Abstract: Intelligent detection of traffic targets is the basic technology in the intelligent collision avoidance system for vehicle blind spots. The research and application of this technology is of great significance to reduce the loss of traffic accidents. In this paper, a detection model for vehicle blind zone collision avoidance system design, which combines two performance improvement strategies in the basic model. The model is applied to domestic and foreign road scene detection data sets to verify the detection performance of the model in all ranges and short-range targets. Experimental results show that the model can show high detection accuracy for close targets and has a high detection speed. Therefore, the model can be applied to the detection requirements of the intelligent blind zone collision avoidance system for traffic targets when the vehicle starts at low speed or when turning.