An Improved Faster R-CNN Method for Car Front Detection

With the further development of deep learning, object detection is widely used in vehicle detection in road environment and plays an indispensable role in the construction of intelligent transportation system. Aiming at the problems of low accuracy of current mainstream object detection algorithms and difficulty in detecting small objects, an improved Faster R-CNN algorithm is proposed. Firstly, in the feature extraction stage, the SE attention module is introduced to focus on the details in the image and improve the detection accuracy of the network model for small objects. Secondly, the Spatial Pyramid Pooling (SPP) structure is used to replace the pooling layer to enhance the robustness of the network and learn feature information of different sizes to further improve the detection accuracy of the model. Finally, experiments are conducted on the CAR-DATA dataset made by ourselves to train appropriate weights for testing. The experimental results show that compared with Faster R-CNN, the AP and APs of the proposed network model in this paper are improved by 1.29% and 1.74% respectively, and the car position can be accurately detected and located.