The Design And Realization Of Robot Vision System Based On DSP

The first problem need to be solved by the intelligent robot when in practice is the perpception to the external environment. Only after this could it complete the specific tasks. The visual sensor can get the most significant environmental information, so highly efficient and stable mobile robot vision system is an indispensable part of the robot. For the robot vision system, study is carried out about the image processing algorithms and the software and hardware designs, and a certain achievements are acquired.Researches have been done to the visiual image processing algorithms, including image pre-processing, image segmentation, moving target detection, etc. Comparing the advantages and disadvantages of the various methods through simulations, the better ways are chosen and realized by program. The pin-hole camera model is used for target positioning, which avoids a large number of matrix operations, and increases computing speed. For the deficiencies of the traditional threshold segmentation, the fuzzy neural network is applied on image segmentation, and a new method with variable threshold based on fuzzy neural network group is proposed for the color image segmentation. The method effectively overcomes the problem of great impact of the external changing light on color image segmentation, and has a self-adjustment function for the color thresholds under different light environments, which increases the accuracy of image segmentation. The sub-networks are trained respectively by rapid compensation learning algorithm. It further improves the convergence rate of the whole network on the basis of general neural network, and shortens the calculation time. The validity of the algorithm is verified through the simulation experiments.The software and hardware of the robot vision system is designed. The hardware adopts a high-performance embedded DSP as the core processor, which could not only complete a large amount of data computation, but also greatly enhance the ease of the vision system. Various hardware modules of the system are designed, including video codec and decode modules, memory modules, power module, communication module, and so on. The software design of the vision system is completed. Program loading and software initialization is arrived using DSP/BIOS operating system under the CCS development environment.Ultimately, the robot vision system is studied. A series of image processing algorithms are realized by programming, including image enhancement, image filtering, image segmentation and moving target detection. What's more, two experiments are carried out, one of which is target location and tracking, and another is speed detection of moving target. Both of them are practical and extensible. The validity is shown through the experiments.