Investigation On The Granularity And Sharp Of Particle By Computer Vision

Particles act as an existing form of material. The techniques related to particle have close relationship with many industrial processes and research departments, including energy, environment, chemical engineering, aerospace etc. The granularity and shape of the particle are the important parameters describing its characteristics, whose measurement techniques have significant influence on the application of the particles in the industrial process. At present, indirect measurement method for particles characteristics is mainly adopted. The measurement results about the granularity and shape of particle have great difference due to different measurement theory and measurement method. A measurement technique of particles granularity and shape based on computer stereo vision is put forward in the thesis. The size and shape information of particle can be obtained by reconstructed results of particle surface, which is a direct measurement method and can accurately get the size and geometry property of particle.How to reconstruct the 3D image of particle is the important constitute part of computer stereo vision based particle shape measurement system, which include the establishment of the camera projection model and the system model, the calibration of the camera model and the system model and the matching of the related points in the images.In this paper microscope imaging was researched in depth and results show the pinhole model can be used to describe the microscope imaging technique. In order to reduce the aberrance error caused by lens, a nonlinear distortion pinhole imaging model is put forward, which can better eliminate the influence of the camera lens and the microscope axis distortion; The imaging process of the microscope system is researched. According to the characteristics that microscope can strictly move around its axis, a coaxial stereo vision model fitting for microscope imaging system is established, through which the 3D coordinate of particle surface can be obtained and further the particle shape can be reconstructed. Based on the geometry characteristics of coaxial stereo vision mode, the equal height property is introduced and is utilized to partition the researched object conveniently. Additionally, before point matching, it can also partition the matching district to reduce calculating quantity.In the calibration process of CCD camera and measurement system, a novel method, which is based on equal height property of the coaxial method, is proposed to calibrate the nonlinear distortion of imaging system. The calibration method doesn't need to know the space point position and the outside matrix parameters of camera. The nonlinear distortion of system can be identified as long as camera move twice equal distance in the axial direction. Compared with standard template method, it has the advantage of higher precision. At last, according to the characteristics of microscope stereo vision and experimental purpose, microscope stereo vision system is calibrated step by step.In the image matching process, multiple steps matching method is adopted in the paper. At first, the interest points are matched and then the mismatching interest points can be removed using the connection relation of points and the characteristic vector of connected lines. At the same time, according to thematching results of interest points, other characteristic points can be matched using the connection relation of interest points and epipolar line restriction condition. The experiment results show the method has higher matching efficiency and accuracy ratio.On the basis of 3D reconstructed results of particle, octree method is proposed to describe the particle in the paper. The 3D configuration of particle can be denoted by chain code. Using the configuration information (circumscribe cube characteristic, curved face direction characteristic), the membership grade function of simple configuration (slice, pin) can be established and for the particle with complicated geometry (sphericity, cube), statistical method is used to construct membership grade function. Finally, a particle identification method based on fuzzy mathematics is introduced, which can classify the particles step by step into slice particles, pin particles, sphericity particles, cube particles and irregular figure particles. Different definitive diameter (volume diameter, surface area diameter, projection diameter) of the particle can be obtained. Further, the statistical analysis results of particle size distribution and shape information (pellet degree, special surface area) can be obtained. Research results show the method can extract a lot of characteristic information of particle and provides a reference for other particle measurement methods.In addition, a track method for movement state of the particles in 3D space is put forward using computer stereo vision technique combined with the camera field depth theory in the paper. The method not only can track the movement but also can realize the track of rotational turn. The solution of tract process can be largely simplified using the characteristics that the change of line segment vector of particle surface points is dependent on rotational turn and is irrespective of movement of particle. At the same time, Kalman filter method is used to forecast the project position of characteristic point of the particles, which improves the matching speed of characteristic point.Finally, the surface shape of coal heap in coal-fired station is similar to the surface configuration of particle except for scale. So computer vision is also applied to measure the storage quantity of the coal heap in this paper. The measurement system based on computer stereo vision was designed and the sources of surface reconstruction error are analyzed. The relation between measurement precision and measurement system layout is also analyzed, from which the system layout can be determined. The experimental results also verify the validity of the measurement method in the lab.