Research on Cone Bucket Target Detection Based on Improved Faster R-CNN Deep Network

To solve the problem of detecting the cone bucket target on both sides of the FSAC racing car, a new target detection method based on Faster R-CNN is proposed. The method uses a deep residual network as a feature extraction network to fully extract sample data features. In the Region Proposal Network (RPN), the anchor boxes are divided into finer parts to enhance the detection effect on the small target cones. Optimize the Non-maximum Suppression (NMS) algorithm to extract the proposal box and reduce the missed detection rate of the adjacent cone bucket target. The target detection network model is obtained by training on the self-made cone bucket dataset. The experimental results show that the improved algorithm based on Faster R-CNN is robust to the detection of cone bucket targets.