Defect Detection System Design For Steel Surface Based On Machine Vision

The quality of the transformer cooling fins is affected by the internal stress in the production process seriously. How to detect the surface defects of cooling fins accurately and remove them from production line in time is the key to improving the quality and production efficiency.In order to overcome the lack of real-time capability and accuracy caused by artificial surface inspection with eyes in the roll forming process of transformer cooling fin, an online automatic defect detection system for steel surface based on machine vision is designed. In the aspect of algorithm, aiming at the poor performance in removing salt-and-pepper noise with traditional linear and nonlinear methods, an improved median filter with combination of traditional ?-median filter and the weighted filter, named weighted ?-median filter, is proposed. The improved method takes both gray information and spatial information into account by giving a specific weight to the each member of the intermediate set of median ?-median filter. The simulation results show that the proposed method reduces salt-and-pepper noise effectively while preserving the original image details. In the aspect of the system hardware, a set of online defect detection system is conducted, which can be divided into three parts, namely image acquisition, software analysis and action execution. In the aspect of the system software, a steel surface defect detection software is developed based on Visual Studio 2013 and Open CV. The software receives the images taken by a CCD camera, and then processes them with the improved Canny algorithm and the BP neural network. The Programmable Logical Controller(PLC) will control the electric actuators to remove defective steel after receives the commands sent out by the software.The system has gotten a good simulation result. It will be applied to the production line.