Blind Spot Detection System in Vehicles Using Fusion of Radar Detections and Camera Verification

摘要

Sensors are the quintessential part of Blind Spot Detection (BSD) systems, which have a profound effect on the performanceof the system. Every sensor has its unique deficiencies that can deteriorate the performance of the system under grievouscircumstances. Hence, making vital tasks in BSD such as object detection arduous. Indeed, previous studies have demonstratedthat data fusion techniques can diminish the adverse effects of sensors and improve detection accuracy in the BSDsystem. One of the main advantages of data fusion is to improve detection accuracy and reduce the processing time bymultiple sensors cooperation. We propose a BSD model that objects are detected in consecutive time intervals in the BSDsystem. Then, association techniques are employed for multi-sensor fusion since all sensors data are not ordinarily ready forfusion simultaneously. It should be noted that the orthodox approach in data association techniques in BSD often includes aglobal nearest neighbor, joint probabilistic data association, and multiple hypothesis tests. We simulate and compare thesetechniques by tracking multiple targets and multi-sensor fusion using virtual data in MATLAB. Furthermore, we illustratethat our multi-sensor fusion detection accuracy in the BSD system is augmented compared to a single sensor BSD system.