Research on the detection of the UAV remote sensing chili images based on superpixel segmentation and SVM

The collection of agricultural field sample sets needs generally to be carried out according to the crop growth cycle, and it is not easy to obtain large amounts of data. For specific crop detection problems, it often takes time and effort to perform data labeling. Aiming at the problem of detecting the unmanned aerial remote sensing chili images, this paper proposes a recognition framework of the chili images, which mainly contains three processes. Firstly, the classifier is designed, and the remote sensing image is automatically sliced to make the sample set. Then, the SVM classifier is trained. Secondly, it is the detection process. The input image is segmented into superpixel to generate candidate region, and then the regions are classified using a classifier to complete the detection process. Finally, the results of detection are fine-tuned by the method of template matching, and the sample set is updated with a part of the detection result in chili image patches.