Vision system is very important for intelligent robots to sense surrounding environment, which influences the performance of robots directly. Although great progresses have been achieved, there are lots of problems needing to be resolved on robot vision. The real-time performance of algorithm is the main bottleneck against robot vision working in practice. Supported by 863 National High-tech Project and Innovation Foundation of IACAS, we have developed an embedded vision system of high performance and studied some algorithms of real-time image processing. The novel work and contribution of this thesis can be summarized as follows: 1.Based on the technologies of DSP and FPGA, an embedded vision system of high performance is developed, which has architecture of modularization and can provide various communication interface. 2.A pseudo omnidirectional vision system and a multi-scale focus strategy are presented. This vision system can provide omnidirectional view of surroundings with high quality. Using the multi-scale focus strategy further improves the performance of vision system. 3.A real-time color image classification method is presented. In order to calibrate the object’s color in different lighting conditions, a kind of statistic ellipsoidal model is constructed. In this method, a 3-D Color Look-up Table (CLUT) is built in which only 18 bits are needed to represent one kind of color, instead of conventional 24 bits. Moreover, this method is implemented on FPGA, which can highly reduce the CPU’s computation burden and remarkably improve the performance of vision system. 4.Based-on FPGA, a fast algorithm for labeling connected component in binary image is presented, in which a novel method of dealing with equivalent labels is used and implemented in FPGA by a ACAM(Address and Content- Addressable Memories) memory. Contrast to traditional method, this algorithm has much higher real-time performance.
修改评论