機器視覺在工業自動化生產中應用相當廣泛，檢測環境往往會影響取像品質，導致後續影像處理和辨識效能的降低，進而影響自動化控制系統的整體性能。本研究研發一個高速視覺檢測系統的影像前處理硬體加速器，用以改善機器視覺系統的取像品質和效能。我們設計了自適應影像增強法ADPHE與自適應Wellner二值化硬體加速器。透過兩種影像前處理降低環境對檢測效率的影響，提升了影像品質，並達到高速的處理效能。為了滿足不同影像運算遮罩的需求，我們設計一個可動態調整遮罩大小的彈性化硬體架構，可藉由指令暫存器設定即可得到新的影像前處理硬體加速器，而不需更改電路。最後，我們以FPGA驗證影像前處理硬體加速器效能，本系統在127.81Mhz的時脈下，每秒能處理61張解析度為 的影像，可滿足即時、高取像品質、彈性化的嵌入式視覺檢測應用。;The machine vision is used in industrial automation production extensively, the inspection environment sometimes influences the imaging quality, so that the subsequent image processing and recognition performance are degraded, and then the overall performance of automatic control system is influenced. This study develops an image preprocessing hardware accelerator for high speed vision inspection systems to improve the imaging quality and performance of machine vision system. We design adaptive image enhancement ADPSHE and adaptive Wellner binarization hardware accelerator. The effect of environment on the inspection efficiency is reduced by two kinds of image preprocessing, the image quality is upgraded, and high speed processing performance is implemented. In order to meet the requirement for different image computing masks, an elastic hardware architecture which can adjust the mask size dynamically is designed, a new image preprocessing hardware accelerator can be obtained by command register setting without changing the circuit. Finally, the image preprocessing hardware accelerator performance is validated by FPGA. On the frequency of 127.81Mhz, this system can process 61 frames with images per second, applicable to real-time, high imaging quality and elastic embedded visual inspection.
