氣囊式安全帶進氣導管之影像視覺特徵檢測

Inspection of Inflatable Seat Belt with Intake Duct using Visual Image Processing

張志忠(Chung-Chih Chang)

指導教授 ： 張文陽

國立虎尾科技大學/工程學院/機械與電腦輔助工程系碩士班在職專班/碩士(2014年)

摘要

本研究主要是鑒於一些汽車零組件在加工或組裝製程中會因為該零件為保安部品，必須確保該零件之可靠度，無遺漏工程或有功能性之重要尺寸錯誤，因該錯誤會造成車輛或人員重大之危安，日後更會因該問題而造成車輛大規模之召回，造成製造商及車廠莫大之損失。因此在製造加工過程中所有會造成安全性危害之尺寸或特徵部分必須獲得完全之確保，在加工過程中大部分之尺寸皆可藉由製程能力來管控，但一些如毛邊、裂縫、漏裝零件…等外觀類還有加工上製程能力不足之尺寸，就必須於最終包裝出貨前必須做全檢之動作，以避免不良品之流出。
本研究是著重在氣囊式安全帶進氣導管之影像視覺特徵檢測，研究上使用CCD鏡頭來擷取影像，並利用Matlab軟體將影像做白平衡、裁切、二值化、像素換算及XOR影像相減等方式處理，其中外徑量測再與外徑分厘卡及市售影像量測儀KM-6210之量測結果作比對，瑕疵檢測則利用標準件與瑕疵件相減後影像差異值來作判斷，本實驗進氣導管之Pipe外徑尺寸為14.1mm公差+0.3mm，Cap瑕疵檢測則是使用限度樣品做為比對標準。
實驗結果顯示，使用影像軟體計算與分厘卡所量測之最大誤差值為0.037mm(0.262％)，佔公差值12.33％，依照一般使用時公差上下會緊縮10％，其量測結果是可以滿足需求的，另外在瑕疵檢驗部分，本實驗所準備之瑕疵樣品與標準件比對後誤差值在0.35％~11.43％之間，與設定之誤差值0.15％，皆能正確判斷出來，故瑕疵檢測部分也是可以達到需求的。

關鍵字

影像視覺 ； 物件辨識 ； 邊緣偵測 ； 瑕疵物 ； 進氣導管

並列摘要

The study investigated the inspection of inflatable seat belt with an intake duct using visual image processing. In general, the manufacture and assembly precesses of vehicles's components are an important issue for safety reliability. The critical dimensions of vehicles's components resulted in secure risk should be totally verified during the manufacturing process. The failure of secure parts will damage the vehicles and people seriously, and the vehicles will be recalled due to this problem that causes automaker and manufactories a great loss. Most of dimensions can be controlled by process capability in machining process. But regarding the appearance classification, such as burrs, cracks and missing parts, or the insufficient process capability of machining dimensions shall fulfill 100% detection process before final packaging and shipping to avoid outputting of defective products.
This study focuses on the visual features of image processing system for active seatbelt tube body, captures the image by CCD camera and utilizes Matlab software to process the image white balance, trimming, binary coding , image pixel translation and XOR subtraction processing etc.,. The outside diameter detection value will compare with the measurement of outside micrometer and Video Measurement System KM-6210. The defect detection judgment is by means of the images subtraction difference between the standard specimen and defective parts. The Cap defect inspection uses limited sample as comparison standard.
The experiment results show that the value of maximum deviation is 0.037mm (0.262%) between calculation of image processing system and outside micrometer measurement, and it accounts for 12.33% tolerance value. In accordance with the normal utilizations, the tolerance will be shrunk 10% so the measurement results meet the requirement. Regarding the defect inspection in this experiment, the comparison deviation between defect samples and standard specimen is about 0.35% ~ 11.43%, the setting deviation 0.15% can be verified precisely so the defect inspection tallies with the expectation.

並列關鍵字

Visual images ； Defect inspection ； Edge detection ； Intake duct

參考文獻

[9]許裕騄，孫永年，"棒鋼表面缺陷自動化檢測系統"，國立成功大學碩士論文，2007年

[11]Vaibhav Gupta, Rachit Puri, "Prompt Indian Coin Recognition with Rotation Invariance Using Image Subtraction Technique", Devices and Communications (ICDeCom), 2011 International Conference, February, 2011, Pages 1-5.

[26]呂儒泓，黃惠藩，"應用不受光照影響之局部二元圖形於影像比對"，亞洲大學資訊工程學系研究所碩士論文，2011 年6 月

參考文獻

[1]Jankywolf FMEA失效模式與效應分析: http://wenku.baidu.com/view/6375d559be23482fb4da4cb5.html