Improved Faster RCNN Object Detection

Chaojie Li ； Peng Zhou

《World Scientific Research Journal》 6卷3期 (2020/03) Pp. 74-81

https://doi.org/10.6911/WSRJ.202003_6(3).0010

摘要

Aiming at the problem that the low accuracy of Faster RCNN object detection algorithm, an improved stereo object detection method based Faster RCNN (Stereo Faster RCNN, S-Faster RCNN) is proposed and used to the vehicle detection. A two-layer feature extraction network is used to extract the left and right image features respectively, and then the features are connected and fed into the mixed RPN network to correlate the slight differences between the left and right images and train the RPN. The features of the associated left and right images will be conducted RoI Pooling operation respectively to form feature maps of fixed size. Finally, the corresponding categories of objects and the accurate location of the bounding boxes will be output through the full connected layers. The experimental results show that compared with the traditional Faster RCNN algorithm. The accuracy of the improved method improves by 11.3%, 9.5% and 6.3% respectively under the three standards of KITTI dataset: simple, medium and difficult.

關鍵字

Object detection ； Faster RCNN ； RPN ； vehicle detection

參考文獻

A. Karpathy and L. Fei-Fei, “Deep visual-semantic alignments for generating image descriptions,” in Proceedings of the IEEE conference on computer vision and pattern recognition, 2015, pp. 3128–3137.

Q. Wu, C. Shen, P. Wang, A. Dick, and A. van den Hengel,“Image captioning and visual question answering based on attributes and external knowledge,” IEEE transactions on pattern analysis and machine intelligence, vol. 40, no. 6,pp. 1367–1381, 2018.

K. Kang, H. Li, J. Yan, X. Zeng, B. Yang, T. Xiao, C. Zhang,Z. Wang, R. Wang, X. Wang et al., “T-cnn: Tubelets with convolutional neural networks for object detection from videos,” IEEE Transactions on Circuits and Systems for Video Technology, vol. 28, no. 10, pp. 2896–2907, 2018.

JOSEPH R, SANTOSH D, ROSS G, et al. You only look once: Real-time object detection [C]//IEEE Conference on Computer Vision and Pattern Recongnition, Lasvegas, IEEE, 2016:27 - 30.

GIRSHICK R, DONAHUE J, DARRELL T, et al. Rich feature hierarchies for accurate object detection and semantic segmentation [C] //IEEE Conference on Computer Vision and Pattern Rec-ognition, Honolulu, IEEE, 2014:580 – 587.