Investigating different methods of vehicle detection from aerial images

Abstract

Superpixel segmentation method designed for aerial images for controlling segmentation with low failure rate is suggested for detecting vehicles from aerial maps with high resolution and accuracy. For greater efficiency of practice and recognition, significant areas are extracted based on segmented superpixel centers. After segmentation, iteration strategy of sample selection based on scattered provision is used to provide a small and complete training subset of the original set. Selected training subset is a method that has the ability to distinguish and differentiate to detect vehicles. Training and detection for the features network of histogram of oriented gradient (HOG) is used to extract features.