Study On Omnidirectional Vision And Panoramic Stereo Sphere Vision Based On Fish-eye Lens

Omnidirectional Vision system can acquire360°degrees horizontal Field of View (hFOV) image in formation with one times hooting. However, because the Omnidirectional Vision Sensor is based on the light reflection principle, the vertical Field of View (vFOV) of Omnidirectional Vision System is limited. In present, it is widely used to control the mobile robot automationa vigation. Panoramic Vision System is able to "see" the whole scenery of the observation space without using the machinery rotating and scanning. But, they project the surroundings cenery information to the two-dimensional images. In this way, one-dimensional information of object, depth, has been missed. Stereo Vision System acquires the depth information bye aptured the corresponding points from different images. Most of theses tudies used or dinaryl enses to build stereo vision system so the overlapped region of these cameras is very limited. In this paper, Omnidirectional Vision and Panoramic Stereo Sphere Vision have been studied to explor new measurement method for Ominidirectional Vision and new scheme for Machine Vision. Major work of this thesis can be included in four parts as follows:1. The basic theory research for fish-eye lens vision system. In this paper, we started the basic theory research of fish-eye lens vision system firstly, including how to apply fish-eye lens to build Omnidirectional Vision System and to get360°×180°field of view, build geometric model for fish-eye lens Omnidirectional Vision System, research the calibration method for the vision system, study the distortion correction algorithm for fish-eye lens images.2. The key technologies study for fish-eye lens vision system. Based on theoretical analysis, some of key technologies such as targets recognition, targets tracking and locating which based on fish-eye lens images have been researched. After that, these theories research achievements have been applied to mobiler obot automatic navigation. The technologies achieved to control different kind of Automatic Guided Vehicle (AGV) platform automatic navigation in different indoor and out door environment.3. Panoramic Vision and Stereo Vision explor. Through the further development of the theories of Panoramic Vision and Stereo Vision, the objective laws of Panoramic Vision and Stereo Vision which can be combined efficiently to realize advantageous complementarities has been revealed in this paper. A new kind of Machine Vision System is proposed—“Panoramic Stereo Sphere Vision”. In theoretical research, buildthe theoretical system for Panoramic Stereo Sphere Vision, abstract a “Sphere Model”,establish the geometry model and mathematic model for this novel vision system,create a real-time simulation system for Panoramic Stereo Sphere Vision and calibratethe internal and external parameters in the stereo calibration space.4. Hardware o f t he embedded omnidirectional vision system a nd th e ima geprocessor of t he P anoramic V ision a nd S tereo V ision design. An e mbeddedomnidirectional vision system based on e mbedded technologies has been designed.The algorithms are transplanted to the embedded platform. In this way, the design ideaof min iaturization, mo dularity a nd in tegration a re r ealized. O n th at b asis, th ehardware structure of Panoramic Stereo Sphere Vision has been proposed, which isbased on f our vision sensors+FPGA+DSP. It is a new vision processor whichintegrates functions with image acquisition, analysis and processing.