Abstract:In order to comprehensively enhance the ability of the reflection characteristics of the ground object spectrum on the remote sensing image and realize the unmistakable extraction of the UAV target, a target extraction technology of UAV remote sensing image based on image segmentation is proposed. With the support of the Haar-like filter structure, the quantization of low-level image features is completed according to the color features of the ground objects in the regional environment, and the acquisition of remote sensing depth maps for UAVs based on image segmentation technology is realized. Define the basic terms related to UAV remote sensing image, determine the average reflection characteristic level of the ground object spectrum through the method of original feature selection, obtain accurate feature element correlation measurement value, and complete the target fusion processing of UAV remote sensing image. On this basis, segment the image target under multi-resolution conditions, while defining the image texture and scale conditions, get the final feature extraction results, and realize the application of image segmentation technology based on image segmentation UAV remote sensing image target extraction. The comparison experiment results show that under the initial acquisition phase conditions and sub-pixel conditions, the spectral width of the target feature is more than 7.0μm, and the reflection characteristic capability of the remote sensing image is greatly improved, which meets the practical need for unmistakable extraction of UAV targets.
