Research On Computational Method Of Traffic Accident Vehicle Speed Based On Surveillance Video Image

With the rapid development of automobile industry,vehicle holdings has also increased,which leads to road traffic accidents occurring frequently and brings a great threat to the stability of society as well as the people's lives and properties safety.On the premise of preventing the accident,it is necessary to do further research for the accuracy and efficiency of traffic accident authenticate result.With the widely application of the monitors in the traffic,much valuable information about road traffic is preserved permanently,so using video image completing accident authenticate work has become an important option for the staff.The space distance and image distance cannot correspond precisely could be caused by photographic equipment,characteristics of imaging system and other factors,.Thus,the direct linear transformation,namely camera calibration process is necessary.In addition,in terms of the vehicle detection and tracking,the results are not ideal due to shadow inference,light change and other factors.In a word,common speed computational method based on monitor's video image is unable to provide accurate results for accident appraisal work because it could not eliminate defects and only as an auxiliary evidence.Regard the speed in the traffic accident as the research object.Utilizes the principles of photography and kinematics,establishes the mathematical model of speed calculation using interpolation method without the the camera calibration process,develops a visual system platform based on Matlab-GUI environment,which enables to calculate the speed quickly and reliably.Using velocity-time curve to acquire the vehicle motion equation,which determine the instantaneous motion(include route,position and velocity)before and during the accident.Use interpolation method to calculate the distance and the time.Described the perspective characteristics in the video image and designed the corresponding experiment.Test results show that in the driving distance of one frame time,it can ignore the effect of the perspective.Described the principle of video speed measuring,applied the interpolation method to the establishment of mathematical model,it can carry the point of make the calculation result meet the requirements even if not carry on camera calibration.It studies three kinds of vehicle speed computational models based on interpolation,and analyzes the application of each model.Designed and developed the traffic accident speed computational system platform based on surveillance video,the development platform select Matlab.According to the function of the platform,developed the system general design,module design and operation process.The applicability of the traffic accident speed calculation was calibrated based on case study.Analyzed three factors includes vehicle speed,scale length and camera angle to gain the precision influence of the traffic accident system,suggestions were put forward to eliminate the error in order to reflect the actual circumstance of the vehicle objectively and fairly.