UAV Image Building Recognition Based On Superpixel Segmentation

In the field of UAV image processing and computer vision,it is of great significance to extract high-precision and reliable information of buildings,roads,bridges and other artificial ground objects from UAV images.Simple linear iterative clustering algorithm is generally recognized as a super-pixel segmentation algorithm with the best comprehensive performance at present.However,when SLIC algorithm is used for image segmentation,it still has the problems of randomly selecting seed points,artificial determination of K value and isolated small areas in the segmentation results.The segmented images are processed in isolated small areas,and the features of the processed images are extracted and the SVM support vector machine is used to recognize the UAV image.The work includes the following aspects:(1)In view of the existing shortcomings of the fast density peak algorithm,the information entropy and min-max standardization are introduced to improve the artificially set cutoff distance and use the decision map to select the clustering center for improvement,and the improved CFSFDP algorithm is used for clustering results and performance testing the experimental results show that the improved CFSFDP algorithm has better clustering effect and higher accuracy than the original CFSFDP algorithm.(2)The improved CFSFDP algorithm and the superpixel optimization merge function were combined into the SLIC algorithm.The optimized SLIC algorithm randomly selected seed points and artificially determined the deficiency of K value,then the improved SLIC algorithm was used for image segmentation,and the Bhattacharyya coefficient was used as the measurement standard to merge the isolated small areas after segmentation.(3)Feature extraction was carried out on the image after segmentation and combined processing.Texture feature and color feature were used as feature vectors of the classifier.SVM classifier was used for classification and building identification of image.