DIMENSIONAL INSPECTION SYSTEM FOR MESO-SCALE ARTEFACTS USING SUB PIXEL EDGE DETECTION

Abstract

Due to the introduction of Nano & Micro technology in the fields of Mechanical, Computer Medical, Chemical and Electronics etc, the world goes towards the manufacturing of small components but inspection of these parts is a big challenge for all. Conventional methods for inspection cannot meet the need of the modern world with required accuracy. Non-contact system has to be introduced to inspect these artefacts accurately without damaging. Computer vision system is used to model the real world data through digital image. Digital image is used to explain and reflect the data in image form. Edge detection technique is used to find the edges in the image to extract the required information. Many edge detection techniques are used for this purpose such as Sobel, Prewitt, canny and LoG etc. Canny and LoG techniques are only for edge detection on pixel level and silent to explain the sub pixel level edge detection. Sub pixel edge detection through curve fitting using sum of least square is used to refine the measurement for inspection of Meso scale (50µm to some mm) artefacts. Different curve fitting techniques are applied. The best curve fitting is implemented on the different features of the different samples. Comparison of this technique is also carried out with Canny pixel level edge detection method. This technique is also implemented on the various resolution i-e for various megapixel Scale. The behavior of the pixel and sub pixel edge detection techniques is tested for different mega pixel scale. Different samples with different features are also tested to check the validity of the proposed inspection method. Accuracy of this inspection system is directly proportional to the accuracy of the edge detection techniques. In pixel level precision can be increased by increasing the number of pixel in the image. Sub pixel level reduced the error of the system by taking the information inside a pixel. The precision is hence increased by locating the edge within the pixel.