Research On Multi-source Remote Sensing Image Water Detection Method

The water body information in high-resolution remote sensing images plays an increasingly important role in various fields,but some existing methods for detecting water body information often cannot meet the actual needs.The high-resolution visible light image and high resolution proposed in this paper are presented.The water detection technology in SAR images has good effect and practicability.The specific research mainly includes the following two parts:(1)For the detection of water body information in visible light images,this paper proposes an algorithm for extracting multiple features in SLIC-divided super-pixel blocks and then using BP neural network to identify them.The whole algorithm is divided into two parts,the training of sample data.And the water body detection of the data to be tested,firstly,the visible light image is subjected to noise reduction processing,and the visible light image is segmented by the SLIC algorithm.After the segmentation is completed,there is a super pixel block,and the data is divided into sample data and data to be tested,and then semi-automatically utilized.The marking program marks the water and non-water bodies in the sample data,extracts their features in the marked pixel blocks,puts them into the BP neural network to obtain the training model,and finally the water body detection of the data to be tested,and extracts the pixels in the data to be tested.The characteristics of the block,the training model derived from these data and training sample data is added to the BP neural network for water body region prediction,and the water body region in the data to be tested is obtained.The accuracy of the water body detection algorithm proposed in this paper is relatively high.(2)For the detection of water body region in SAR images,this paper proposes a circular sliding window template traversal algorithm,which firstly suppresses the speckle noise in the SAR image,and then traverses the entire SAR image with a circular sliding window template.Finally,according to the overlapping sliding window area,the connected area of the candidate pixel is counted.According to the area of the candidate water body connected area,the false detection area is eliminated and the missed detection area is eliminated.Experimental results show that this method can not only detect water body information well.Come out,and make the detected water body information and the boundary part of the land information smoother,there will be no obvious rectangular corners,and it is closer to the characteristics of the water body.By analyzing the experimental results,it is verified that the proposed algorithm can improve the accuracy of water detection in visible light images and SAR images.In order to verify the effectiveness of the proposed algorithm,the proposed algorithm and other different algorithms are compared.The experimental results show that the proposed algorithm for extracting multiple features in SLIC-divided super-pixel blocks and then using BP neural network for recognition and circular sliding window template traversal algorithm improves the accuracy of water detection.The accuracy of water detection is 89.55%,and the accuracy of water detection in SAR images reaches 91.25%.It is ideal to detect water bodies in visible light images and SAR images.