Research On Image Processing Of Accident Scene Based On Multiple And Unfixed References

In recent years，with the development of the social economy，the total numberof automobiles in China were increased rapidly. The road traffic safety is faced withserious hidden dangers, which not only had a serious influence on the developmentof social economy, poses a serious life and property losses, but also affect thestability of the society and the construction of harmonious society. Because of itsblindness, time consuming, and many other problems of itself, the traditional way oftraffic accident scene investigation result in more huge resources loss for human,material and financial. Therefore, a rapid, accurate, scientific research of trafficaccident scene investigation method is more and more important and urgent.On the basis of the research and application status’s analysis of the currenttraffic accidents for site survey and disposal technology at home and abroad, themethod of image correction was proposed based on unfixed references, that is to saythat it can take several scene photos with the references objects from differentviews(about four position relative to an angle of90degrees). Then on the basis ofthe scheme, firstly introduce the necessary photographic measurement imagepreprocessing techniques and methods of image edge detection technology, throughthe contrast of the processing target image and then get the method-the improvedCanny edge detection algorithm. Use traffic accident image correction software forimage correction of target images on the scene of the accident, and measure thedistance of the point spacing, finally compare with the actual measurement point andproved its feasibility and accuracy, provide the basis for drawing traffic accidentscene scale map and other works. The experiment results demonstrate that theproposed method is not only accurate and simple, but also flexible.At the actual scene of road traffic accident, only need to take a picture thatcontain the specific references and take the mentioned operations like preprocessing,edge detection and image correction and then measure the spacing and size of theelements of the accident scene directly. It reduced the time of the accident handlingand traffic dispersion, and achieved the purpose of handling the traffic accidentsrapidly, accurately, and conveniently.The main contents are as followings: In order to make up for the inadequacy of the image acquisition method of roadtraffic accident scene, selected target image preprocessing method, by taking someoperation such as smooth and denoise, threshold segmentation and the processing ofsharpening by C++, it keeps the original information of the target images, enhancingthe definition of the target images, and preparing for the edge detection andcollecting the features for the further part in the dissertation.After comparing with several common image edge extraction algorithm, thisthesis puts forward the new method of edge detection based on the Cannyalgorithms, do the necessary preparation conditions for edge feature point extractionof the images lately.Firstly summarized the camera working principle and imaging geometric model,Proposed a method that using multiple and nonstandard objects for image geometriccorrection on accident scene. This method is that by putting multiple and unfixedreferences object, firstly do processing the images of the traffic accident scene, thenget the edges of the references’ feature points and lastly using the image point tomeasure the distance, for the subsequent processing of the accident. The selectedreferences object has the characteristics of uncertainty, be not restricted ofenvironment, be more practical for some accidents which happened on non-fixedreference sections, solved the time delay of field measurement, made the survey ofthe scene of the accident more convenient, flexible and accurate.Use the preprocessing and edge extraction technology to processing the imagefeature points, identify each image feature points, then correct the images, lastlymeasured distance of image point, comparing with the actual measurement point,then test error. Analyzed the factors based on the unfixed references imagecorrection method, through the comparison of multiple sets of test data, proved itsfeasibility and accuracy.