Edge Detection Research Used For Vehicle Identification

Edge detection is an important tool in image processing and analysis. Image edge detection and extraction of features, has been a hot spot of research in image processing and analysis techniques. Through the vehicle identification and edge detection research, This paper make the study of edge detection has practical relevance and closely link with the production and living.Paper use the study system of the traditional edge detection methods and research as a starting point to analysis and comparison strengths and weaknesses of the known algorithms step-by-step, sum up the difficulties and focus of the existing edge detection, than it give the targeted improvement algorithm. Because the wavelet transform has good time-frequency localization properties and multi-scale analysis capabilities in the gray image of the vehicle, we give the improved multi-scale edge detection algorithm based on Contourlet Transform. Contourlet transfom design adaptive threshold associate with multi-scale wavelet transfonn use modulus maxima method to get the vehicle edge images at different scales. In order to effectively identify the vehicle image information, and remove noise introduced by pseudo-video content, reducing the influence of noise to enhance the robustness of the edge detection of the vehicle identification.Focus on the Gabor filter applications dimension and the computational load of the bottleneck problem, the paper gives the vehicle identification algorithm based on edge features. Based on the geometric characteristics of the vehicle, the algorithm separate the vehicle from the background, through the background subtraction algorithm to extract the Gabor features. The paper use non-uniform sampling strategy to intensive sample in key component of the sample image and sparse sample in non-critical parts to settle the Gabor dimension in the feature extraction and computational load. The algorithm compare the Gabor set of the identifying image to Gabor set of different template images, It confirm the types of identifying vehicles by the highest similarity values. The experiments show the algorithm reducing the dimension of Gabor feature vector effectively without lowering the recognition rate.