Screening Of Irregularities In Train Driving Based On Intelligent Video Analysis

Object tracking and action recognition are the core issues of intelligent video analysis.It is of great significance to study related algorithms to improve the performance of intelligent video monitoring system.The main works are as follows:1.An effective algorithm based on improved correlation filters for long-term visual tracking was proposed.Long-term accurate visual tracking is a great challenge.Most of tracking algorithms use simple templates for short-term tracking and generally fail when the target objects undergo heavy occlusion,pose variation and motion blur.We use correlation filters to train the tracking model.We adopted a high-confidence model update strategy to ensure the correct detection.To restore the false model caused by occlusion,we construct a conditional target re-detection mechanism.The experimental results indicate that the proposed tracker improved its robustness and accurate to heavy occlusion.2.An action recognition algorithm based on multi-feature fusion was proposed.To address the research of behavioral recognition stage,the traditional Bag of features ignores such issues as the timing in the action and the traits of some features themselves.We extracted HOF,HOG,and CN features for feature fusion.We used a weighted BOF coding method to add time information.Finally,the weighted BOF features are classified by the SVM classifier.Experimental results show that the algorithm can effectively improve the accuracy of action recognition.3.A smart screening system was designed for train driver violations in the monitoring videos.The proposed algorithms are applied to the intelligent screening system,which enables the intelligent screening of train driver irregular driving behaviors in massively offline surveillance videos.This helps reducing the burden of reviewers and improves the work efficiency.