Research On Motion Parameters Measurement Based On Video Image

The motion parameters measurement method based on video image theory, whichdirectly simulates the way of human eyes processing scenes, computesthree-dimensional information of space object from images captured by camera, andthus measures its motion parameters. It has a wide range of applications in the field ofmodern industry. In this paper, camera calibration, stereo matching, feature extractionand motion parameters measurement techniques and algorithms are deeply researched.Finally, the subject validates the precision and accuracy of this method through theexperimental results.More specifically, the work and innovations of the subject are expressed asfollows.(1)Camera imaging model and calibration parameters are researched and analyzed.Then, comprehensive calibration method, which based on Zhang Zhengyou’s principle,and integrated the Matlab calibration toolbox with OpenCV, is proposed. Experimentsprove that the method has the higher accuracy, and its precision is directly influenced bythe number of collecting sample and the sample’s quality.(2)Combined with the Differential Morphological Decomposition theory ofmulti-scale extraction algorithm, which using Harris corner feature as the target’sgeometric feature, is analyzed. Differential Morphological Decomposition (DMD) asthe nonlinear scale space decomposition, not only preserves edges in higher scales, butalso makes it easy to choose the Harris operator’s integral scale. It overcomes thedeficiencies of single-scale Harris corner detector, such as poor detection accuracy,noise interference, difficult selection of the integral scale. The experiments show thatthis method improves the detection accuracy, preserves the image edge well, and has a low rate of the false detection, which especially has a great advantage on corner regionsand overlap regions.(3)Segment-based stereo matching method is analyzed, which effectively solve theproblem of fuzzy boundary and low texture regions prone to mismatching. The first stepof the algorithm employs the mean-shift algorithm to segment the reference image.Then, it’s followed by the use of Adaptive Support Weighted Self-Adaptationdissimilarity algorithm (ASW-SelfAd) for the estimation of initial disparity. The thirdstep is the application of Singular Value Decomposition (SVD) in solving the robustdisparity plane fitting. Lastly, we apply improved clustering algorithm to merge theneighboring segments, and refine the disparity map by the new energy function.Matching experiments show that this method can not only improve the matching speedand accuracy, but also be an exact match in textureless regions, disparity discontinuousboundaries and occluded portions.The hardware and software architecture of the motion parameters measurementsystem based on video image are constructed by above theories and techniques. All thealgorithms proposed in this paper are programmed by VC++6.0and MATLAB. Cameracalibration, image matching,3D position, speed, acceleration measurement and othersystem functions are realized.