North China Electric Power University (Baoding)
The Science and Technology Project of Shenzhen Power Supply Bureau Co., Ltd. funded "Research and Application of Project Management and Control Based on Visualization Technology"
近几年，可视化设备应用的普及，使得目前图像信息资源庞大，图像信息作为一种极为重要的信息，它与整个人类的生活息息相关。同时，我国工程建设行业蓬勃发展，使得施工现场数目迅速增加，施工安全问题的重要性成为亟需解决的问题，如何充分利用目前已有信息资源实现由于围栏摆放不合规导致的施工安全隐患的检测与告警，文章提出了一种基于Open CV的围栏合规性摆放检测方法。利用Open CV对目前电力施工现场可视化设备所收集的海量视频图像信息，对于施工现场围栏摆放的合规性进行检测。通过对于施工现场图片的处理，首先对于图象中目标围栏部分进行预处理，并通过连通区域分析算法与区域生长算法相结合，实现对于该围栏群围栏部分的提取以及缺口存在性的初步判断，然后训练专用于检测围栏缺口的分类器对于存在缺口的围栏群进行再次检测，并对检测结果中缺口数量进行统计。通过对于测试集样本进行检测，分析分类器检测结果，总结并解决分类器检测结果不准确的问题，对分类器重新训练并优化，最终该算法可以实现对于围栏摆放合规性的判断。通过一系列图像处理算法的应用以及专用分类器的训练，以缺口数量作为判断围栏摆放是否合规的突破口，首次实现了使用Open CV对电力施工现场围栏群摆放是否合规的检测，并为该类特征模糊物体的检测拓展了思路.
In recent years, the popularization of the application of visualization equipment has made a huge image information resources. As an extremely important information, image information is closely related to the whole human life. At the same time, the vigorous development of our country's engineering construction industry has led a rapid increase in the number of construction sites, and the importance of construction safety has become an urgent problem to be solved. How to make full use of existing information resources to realize the detection and warning of hidden dangers in construction safety caused by non-compliant fence placement, A method for detecting compliance of fences based on Open CV is proposed. Open CV is used to detect the compliance of the construction site fence placement on the massive video image information collected by the current power construction site visualization equipment. Through the processing of the construction site pictures, the target fence part in the image is first preprocessed, and the connected area analysis algorithm is combined with the region growth algorithm to realize the extraction of the fence part of the fence group and the preliminary judgment of the existence of the gap. Then, a classifier dedicated to detecting fence gaps is trained to re-detect the fence groups with gaps, and the number of gaps in the detection results is counted. By testing the samples of the test set, analyzing the detection results of the classifier, summarizing and solving the problem of inaccurate detection results of the classifier, and retraining and optimizing the classifier, the algorithm can finally realize the judgment of the compliance of fence placement. Through the application of a series of image processing algorithms and the training of special classifiers, the number of gaps is used as a breakthrough for judging whether the fences are placed in compliance. It expands the idea for the detection of such characteristic fuzzy objects.
余萍,宋祥宇. 基于Open CV的围栏摆放合规性方法研究[J]. 科学技术与工程, , ():复制