Unlock instant, AI-driven research and patent intelligence for your innovation.

A cloud detection method for remote sensing images based on convolutional neural network

A technology of convolutional neural network and remote sensing image, which is applied in the direction of instrumentation, scene recognition, calculation, etc., can solve the problem that it is difficult to adapt to remote sensing image and achieve good robustness

Active Publication Date: 2019-02-19
BEIHANG UNIV
View PDF2 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

However, most of the existing cloud detection algorithms only use the low-dimensional features of the image, and it is difficult for them to adapt to remote sensing images with complex backgrounds, especially when there are thin clouds with low contrast to the background in the image.

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
  • A cloud detection method for remote sensing images based on convolutional neural network
  • A cloud detection method for remote sensing images based on convolutional neural network
  • A cloud detection method for remote sensing images based on convolutional neural network

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0045] Example 1: An example of remote sensing image cloud detection with a simple background

[0046] Step 1: Establish training sample set

[0047] Firstly, the ground truth is manually marked for the selected sample image, and the cloud region is marked out. Then N sub-blocks of size K×K are selected from the cloud area as positive samples, and 6*N sub-blocks of size K×K are selected from the non-cloud area as negative samples. In this experiment, N takes 42000 and K takes 55.

[0048] Step 2: Convolutional Neural Network Classification Model Generation

[0049] Build a convolutional neural network with 6 layers, where the first 4 layers are convolutional layers and the last 2 layers are fully connected layers. The input of the network is an RGB three-channel remote sensing image sub-block with a size of 55×55, and the output of the network is two values, which respectively represent the probability of being a cloud and the probability of not being a cloud. For the conv...

Embodiment 2

[0067] Example 2: Example of Cloud Detection in Remote Sensing Image with Complicated Background

[0068] At this time, there is no need to perform steps 1 and 2 in Example 1, and the detection results can be obtained by directly performing steps 3 to 6 in Example 1. Figure 3(a) is the original remote sensing image with complex background; Figure 3(b) is the result of superpixel clustering and segmentation using SLIC algorithm; Figure 3(c) is the cloud probability map generated by convolutional neural network; Figure 3 (d) is the cloud detection result map after refining the cloud probability map.

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

A method for cloud detection of remote sensing images based on convolutional neural networks, comprising the following steps: 1. Establishment of training sample sets; 2. Generation of classification models of convolutional neural networks; 3. Extraction of superpixel sub-regions; 4. Generation of cloud probability maps ; 5. Rough detection of cloud area; 6. Fine detection of cloud area; Through the above steps, we can detect the cloud area in the remote sensing image more correctly, and for the remote sensing image with complex background or containing translucent clouds, we can also obtain Better detection results can solve the problems of wrong judgment and analysis that the cloud may cause to analysts, and bring convenience to its subsequent processing and analysis. The invention can make the cloud area have better edges in the final cloud detection result, and has better robustness to complex scenes, so that better detection results can be obtained.

Description

technical field [0001] The invention provides a remote sensing image cloud detection method based on a convolutional neural network, which belongs to the field of optical remote sensing image processing. Background technique [0002] With the rapid development of remote sensing technology, remote sensing images have been widely used in many fields such as surveying, geographic mapping and resource monitoring. However, under normal circumstances, there will be clouds in remote sensing images, which will not only cause the loss of collected information, but also bring difficulties to subsequent processing such as target detection or target recognition, so that analysts can get wrong analysis results. Therefore, cloud detection and cloud removal have become one of the most important problems to be solved in remote sensing image processing. [0003] At present, cloud detection methods in remote sensing images mainly include: segmentation method based on physical threshold and p...

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): G06K9/00G06K9/34G06K9/62
CPCG06V20/13G06V10/267G06F18/2321
Inventor 谢凤英资粤史蒙云姜志国尹继豪史振威张浩鹏
Owner BEIHANG UNIV