“…Alba Flores[9] presented two threshold algorithms, threshold and Ostu threshold method, to detect vehicles by clustering analysis on the highway in the United States from 1 m resolution IKONOS panchromatic images. Experiments show that vehicle detection and counting has the very good effect, the error is less than 10%, the main reason for the error is caused by a shadow over roads, ramps, trees and the traffic jams.Xiaoying Jin Curt and h. Davis[10,11] using morphological neural network weights sharing technology road to divide image pixels into targets and non-targets. A. Gerhardinger et al[12] evaluated the potential of vehicle detection and counting from high resolution satellite imagery, and compared the two sources, IKONOS and QuickBird data, put forward a method to enhance the classification performance of image preprocessing, and tested automatic feature extraction process.…”