影像處理在這幾年因為工業4.0的崛起，而在圖像辨識上有了大幅度的進化以及需求性，包括自動駕駛，臉部辨識甚至是醫療方面，美術美工等等…機器人在這幾年進步飛速，使得機器人工業自動化進步更為快速，其中的圖像辨識更是不可或缺的一部分，本論文將樹梅派視覺鏡頭用來做圖像的辨識，將模糊分群法帶入其中，希望在未來可應用在機器人上，輔助其辨識物品，進而做出捕抓等應用。
圖像辨識在機器人的應用是重要的一部分，本研究提出了將模糊理論與圖像識別相結合的方法，並經由模擬與實作驗證其有效性，論文中利用樹梅派微控制器(Raspberry Pi Microcontroller)搭配OpenCV的視覺辨識函數對捕獲的待抓物進行圖像前置處理，座標及不變矩計算，最後做種類辨識。所使用的方法是由影像不變矩與模糊分群法整合而成，將圖像的不變矩值當作輸入，匯入模糊分群法後，建立各分群之中心座標，再將這些中心座標值植入樹梅派中，對機器人捕獲之圖像進行圖像辨識。
為驗證其有效性，研究過程將此辨識方法先以MATLAB模擬，然後使用實驗室的履帶夾爪機器人，樹梅派微控制器及攝像鏡頭等設備，將圖像辨識功能實作在搜尋目標之應用，從實驗數據來看，所提之以模糊分群法為基礎之之樹梅派圖像識別應用架構確有其效果。
Currently, because of the rise of Industry 4.0, image processing has undergone significant evolution in image recognition and demand, including autopilot, face recognition medical treatment as well as art. The rapid progress in robotics makes the automation of the industry evolve faster, and image recognition is an indispensable part of it. In this paper, the Raspberry visual lens is used to identify images with the fuzzy clustering method. In the future, it can be applied to robots to identify object and grasp it.
The application of image recognition in robotics is an important part. This study proposes a method that combines fuzzy theory with image recognition, and validates its effectiveness through simulations and experiments. The architecture includes the Raspberry Pi in which. OpenCV's visual recognition functions are used to image pre-processing, coordinates and moment of its the captured object, and finally identify its type. The used method is the integration of the image invariant moments and the fuzzy clustering method. The invariant moment values of the images are taken as input. After the gathered data is imported, the center coordinates of each group are established, and then these center coordinate values are used.in Raspberry Pi.
In order to verify its effectiveness, the proposed method was simulated by MATLAB first. Then the lab’s equipment including the gripper robot, Raspberry Pi microcontroller and camera lens are adopted to implement to search for the object. From the experimental results, the proposed architecture of fuzzy clustering based image recognition has good performance of recognition and quicker identification time.
