Improved Wavelet Feature Extraction Methods Based on HSV Space for Vehicle Detection

The focus of this work is on the problem of feature extraction for vehicle detection. Feature extraction is a key point of pattern recognition. In particular, we propose using improved wavelet feature extraction approaches based on HSV space for rear-vehicle detection. Wavelet features are attractive for vehicle detection because they form a compact representation, encode edges, capture information from multi-resolution, and can be computed efficiently. Currently, the wavelet features based on coefficients and grayscale space are easily affected by the surroundings and illumination conditions and cause high intra-class variability. In order to deal with this problem, three improved wavelet feature extraction approaches based on HSV space are proposed. The experimental results indicate that the improved approaches based on HSV show super performance compared with the current methods based on both HSV space and Grayscale space. Furthermore, they also show better results than themselves based on Grayscale space.