Abstract
Video surveillance systems can be applied in coal mines for remote monitoring and for production control. Stitching video images into a panorama enhances the usability of video systems, since a panorama offers a wider view than single images do. But there are big challenges when conventional image stitching methods are applied to the domain of coal mine, especially in the mining faces. These challenges consist of non-uniform illumination, missed scenes and oblique panoramas. In this paper, a robust method was proposed to solve these three problems: (i) to overcome the non-uniform illumination on a mining face, the wide dynamic range technology and the histogram matching algorithm were used to enhance single images and reduce differences among images, respectively; (ii) to eliminate the missed scenes, overlapped images were quickly taken, then the feature matching method and template recognition method were adaptively used to achieve robust stitching and (iii) to mitigate the obliqueness of panoramas, vertical correction technology was used, which exploited the posture information of the camera. Next, the adjacent panoramas were concatenated and experiments were conducted on a fully mechanized mining face. The results show that the proposed method solves these three problems well and a dynamic panorama of the partial long-wall mining face is outputted. The research provides a new approach for displaying extended scenes of stope faces in intelligent collieries.
Multi Image Stitching with Cylindrical Surface Base on Local Feature Matching for Solving the Distortion Problem
