Desert Vehicle Detection Using Adaptive Visual Attention and Pulse Coupled Neural Network

Considering deficiency of existing methods in desert vehicle detection, this paper proposes a novel approach for vehicle detection in desert based on adaptive phase spectrum of quaternion Fourier transform (APQFT) combined with pulse coupled neural network (PCNN) .We propose an adaptive visual attention model which uses information of background, image colors and intensity to generate a visual saliency map by parallel computing.Moreover, this model uses bilinear interpolation to improve computing efficiency.After that, using PCNN extracts regions of interests (ROIs) .Then, using scale-invariant feature transform (SIFT) feature extracted from ROIs detects vehicle areas which combines with hierarchical discriminant regression (HDR) tree.The experimental results shows that compared with the morphology and the SVM methods, the recognition rate of proposed approach increases 5.8%and 15.4%respectively.