Implementation of Edge & Shape Detection Techniques and their Performance Evaluation

In this thesis, we develop an industrial image processing application for baked goods. Images of baked objects on a conveyor belt are taken by high resolution cameras batch wise throughout the baking period. The network is designed with high performance equipments and the application is fast enough to complete all the steps within the allowed time window. The application uses Canny edge detection method [6] which optimizes the performance compared to other applications used in the industry. We have analyzed the performance of different key properties; such as processing time of an image, dimensions of an object shape, average color value to detect if an object is properly burned or damaged. Performance in detecting shapes shows higher accuracy for the developed application against other applications used in the baking industry.