DSP-based vision sensor applied to welding robots

Design and implementation of a DSP-based vision sensor are presented. The vision sensor uses high-speed DSP micro-processor to capture images, and carries out image processing algorithms locally. After making a decision, it directly transfers the result to the control system to execute the actions. Therefore, the need for high-bandwidth communication channels for transferring huge image data is avoided, and the performance of the system is then enhanced. Using embedded micro-processor and socket-based configuration, the developed vision sensor is characterized by compact size, great real-time performance and flexible expansibility. Software libraries for image processing are also supported for second-level development. The developed vision sensor is applied to the welding robot platform, and the experiment validates its practicability in real-world applications. Future work about using the vision sensor on robots is discussed at the end of this paper.