基於修改電位能函數的新穎銲點定位方法

Abstract

近年來，資訊產品已成為我們日常生活中的必需品之一，由於人們對資訊產品的依賴性愈來愈高，因此，工業生產的技術也逐漸受到重視。在工業生產中，自動化系統已被廣泛利用多時，銲點定位技術是其中一項重要的課題。由於現今工業生產所需的微電子元件 (Microelectronics) 體積較小，若是用人工技術去做銲點定位的工作，勢必會消耗大量的人力成本，在效率上也會大打折扣，因此，我們希望能夠開發一套定位系統，利用影像處理的技術取代人工技術來實現銲點定位的工作。 在拍攝環境中，往往會受到光源的影響而造成影像色彩的不穩定性，因此我們開發一套不需要色彩資訊的自動化銲點定位系統，藉由使用電感的梯度特徵 (gradient feature) 以及輪廓資訊 (contour information) 作為我們的銲點特徵，最後，我們提出一套以電位函數 (potential function) 為概念的定位演算法。另外，在實驗當中，發現有些測試樣本利用電位函數定位結果有些微的偏移，為了能改善這類的問題，我們進一步提出一個修改過的電位函數 (modified potential function)，利用一個自動建立外加電場的概念，將最後的定位點吸引到我們期待的位置上。 在最後的實驗比較中，對本篇論文處理的三種元件而言，我們可以發現利用修改過之電位函數 (MPF) 定位的結果比電位函數 (PF) 定位的結果來的好，在準確性方面都有著大幅度的提升。

In recent years, information products have been one of the necessaries in our life. People pay more attention to the industry techniques of production due to the high dependence on information products. In the industry application, automatic system has been utilized for a long time, and solder-joints locating technique is one of the important tasks. Due to the tiny size of microelectronics, implementing the locating task by human eyes results in low efficiency and high consumption of man power. Therefore, the locating system based on image processing is proposed to locate solder joints instead of human eyes. Because the influence of light source the color information of images is unstable in the locating environment. Therefore, we proposed solder joints locating system without color information. We use the gradient feature and contour information of inductor as the features of solder joints. Finally, we propose a locating algorithm based on the concept of potential function (PF). In addition, we observe that the locating results by utilizing potential function are deviated. In order to improve the locating results, we modify the potential function further as modified potential function (MPF), the concept of MPF is that we add the external electric field to attract the locating points to the expected position. In the experiments, we compare between PF and MPF for the three components .The results demonstrate that MPF outperforms PF.