Looking for breakthrough ideas for innovation challenges? Try Patsnap Eureka!

Multi-source Image Change Detection Method for Cluster-Guided Deep Neural Network Classification

A technology of deep neural network and change detection, which is applied in the field of multi-source image change detection of cluster-guided deep neural network classification, can solve the problems of dimensionality reduction, difficulty in adapting to production requirements, difficulty in meeting precision, etc., and achieve accuracy High, excellent feature learning ability, and the effect of solving classification problems

Active Publication Date: 2018-11-20
XIDIAN UNIV
View PDF2 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

For the multi-source image change detection method acquired by different sensors, also known as multi-source image processing, scholars from various countries have studied a large number of change detection methods and theoretical models from different perspectives. Traditional methods such as algebraic methods, time series analysis methods, etc. , according to the image difference method or ratio method to generate a difference map, and then select the threshold. The algorithm of this type of method is simple and easy to implement, and some transformation methods can effectively reduce the dimension. The disadvantage is that it is difficult to overcome due to atmospheric conditions, sensor noise and The interference caused by the difference of atmospheric radiation affects the final detection results
With the deepening of the application of change detection methods, the accuracy of the detection results of the change detection method that first generates the difference map for the images obtained by different sensors is difficult to meet the current image processing requirements, and it is difficult to adapt to the current human production needs.

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Multi-source Image Change Detection Method for Cluster-Guided Deep Neural Network Classification
  • Multi-source Image Change Detection Method for Cluster-Guided Deep Neural Network Classification
  • Multi-source Image Change Detection Method for Cluster-Guided Deep Neural Network Classification

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0030] The present invention is a multi-source image change detection method for clustering-guided deep neural network classification. In the present invention, the multi-source image to be processed includes an optical image and a TM image. Grayscale matrix of optical images in images acquired by different sensors. see figure 1 , image change detection includes the following steps:

[0031] (1) Input the optical image to be tested, hereinafter referred to as the optical image: convert the optical image whose data type is three-dimensional into a grayscale matrix whose data type is two-dimensional, and input the grayscale matrix of the optical image.

[0032] (2) Segmentation of optical images: Fuzzy C-means clustering method is used to perform fuzzy clustering on the grayscale matrix of the optical image to be detected to obtain the grayscale matrix after clustering and segmentation of optical images. The fuzzy C-means clustering method is referred to as FCM .

[0033] (3)...

Embodiment 2

[0046] The multi-source image change detection method that clustering guides deep neural network classification is the same as embodiment 1, wherein the process of the training stacked autoencoder SAE described in step (6) includes:

[0047] (6a) The stacked autoencoder is composed of two layers of sparse autoencoders, and the activation value of the hidden layer nodes of the first layer of sparse autoencoder is used as the input of the second layer of sparse autoencoder.

[0048] (6b) Set the input layer of the first layer of sparse autoencoder to 9 nodes, the hidden layer to 49 nodes, the input layer of the second layer of sparse autoencoder and the hidden layer of the first layer of sparse autoencoder The nodes are the same, set to 49 nodes, and the hidden layer of the second-layer sparse autoencoder is set to 10 nodes.

[0049] (6c) Initialize the weights on the stacked autoencoder SAE with random numbers on the interval [0, 1].

[0050] (6d) Input the selected training s...

Embodiment 3

[0056] The multi-source image change detection method of clustering-guided deep neural network classification is the same as embodiment 1-2, wherein the specific steps of carrying out block sampling to image data described in step (4) include:

[0057] 4.1: If image X 0 The dimension is p×q. In this example, 3×3 is taken as an example to describe in detail. An all-zero matrix X with a dimension of (p+2)×(q+2) is initialized.

[0058] 4.2: Put the image matrix X 0 The value of each element in is assigned to all elements in the rectangular range from the 2nd row to the p+1th row, the 2nd column to the q+1th column in the all-zero matrix X;

[0059] 4.3: Assign all elements of row 3 in the all-zero matrix X to all elements of row 1, and assign all elements of row p to all elements of row p+2.

[0060] 4.4: Assign all the elements in the third column of X to all the elements in the first column, and assign all the elements in the qth column to all the elements in the q+2th colum...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

PUM

No PUM Login to View More

Abstract

The invention discloses a multi-source image change detection method based on clustering guided deep neural network classification, which avoids the step of difference figure generation in the early stage in traditional change detection and overcomes the defect that difference figure generation is needed in multi-source image change detection. The method comprises the following steps: inputting a gray matrix of an optical image; carrying out fuzzy clustering on the optical image to get a segmented gray matrix; marking the optical image after clustering segmentation; sampling the optical image and a TM image; selecting a training sample from the TM image; training a stacked sparse automatic encoder SAE; using a tag to fine-adjust network parameters; inputting the TM image to a network, and outputting a classified image; working out the log ratio of the two classified images; and getting a change detection result. The link of difference figure construction is abandoned. The method is applicable to multi-source remote sensing image change detection, and has the advantages of little noise influence, high precision of change detection result classification, and the like.

Description

technical field [0001] The invention belongs to the technical field of image processing, mainly relates to the combination of deep neural network and remote sensing image processing field, mainly solves the problem of change detection of remote sensing images, and specifically provides a multi-source image change detection method guided by clustering deep neural network classification, which is used for The change detection of multi-source images is widely used in aerospace, ground object coverage and utilization, earthquake disaster detection and evaluation and other fields. Background technique [0002] The continuous development of computer digital image processing, pattern recognition, artificial intelligence, and sensor data fusion technology provides more technical support for automatic change detection of remote sensing images. In the past two decades, the change detection methods of remote sensing images have been continuously updated, the change detection technology...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

Application Information

Patent Timeline
no application Login to View More
Patent Type & Authority Patents(China)
IPC IPC(8): G06T7/00G06T7/10
CPCG06T7/0002G06T2207/10032G06T2207/20081G06T2207/20084G06T2207/30181
Inventor 马文萍李志舟焦李成马晶晶张普照赵暐
Owner XIDIAN UNIV
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Patsnap Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Patsnap Eureka Blog
Learn More
PatSnap group products