Abstract:

Taking the road under the influence of occlusion and perspective in the surveillance video as the research object, aiming at the difficult problem of vanishing point detection and instance segmentation under segmentation caused by the occlusion of pedestrian and driving on the road boundary line, an improved instance segmentation road detection method is proposed. Firstly, the road area is extracted by instance segmentation, and then the convex hull algorithm is used to compensate the occlusion of pedestrians and driving on the road boundary line. Finally, the road area is simulated into a trapezoid model which conforms to the scene structure, so as to optimize the detection of the road. The experimental results show that this method can solve the problem that the road is affected by occlusion and perspective, which leads to inaccurate detection. It can meet the needs of road detection under the security monitoring video, and has practical significance.

Key words: instance segmentation, road detection, image fitting, occlusion, perspective

摘要：

以安防监控视频下受遮挡与透视影响的道路作为研究对象，针对行人、行车对道路边界线的遮挡造成的消失点难以检测和实例分割欠分割难题，提出了一种改进的实例分割道路检测方法。首先使用实例分割对道路区域进行提取，然后通过凸包算法补偿行人、行车对道路边界线的遮挡，最后将其拟合成符合场景结构的梯形模型，从而优化对道路的检测。实验结果表明，该方法解决了道路受遮挡与透视影响而导致检测不准确的问题，可以满足安防监控视频下道路检测的需求，具有实际意义。

关键词: 实例分割, 道路检测, 图像拟合, 遮挡, 透视