隨著機器視覺的進步,逐漸以機器取代人眼進行視覺檢測等應用,然而開發一嵌入式視覺應用過程複雜、耗時,且仰賴開發人員的技術能力,需耗費大量開發成本。本研究提出一個嵌入式攝影機的軟硬體協同開發平台,於微控制器中實作嵌入式影像處理中介軟體與Grafcet虛擬機器,於FPGA中實作取像與影像處理硬體加速器,兩者形成軟硬體協同處理器。最後結合PC端Grafcet圖形開發和驗證軟體,形成一個可開發嵌入式視覺應用的軟硬體協同開發平台。此開發平台中最為核心的模組為嵌入式影像處理中介軟體,其具有多種影像處理、分析方法,協助使用者開發視覺應用,且為Non-OS系統,使用的記憶體資源少、可移植性高。最後,以兩個嵌入式視覺應用開發案例,驗證本研究設計之嵌入式攝影機軟硬體開發平台,於開發視覺應用上具有快速開發且減少開發複雜度的優點,並且開發的視覺應用可獨立運作於嵌入式裝置中。;As machine vision advances, machines have gradually replaced human vision for visual inspection and other applications. However, developing an embedded visual application is complicated and time-consuming, and heavily dependent on professional capacity. This paper proposed a hardware/software collaborative development platform of embedded camera. This development platform includes hardware/software collaborative processor which is composed of microcontroller and FPGA, and PC software. This paper implements embedded image processing middleware and Grafcet virtual machine on microcontroller, and implements image capture and image processing hardware accelerator on FPGA. Then, we design a PC software which contains two function: development using Grafcet description language and verification of user design. In this paper, the core of development platform is embedded image processing middleware. This middleware includes many methods of image processing and image analysis to helps user develop vision applications. Moreover, the middleware is Non-OS system and thus it has the advantages of less memory usage and high portability. Finally, we verify this development platform through the two experiments of embedded vision application development. The result shows that the development platform features rapid development of vision applications and reduces complexity of development. Additionally, developing vision application can run independently on embedded devices.