method based on multi-feature fusion | |
CN104915672A (en) | Rectangular building extraction method and system based on high-resolution remote sensing image | |
Chen et al. | Automatic scaling of F layer from ionograms | |
CN107909083A (en) | A kind of hough transform extracting method based on outline optimization | |
CN102663385A (en) | Detection method for spot target on satellite | |
CN116630802A (en) | SwinT and size self-adaptive convolution-based power equipment rust defect image detection method | |
CN104050674B (en) | Salient region detection method and device | |
CN101739667A (en) | Non-downsampling contourlet transformation-based method for enhancing remote sensing image road | |
Kong et al. | Lightweight algorithm for multi-scale ship detection based on high-resolution SAR images | |
CN106971402B (en) | SAR image change detection method based on optical assistance | |
CN103065296B (en) | High-resolution remote sensing image residential area extraction method based on edge feature | |
CN105096274B (en) | Infrared image noise-reduction method based on non-down sampling contourlet domain mixing statistical model | |
CN115115573A (en) | Transform model-based rivet defect detection method and system | |
CN114022794A (en) | Unmanned aerial vehicle offshore oil contamination monitoring method based on fusion of multi-machine learning algorithm |
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 | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20171222 |
|
WD01 | Invention patent application deemed withdrawn after publication |