Study Of Key Technologies For Image Measurement In SpecialEnvironment

Measurement of tube static parameters is a prerequisite and important technique process in designing, studying and use of the special type of tube. The laggard measurement method based on optics glass and data processing artificially cannot satisfy new requirements that following progression of the development, manufacture level and use of the special type of tube. A high efficiency, precision and automatism measurement system need be constructed.Based on thorough study of the measuring principle, measuring method and performance and limitation of the special type of tube, a project to the tube static parameters measurement system with the core of crawler control, illumination, image pretreatment and image measurement at high accuracy has been designed and realized. These technologies are absolutely necessary for all the static parameters measurement, so key technologies for camber measurement are only studied thoroughly in this dissertation. The following is the main contents that studied in the dissertation.1. A kinetic model on the movement of the crawler control system and an even illumination model are proposedAccording to the operation demands of the crawler, a kinetic model of the crawler entrance motion is proposed. And the load, the carrier's moment of inertia and motor torque are studied thoroughly. At the same time, the factors effecting motion precision have also been studied.According to even illumination method, an even illumination model is proposed. In view of narrowness of interior space, small size of the measurement system, short object distance and large field of view of imaging optical system, the condition illuminating evenly and realization method are studied in the special environment.2. Many kinds of noise can be distinguished by the local gradient module of pixels, and the corresponding noise can be denoised with complex-valued log Gabor wavelet filterThe local gradient module of pixels is regarded as the threshold value, by which many kinds of noise can be distinguished. Then the corresponding method can be chosen to remove noise. The median filter is worked on pixels polluted by salt and pepper noise. With the use of complex-valued log Gabor wavelet, phase information and amplitude information of a pixel can be obtained in the partially denoised image. And an estimate of noise amplitude distribution for the smallest scale filter pair can be got. From the statistics of the smallest scale filter response, estimates of the noise amplitude distributions at all the other scales can be obtained, and shrinkage thresholds can be set automatically. Thus the noisy image can be denoised completely.3. In view of target images with nonuniform illumination, an adaptive threshold surface method based on variation and a fast image segmentation method based on maximum fuzzy entropy are proposedAn adaptive threshold surface method based on variation can segment target images accurately. Because of time-consuming iterative operations of partial differential equations, the adaptive threshold surface method based on variation cannot satisfy adequately the requirement of real-time image segmentation. Subsequently, a fast algorithm of image segmentation based on maximum fuzzy entropy is proposed to satisfy the requirement of real-time segmentation. Genetic algorithm and simulated annealing algorithm are adopted to not only generate better solutions but also accelerate the speed of convergence in the proposed fast algorithm.4. Nonlinear distortion correction methods based on Radial Basis Function neural network and linear characteristic are proposed to correct target images A nonlinear distortion correction method based on Radial Basis Function neural network can keep away from the difficult problem of singular solutions of equations in the polynomial warping method. However, the method requires the accurate determination of point correspondences between a scene and an image of that scene, or else corrected images are unsatisfying. Thus a nonlinear distortion correction method based on linear characteristic in the target pattern is proposed. The method does not require the determination of point correspondence between a scene and an image of that scene. The method simply requires images of known linear segments, which do not pass through the center of field of view and are long enough to be calculated in the distortion correction operation.5. A two-step edge detection method at high accuracy combined with LoG operator and Zernike moment-based operators is proposedEdges of the target pattern and the laser spot are reached at pixel accuracy by using LoG operator, then they are reached at sub-pixel accuracy by using Zernike moment-based operators in the extending concourse of edge pixels. The edge pixels coordinates at sub-pixel accuracy are used to fit spot circle and calculate the centroid of spot in least square estimation method.At the same time, error sources affecting measurement accuracy in the tube static parameters measurement system have been analyzed thoroughly, and the corresponding methods have been proposed to overcome the deficiencies. And we look forward to more researches on the system.