A Novel Target Detection Method based on Visual Attention with CFAR

Based on visual attention theory and local probability density function statistical feature, a novel target detection method with Constant false alarm rate (CFAR) is proposed in this paper. Visual attention model mimics the effective and efficient visual system of primates to deal with complex scenarios. The proposed target detection algorithm inherits the advantages of both visual attention model and CFAR, which is applied to complex circumstances for target detection. By computing the phase of Fourier Transform, the saliency map is calculated by applying the adaptive Gaussian Filters. In order to extract the ground targets rapidly from CFAR detection images, the gradient feature is extracted to detect visual saliency area. By using watershed transform method, the segmentation image for target detection is obtained. Experimental results show that the adaptive Gaussian Filter could not only de-noise images effectively, but also can reserve as much original information as possible. The proposed method is proven to be capable of detecting ground targets in complex scenarios. In addition, the calculation procedure of the proposed method is pretty simple, which enables it to be suitable for engineering application.