Abstract: To address the real-time vision requirements of robot control, an embedded robot vision system based on the TMS320C6416 image processing platform was designed. This system uses VC++ 6.0 as the platform to develop the DSP control interface. This interface controls the DSP to acquire, process, and recognize images of rectangular objects. The DSP then transmits the recognition results to the host computer, providing target object information for the robot arm's trajectory planning. The hardware design scheme and software implementation process of the system are presented in this paper. Keywords: Embedded system; Robot vision; PCI interface; DSP; Image acquisition and processing; Pattern recognition. For machine vision systems, a large amount of work involves image processing and pattern recognition. Many technologies considered to date require digitization before image processing. Although the images are not stored, the processing consumes a significant amount of computation time, indicating a long time interval between image acquisition and obtaining the processing results. In some cases, machine vision systems need to process images in real-time to obtain results in real-time or within a sufficiently short time, making real-time operation essential. There are generally two methods to achieve real-time processing in machine vision: one is to design dedicated hardware circuits to achieve sufficiently fast processing speeds to meet real-time requirements; the other is to optimize both hardware and software to reduce processing and computation time, bringing it close to real-time speed. This paper focuses on the real-time design of a machine vision system. Utilizing the message communication mechanism between the TMS320C6416 DSP and a PC, image acquisition and processing algorithms are embedded in the DSP's Flash memory. The PC's control interface uses the PCI interface to control the DSP to execute the application program in the Flash memory via message commands. Because of the DSP's powerful computing capabilities, image acquisition, processing, and recognition are completed within the DSP, thus improving the real-time performance of machine vision and meeting the real-time requirements of robot control for its vision system. Although the PC controls the DSP, a large amount of time-consuming image processing and recognition algorithms are implemented entirely within the DSP; only minor information transfers (such as control commands and DSP image recognition results) occur between the PC and the DSP. (Source: Research on Rectangular Object Recognition Based on DSP Machine Vision)