Application Research Of Video Detection And Recognition On Safety Monitoring For Workers Of Train Maintenance

There a complicated working environment and many workers in the site of train maintenance, so it is most likely to lead to the conditions of invading operating lines, crossing operating lines or climbing up trains to cross operating lines, it is particularly important to ensure the safety of workers of the site of train maintenance. The existing video monitoring systems of the site of train maintenance are the important way to ensure that the workers are safe. These existing video monitoring systems have no the function of automated monitoring.The technologies of video detection and recognition bring unprecedented development to all kinds of video surveillance systems, so it is of important practical significance that the two technologies are applied to safety monitoring for workers of train maintenance, which can be early warning of workers.A moving object detection algorithm based on background modeling of GMM and C-means clustering is adopted on the basis of analyzing several moving object detection algorithms for the complicated working environment. Background modeling of GMM to images can adapt to the variability of external environment well, so the binary areas extracted are close to the real situation before clustering of moving object areas, then the binary areas of binary images are classified with C-means clustering algorithm to more accurately detect moving objects. The experimental analysis shows that moving object detection can get good results if the clustering value is properly chosen.The binary areas of moving objects are classified by BP neural net after moving object detection, and then the shape features of the human body areas are selected as the inputs of neural net, at last, the output is human or inhuman. The classification features of positive and negative sample images are entered into neural net in order to finish learning and training of neural net, after which live images taken can be recognized in real time.A safety monitoring system for workers of train maintenance based on video detection and recognition technologies is realized by applying the two technologies to existing video monitoring system and this system can monitor the workers of train maintenance automatically.