Experimental study on vehicle extraction based on wv-2 image data

Abstract

Real-time dynamic traffic information is an important basic information source for traffic monitoring and management. In view of the limitations of existing ground traffic detection equipment, people hope to be able to obtain a larger range of real-time dynamic traffic flow data with higher application value to monitor road traffic conditions far from the target. In this paper, features of wv-2 image data are analyzed. Firstly, gray scale transformation, filtering and mathematical morphology are used to preprocess the remote sensing image. Then, edge detection is used to extract the road and limit the area for vehicle identification. Then the method of double threshold and support vector machine is used to identify the vehicles on the road. Finally, the accuracy of the recognition results is analyzed. In the experimental results, the accuracy of the recognition using the double threshold method is 82.7%, and that of the support vector machine method is 95.2%. The latter has a better recognition rate.