Research On The Algorithms For Detection Of Vehicle Converse Based On Object Tracking

Detection and tracking of moving object is an important area in the domain of digital image processing,pattern recognition and computer vision.It is becoming a hotspot,and its applications is great important in the military,civil and many other fields.In this paper,we study several vehicle detecting and tracking methods and study filter technology of vehicle tracking,propose some novel algorithms aimed at the implementation of traffic incidents detection.Our work mainly focus on the following aspects:(1) Based on Simulated annealing algorithm the image segmentation of threshold selection.Because the video sequence have great relevance, and threshold of adjacent frames is very close,the traditional OTSU method ergodic all gray value and obtain threshold,it is clumsy.In this paper,according to Simulated annealing theory,we save an threshold of the previous frame and let it be initial value of Simulated annealing algorithm,then,intelligent search of the current frame.Finally,we can obtain a satisfied threshold.The experiments show that our algorithms shorten the running time and obtain a good result.(2) Detection of the moving vehicles.We use block-divided methods to initialize the background first and then update it dynamically.By the background subtraction and connectivity judgment method,the vehicles can be located.The experiments show that our algorithms' performances are satisfactory.(3) Based on Kalman filter tracking of forecast region vehicles and detection of converse.After understanding filter technology,studing Kalman filter and Particle filter and analyzing both similarities and differences,we estimate vehicle position with kalman filter,reduce detection region,combine vehicle characteristic information and accomplish vehicle tracking.At last,this paper realize the detection of converse by using tracking information.The experiment results show that our algorithms work effectively and stably in the vehicles tracking and the detection results are accurate.