Research On The Parameters Calibration Methods Of Omnidirectional Vision System Based On Fish-eye Lens

In recent years, with the optic lens and image sensor technologydevelopment,conventional vision systems which are based on the normal lenses can’t satisfythe requirements of many applications because of their limited Field of View(FOV).Omnidirectional Vision system can acquire 360 degrees horizontal Field of View(h-FOV) image information with one time shooting. Because of the advance ofthe large FOV,it became to the research focus and hot topics in computer vision field. Thanks to the effortsfrom worldwide scholars, there many study achievements are published. Omnidirectionalvision system has been widely applied in safety monitoring, mobile robot automaticnavigation, entertainment, driving assistance and clinic medicineetc.Unlike the traditional omnidirectional vision, in this paper, fish-eye lens has been used tobuild the vision system. Although, Omnidirectional Vision Sensor (ODVS) can acquire 360degrees h-FOV image information, the vertical Field of View (v-FOV) of all kinds of ODVSis limited, and there exists a central blind area or some useless information such as camera orcamera stand in the acquired image. Fish-eye lens maps all information of the hemispherespace in the shape of a concentric circle. Compare with ODVS, there is no blind area, and thev-FOV can reach 180 degrees. Therefore, using fish-eye lens to build omnidirectional visionsystem has been becoming more and more received attention and approbated.By studying and reviewing some classical calibration methods for the conventionalvision systems, a special calibration method of fish-eye omnidirectional vision system hasbeen proposed firstly in this paper. This method is based on the idea of the physicalparameters separation. The advantages and deficiencies of this method have been analyzedthrough experiments. After that, a novel digital parameters calibration method is proposed,which is based on a general mathematic model. This model can satisfy all kinds of fish-eyelenses. Finally, the corresponding calibration software has been programed.The main works of this dissertation are summarized as follows:1. Analyze the imaging principles and imaging modelsofconventional vision system.Compare different classic calibration methods and find out the advantages and deficiencies ofthem. Establish the theoretical foundation to the fish-eye omnidirectional vision systemcalibration research.2. Summarize the imaging principles and imaging modelsfor fish-eye lens.Build thegeometry, mathematics and distortion model for it. Select lens and camera to establish anomnidirectional vision system for this research.3. Build a geometry model for fish-eye omnidirectional vision system. Research thephysical parameters separation calibration method based on this geometry model. Verify themethod by experiments. 4. Create a new mathematics model for fish-eye lens. This model can suit all kinds offish-eye lenses. Research a novel digital parameters calibration methodbased on it.Thecalibration software has been programed by MATLAB. The experiment results improved that,the method is correct and applicable.