In recent years, with advances of technology, computer equipment, both in computing power or cost, has gradually broken through numerous limitations to bring about many new applications. For instance, national highway electronic toll collection systems and parking automatic toll collection systems, thanks to information technology, have reduced labor costs and improved quality of service. Nonetheless, despite availability of mature technology, we still rely on human labor in many aspects of life which falls short of curbing the rising trend of highly specialized professional vehicle theft. Given the trend, how to effectively exploit technology and the limited police force to increase the effectiveness of law enforcement is one of the imperative research topics. In the past, there have been many implementations on PC or researches on license plate recognition embedded system development board. However, the use of personal computers, albeit with an excellent performance, are financially and energetically at high costs whereas the use of conventional embedded board, despite lower energy consumption, is at times constrained by the efficiency of the hardware to yield satisfactory results.
Therefore, this thesis proposes employing the SoC FPGA embedded platforms to fabricate high performance real time license plate recognition system. We attempt at striking a balance amidst low cost, low power consumption and computing performance that may induce law enforcement officers to accept it. Compared to pure ARM embedded system, the ARM embedded system compounded with FPGA programmable logic array not only retains the advantages of low cost and low power consumption, but the collaborative computing of FPGA programmable logic array and ARM processor in accordance with the characteristics of different hardware avoids some calculations bottlenecks and improves overall system processing performance.
The implementation shows that the license plate recognition system proposed in this thesis effectively makes use of the characteristics of the hardware to reduce the computation complexity. Additionally, some computations of this paper even produce better results relative to the general mid-level PC. Last but not the least, it attains high recognition performance in the general environment on implementation, proving itself a viable license plate recognition system economically and functionally.