Development of Real-time Digit Recognition System Based on FPGA

Abstract

With the rapid development of intelligent devices, the demand for automatic recognition of digital information as license plate recognition, instrument reading recognition and industrial coding recognition is growing increasingly. Digit recognition relies on image processing and pattern recognition technologies, but traditional software-driven methods and early FPGA-based systems face critical limitations in embedded scenarios. Early feature matching and geometric feature-based recognition methods verified accuracy but ignored embedded resource constraints, while existing FPGA applications suffered from cumbersome algorithms leading to insufficient real-time performance. Additionally, traditional PC-based algorithms fail to meet embedded requirements due to excessive computational complexity and power consumption, whereas FPGA’s high parallelism and low power consumption make it ideal for data-intensive image processing. To address these gaps, this paper proposes a simplified and efficient FPGA-based real-time digit recognition system. Combining the OV5640 camera to complete image acquisition, and realizes real- time recognition of printed digits through steps such as color space conversion, binarization processing, projection segmentation, and digital feature matching. to resolve computational complexity and real-time issues, enabling stable printed digit recognition in embedded vision applications. Experimental results show that the method has a clear structure and good real-time performance, can stably realize the digit recognition function on the embedded hardware platform, and has certain engineering application value.