基于双目图像的行人检测与定位系统研究

针对半全局匹配算法（Semi-Global Matching，SGM）的视差图匹配度较低的问题，以及目标检测算法Fast-YOLO对小目标的检测能力不足的问题，提出一种基于双目图像的行人检测与定位系统。系统首先利用图像分割块具有视差相似性的特点，使用基于快速图像分割的SGM算法对双目图像进行立体匹配，然后修改Fast-YOLO网络模型，提高网络分辨率，使用改进的Fast-YOLO网络进行行人检测。实验结果表明，基于快速图像分割的SGM算法较好地解决了匹配度较低问题，基于Fast-YOLO改进的行人检测网络明显地提高了对小目标的检测能力。系统实现了对行人的检测和定位，并使用GPU达到实时的计算效率。

Research of pedestrian detection and location system based on stereo images

Aiming at the problem of low matching rate of Semi-Global Matching (SGM) and the problem of poor performance for small targets detection of Fast-YOLO algorithm, a pedestrian detection and localization system based on stereo images is proposed. Firstly, the system take advantage of the characteristic that the same segmentation block has similar disparity, and use the SGM algorithm based on fast image segmentation for stereo matching. Then, the Fast-YOLO network model is modified to improve the network resolution and the improved Fast-YOLO network is used for pedestrian detection. Experimental results show that the SGM algorithm based on fast image segmentation solves the problem of low matching rate, and the improved pedestrian detection network based on Fast-YOLO obviously improves the detection ability of small targets. The system completes the detection and localization of pedestrian, and achieve real-time calculation by GPU.