Research on Vehicle Object Detection Method Based on Convolutional Neural Network

In the current systems of vehicle object detection, there exist three major drawbacks: inappropriate manual feature selection, difficulty of small object recognition and low detection rate of the whole system. A new vehicle object detection based on Faster R-CNN is proposed. Firstly, a new anchor scale of 642 is added into the system. Then, the object detection problem in the scenario is converted into a binary object classification problem. Compared with Faster R-CNN, the modified Faster R-CNN has an obvious improvement in detection accuracy. Finally, simulation results are given to verify the validity of the modified Faster R-CNN method and improve the precision.