Vehicle Detection in High-resolution SAR Image Using Saliency Visual Attention Method

Benefiting from the characteristics of low-cost, small-size, and operational capacity of all-day and all-weather, high-resolution millimeter-wave (mmWave) synthetic aperture radar (SAR), is widely used in advanced driving assistance system (ADAS). In this paper, inspired by human visual attention mechanism, a saliency visual attention algorithm is first proposed to suppress background noise while prominent foreground targets. After generating the saliency map, a watershed-based SAR image segmentation algorithm is used to Figure out the candidate regions. Finally, image morphology analysis and connected components analysis is adopted to reduce false alarms and detect the vehicle in the imaging scene. Experimental results based on measured data in a 77GHz automotive radar platform show that the proposed method is well worked and has incremental performance in parking plot scenes.