Public Bus Emergency Detection And MFC System Implementation Based On Image Processing Algorithm

Echoing to the Internet Plus strategy,making use of advanced technologies and ideas such as mobile Internet,cloud computing,big data and car networking,the concept of Smart City has come into being.Riding the era express of Internet Plus and combined with the intelligent video surveillance technology,Smart Bus,as a crucial element of the bus priority strategy and Smart City,provides strong support to improve the quality of service and protects the public safety.Considering the fact that there are scarcely any abnormal behavior intelligent video surveillance of bus scenes in current market,this paper proposes an algorithm of intelligent video surveillance in bus scenes.Through the technologies of computer vision,pattern recognition,video analysis and digital image processing and combined with basic image processing algorithms and machine learning algorithms,this algorithm provides the basis for the realization of the bus emergency detection.The main research and implemented functions in this paper are listed as follows:(1)In the aspect of basic image algorithm,original images are preprocessed before establishing regions of interest.Afterwards,moving targets are detected through an improved ViBe algorithm.Then we extract speed information of moving targets by Shi-Tomas corner detection and pyramid Lucas-Kanade optical flow method with correction coefficients,setting thresholds and determining whether an emergency occurs or not.Through theoretical analysis and video experiments,the detection accuracy of this method can reach more than 86.4% under different illumination conditions.(2)In the aspect of machine learning algorithm,to locate detection targets more precisely and make it easier for improved CNN algorithm to operate,the multi-scale sliding window algorithm is proposed on the basis of the detection results of the improved ViBe algorithm.Then we operate improved CNN algorithm by combining the motion information of 10 frames forward and afterward,integrate the information of full connection layers of moving targets,carry out pattern recognition to determine whether the emergency occurs.The experimental results show that the detection accuracy of the improved CNN algorithm is 93.5%.Comparing with the original CNN algorithm,the accuracy has been increased by 9.6% and the false detection rate is reduced by 3.1%.(3)Based on basic image algorithm and machine learning algorithm,a bus emergency detection method is presented in this paper.The experimental result shows that the detection accuracy rate is 96.8% and the algorithm takes 294ms/frame.In addition,this paper gives the realization of MFC system based on bus emergency detection,the system requirements analysis,environmental configuration requirements and the overall framework of the system.This paper also analyzes in detail the software implementation methods of four main system modules,which are video reading module,video preprocessing module,motion detection module and emergency identification module.Moreover,the system interface design and function description are also presented.