| This thesis describes an automatic reading system for general instruments based on image processing and ZigBee wireless communication which is a new solution of instrument automation. Comparing with installing a new intelligent instrument, the automatic reading system has the advantage of avoiding disassembly of existing instruments, needless to wire layout, low economy cost and reusable, which can bring a widely usage in public power networks, public gas systems, mine detection systems, hydrological monitoring systems and so on.A DSP chip TMS320DM643 is chosen as the main processor for the reading module, which is expanded by external memories, video decoder and other devices. Two ZigBee wireless communication modules is designed based on CC2430 which is a new knid of wireless micro control unit containing a radio sender and receiver. Reading module and wireless communication modules make up of the whole system.The software architecture is based on DSP/BIOS. Drivers for image capture and serial communication are designed firstly; followed by image pre-processing, rotate detection and correction, and then segmentation. Characters are located and cut out from the result of segmentation using rejection method. Binary character images are put into a BP neural network to recognize its value.To avoid distortion in image scaling advances a regional average method. In image filter process raises an average-product method which verified to have a good effect. To enhance adaptability to uneven lightness in image segmentation, a new union thresholding method based on gray-scale transformation is proposed. In character location, iterative rejection is applied to reduce its error.For BP neural network training and testing,1400 character images are sampled and reach a recognize rate of 99.7%. Finally an integrated test, including remote control and data transformation, is carried out which proved the reliability and stability of the system, and ultimately the feasibility of the design. |
PDF Full Text Request