| Visual technology, as an important method in the fields of target positioning, has been widely applied to the industrial measurement, surveillance and control system, navigation, positioning and tracking for military target, and so on. Because compound eye in nature is endowed with especial surface construction, imaging system with multi-channel and high centralized process unit for information with neural network, and it is capacity of target positioning with large field of view and high sensitivity to capture moving target and deal with it quickly, it is very fit for the requirement of light weight, small size, low power consumption and high reliability of positioning equipment. Based on these characteristics of compound eye, a novel compound eye system for target positioning with large field of view is put forward in this paper, and the adopted technology to capture the entire target within large field of view and the solved method to project lenses array within large field of view on limited imaging area of one CMOS are discussed. After installation requirements of lenses array is introduced, lenses installation platform is set up and installation task is completed. Based on the survey of positioning technology with binocular vision, the mathematical model fitting to the compound eye is set up. Next, calibration task for imaging distortion of multi-lenses of compound eye and the image processing method to recognize imaging point of target are studied and summarized in the ways of theory and experiment, and the algorithm to extract the center of every imaging point is analyzed. Subsequently, based on the unit partition for compound eye, a new algorithm to match every imaging point and its corresponding lens is put forward, and thus to complete further task of target positioning. Finally, according to the target positioning algorithm, calibrated performance of compound eye imaging system is evaluated and simply applications are initially studied.The main research contents of this thesis are listed as follows:1. A novel compound eye device for target positioning with large field of view is designed. The device makes use of the lenses array with surface distribution to obtain targets within large field of view. The arrange type which could not only meet high fill ratio but also eliminate imaging blind region by the greatest extent is studied. Installation scheme for lenses array is initially designed and the task of installing lenses array is completed. After the question projecting lenses array with large field of view on the limited imaging plane of one CMOS and the question of serious defocus and inclination of received lenses locating at the edge of compound eye are studied, one solution is put forward that refractive lens is added to the location between lenses array and CMOS plane. For the selected CMOS image sense, its driver is developed, and pixel data is collected by FPGA. Then the data is sent to PC by USB2.0. Finally, the received images with compound eye are displayed and the imaging information is further copied with.2. After the current technology for target positioning with binocular vision is studied and summarized, sub-coordinate system of compound eye and its world coordinate system are set up, then the mathematical model and calibrating content for target positioning with compound eye are established. In order to help express complex relationship of incident ray and solve the question setting up mathematical model, the ray, form target point to received lens and to refractive lens and to CMOS plane, is split into two parts. One is the linear part from target point to received sub-lens of compound eye, and the other is the nonlinear part from received sub-lens to refractive lens and to CMOS plane. Based on the existing calibrating model, after those methods are analyzed, the calibrating method which establishes the corresponding relationship between imaging points and its corresponding incident angels of every lens of compound eye is put forward.3. According to the calibrating theory, calibrating scheme is designed, and the calibrating goal is completed. The experimental equipments and laboratory instruments of plasma television, plane mirror interferometer, semiconductor laser, level instrument and horizontal moving guide rail are selected. The plasma television is selected as the target plane. The parallelism between planes of compound eye and of target is adjusted by interferometer. The verticality between horizontal moving axle and target plane is regulated by the superposition of laser dot and the fixed point in the target plane in the moving process of target plane. The uniformities between X-axis of coordination system of compound eye and row on the target, and between Y-axis of compound eye coordination system and column on the target are adjusted. Based on the work above, the intersection of optical axis of center lens of compound eye and target plane is found out, and the position coordinate on the target plane is calculated by the vanishing point. The distance between the center lens of compound eye and the intersection is obtained by coincidence imaging point of different points in the Y-axis. Then the three dimensional coordinates of all points on the target plane are known. After every center coordinate of imaging point and its corresponding angles of incident ray for every lens are obtained, the corresponding relationship between imaging points and its incident angles is set up. Finally, the calibrated task for imaging distortion of compound eye is completed. In the process of system calibration, image processing method to extract imaging point of target is introduced. And the algorithms to extract imaging point center and of imaging denoising are analyzed.4. The key issue of three dimension target positioning based on compound eye is studied, and initial application on target positioning with large field of view is discussed. The data interpolation methods based on Delaunay triangulation and neuron network and biharmonic spline are introduced, and interpolation accuracy and interpolation characteristic are analyzed. After the target imaging point center coordinate is extracted, it is interpolated into calibrating results of corresponding lens and the angles is obtained. This paper puts forward to divide lenses array of compound eye into169units of cluster eyes. For received some imaging points of one target, the imaging points that belong to a cluster eyes are found out firstly. Then the waving error is defined by normalized3D coordinates different. Finally, their corresponding lenses are matched successfully and the key issue to match received imaging point and its corresponding lens is solved. After target plane is moved to different position and the points on the target plane are calculated, the calibrated results of compound eye imaging system are evaluated. The results show that the mean square error of angle of every lens is about0.02°, and the relative error of X or Y or Z of three dimension coordinate of every point is about2%. In the process of target positioning, the influence of distribution relationship of interpolation point and referential points on interpolation accuracy is analyzed. Then the interpolation point that is within quadrangle formed with four neighbor points among referential points is only selected to interpolate. Finally, by the designing experiment scheme and putting up experimental platform, the intersection points of different parts with different field of view are initially measured with the designed compound eye positioning system, and their three dimensional contours of are reconstructed. |
PDF Full Text Request