The Technology Of Feature Points Extraction And Matching Of Photogrammetry In Traffic Accident Scene

In recent years,along with the quantity of mobile car retention keeping onincrement,the number of rode traffic accidents keeps on increment .It is more andmore seriously that handling traffic accident to cause on the scene of thetransportation jam and two trouble and so on hazards. Along with the exaltation ofpeople legal system consciousness and the change of the transportation principle,put forward the request for intelligentizing means of handling traffic accidents.Shorten the continuously time of handling traffic accident on the scene becomes theimportant target of the transportation managements. The photogrammetry is a validmeans to carry out intelligentizing to handling traffic accident scene. The featurepoints extracting and matching are the key techniques to Photogrammetry. Thefeature points extracting and matching become a little bit hot and crux in scienceresearch.Based on the sub-item" Research on Traffic Accident Scene Fast InvestigationTechnique "of the national natural science fund keystone item" Research on KeyTheory and Simulation technique of dynamic Traffic Management and Control inCity Road Network " , the paper aim at the 2008 the Peking Olympic gameprimarily, expect to study a kind of fast and efficient new technique of handlingtraffic accident scene——Photogrammetry. The feature points extraction andmatching is the key technique to carry out Photogrammetry . Firstly this paperbuilds up an relation of between the photograph picture and the space object ,andpoint out the feature points extracting and matching is a technical importantposition in this process. And study a few exaltation feature points extraction andmatching result to the image pre-processing technique. Then pivotly study thefeature points extraction and matching related theories and techniques. The mainresearch contents of this thesis as follows:(1) Introduce the road traffic accident present conditions and the troubleprocessing present conditions. Compare the domestic and international currenttraffic accident processing method, and point out the main spot of handicraft andmeans of making use of the electronics diagraph equipments, and also point out thephotogrammetry method is fast and usefulness in the scene. Explicit the fearurepoints extaction and matching to photogrammetry is important in the trafficaccident scene, and introduce domestic and international research presentconditions of points extraction and matching.(2)Introduce camera models ,camera calibration ,3D reconstruction etc.basic principles of photogrammetry. Summarize and analyzes the research progressof camera calibration and 3D reconstruction .The technique of camera calibrationfrom setting reference object develops to self-calibration that makes use of spacerelation to carry on, 3D reconstruction technique from a calibrated picture developsto make use of the foundation matrix to carry on. Through the theories analysis,point out the technique of fearure point extaction and matching not only is a key toinfluence the accuracy, also is the important technique support that pushesphotogrammetry theories to develop.(3)Study the technique of image pre-processing. After getting the spot picturein the traffic accident, it needs to take of the interference of the factors in thephotograph, such as noise and sunlight...etc. It wants to carry on the means ofimage pre-processing, improving the quality of the photograph, raising the accuratedegree that feature points extraction and matching .Mainly study methods of imageenlargement, image enhancement through equilibrium histogram, imagesmoothness through area averaging or median filtering work those well prepare towork with feature point extraction and matching.(4) Study the technique of feature points extraction. Introduce the definitionof feature points , and point out the substance of feature points extraction is pickingup the points token object characteristic in the picture. Study the traditionalmethods of feature points extraction Moravec and Forstner, analyzing theircharacteristics, indicating that they are not suitable for to feature points extractionin complex picture. Study the classic method-Harris, and introduce the pictureprocessing knowledge involved scale ,transformation etc. Point out Harris has theadvantage that principle intuitionistic ,computing in brief ,effective extraction,anti-revolve etc. and also demonstrate that Harris do not have the anti-scale &anti-affine functions. Combine to photogrammetry in camera calibration accordingto its characteristics and an actual circumstance, Study the SIFT method which isforeland in the feature point extraction theories. The SIFT method is according tothe thought of the choice of scale. Also be say that the point with correspond of thefeature scale. In order to examine the same point under the different scales, theSIFT method establishments the space of many scales, Attain the scale of thefeature points, in the meantime ,when making sure the feature points .It is thepurpose that anti-scale.(5)Study the feature points matching technique.The feature points matchingclosely relates to the feature points extraction. The substance of match is that assureextracting the same point between two pictures .In matching principle introduce theepipolar constraint and match thought etc. related theories. Introduce and analyzethe two match methods based on intensity and features. Study SSD,NCC two kindsof points match which are normal regulation method. Pivotly study the SIFTmethod. Description sign for points is key to the SIFT method, which is equal tothe small area around the point differentiation with others. Being concretelyordering of match up adopt an European distance method. Study the evaluation forfeature points matching .Summarize and introduce the parameter evaluationmethod and the relatively evaluate method.(6) Analyze and compare the description feature points matching result ofseveral kind basis. Adopt to pass similarity transformation matrix to describe thequality of matching methods. Draw up an experiment project, and draw upprocedure through MATLAB. Verify the usefulness of the SIFT way.Finally, the dissertation summarizes the findings and achievements, and bringsforward the issues for further research.Photogrammetry in traffic accident is very useful and important to alleviatetraffic pressure and traffic jam. It not only in some certain represent the level oftraffic accident handling in our country, but also is a new need to the trafficaccident handling because of rapid economic development.