Study On The Image Reading System Based On DSP

Traditional ocular optical instruments, which have occupied an important position in measurement field, are characterized by their high accuracy, strong anti-interference capacity and long-term stability. However, there are many limitations in their data-reading system, including tedious reading procedures, time-consuming reading method, possible man-made errors and low automation degree. Due to the requirement of high precision, high efficiency and high intelligence in modern measuring apparatus, it is necessary to accomplish digitalization to the old-pattern.UTM (universal tools microscope), whose reading system is composed of reticles, staffs and spiral plates, is a typical and representative optical measuring instrument. This paper introduces a combinative solution of image measurement technology and digital signal process technology. Based on thorough research of the reading system structure and working principle of UTM, an image-type automatic reading system based on DSP is developed, which could achieve automatic reading and result displaying without changing the original instrument structure. The advantages of this system consist of miniaturization, intelligence, low power consuming, low cost and high processing speed. Meanwhile, the system is highly general and flexible, so that it is suitable for the upgrading to different types of traditional ocular optical instruments.The hardware system platform is established based on TMS320VC5410A DSP. In the platform, a single-chip analog CMOS, which is small in size, light in weight and low in power consuming, was adopted for image acquisition, so that the requirement of system portability can be realized. As the system logical control center, CPLD is mainly used for data buffering and address decoding in order to achieve real-time image storing. USB 2.0 data transfer protocol is adopted in communication module, which is able to accomplish high-speed data interchange between the system and the PC computer. Experiments results demonstrate that this hardware platform is capable of the acquisition, transferring, storing and processing of large capacity video data.In accordance with the features of spiral reading lens view images, an automatic reading algorithm for UTM is designed after deep analysis of related digital image processing technology, especially character recognition and sub-pixel locating methods. Based on gray scale center-of-mass algorithm, bilinear interpolation technology is inducted to improve the locating accuracy of millimeter reticles, whose precision is better than 0.3 pixels. Template matching method and BP neural network algorithm based on Tchebichef invariant moment are applied in digit character recognition, which has a high identification effect.The results of experiment reveal that the measurement accuracy of the current system can achieve±3μm, while it may reach or surpass ocular reading precision if the reticle is targeted with drum wheel firstly.