基於機器視覺與次像素邊緣偵測於LED探針之自動化檢測系統

Abstract

LED (light-emitting diode,發光二極體)從原先做為電子裝置的指示燈使用，如今已被廣泛的應用在工作或一般用途的照明上；而LED需要由LED探針做燈泡特性的量測，以篩選出LED的好壞，因此越來越多LED探針被大量地生產。本論文的研究目的就是提出一套自動的光學檢測系統，以提升LED探針的品質，提高生產效率，降低不良品的產出，並能夠即時回饋產品的資訊，而檢測的流程則使用許多機器視覺與影像處理的技術，包含運用Otsu門檻值搭配Canny的邊緣檢測得到初步的邊緣位置、次像素邊緣檢測取得更精確的邊緣位置、以及物件分群等方法，以提升量測的準確度，最後可以控制探針角度誤差在1%而半徑誤差在2%左右，並且提出一套能夠精準的區分出有瑕疵探針的方法，實驗證明本論文所提出的方法能夠快速且精準的分析LED探針的尺寸以及好壞。In the beginning, LEDs (light-emitting diode) were used as indicator lamps for electronic devices, and nowadays LEDs have been widely utilized in general lighting devices[1]. To test the quality of LEDs, more and more LED probes are required. In this paper, the efficiency of autonatic optical inspection system has been proposed. The proposed process is able to ensure high-quality of LED probes, improve the efficiency of production, reduce the output of defective products and feedback information of products in time. The detection framework use wide image processing, consists of Otsu threshold and Canny edge detection to get coarse edge, sub-pixel edge detection, object extraction, i.e. The proposed method can quickly and accurately analyze the size of LED probes whose angle error about 1% and radius error about 2%. The experimental results verify the effectiveness of our methods.