Research And Development On The Defect Image Recognition Technology In Automatic Strip-Steel Surface Inspection System

According to the steel company's requirement about the strip steel's surface quality, based on the advantage of the image processing and high-speed calculating-ability of the computer language, this system is developed with analyzing the information in the strip steel's surface image for the detection of product. This paper puts emphasis on the technical research of image recognition, and utilizes the image's features about color, texture and morphology in the image's similarity-comparison, which can recognise which typical flaw it is. So this inspection system can liberate workers from the dull, heavy and traditional artificial-detecting work, as well as, can improve the working efficiency under ensuring the product quality.The main achievements of this paper are listed as follows:1. According to the company's technical requirements and practical circumstance of the strip steel's production-spot, the detection system's overall design plan is put forward. It includes the parts of acquiring image (IA), image's pre-processing (IP) and image-recognition (IR), which transits image data through local area network.2. Advancing the method for the acquirement of the noisy image in IP part and the design scheme through analyzing image data.3. Putting forward the design scheme about the image-recognition (IR) part'ssoftware and hardware, which includes the module of image recognition, management of flaw-image database and system.4. According to comparing characteristic, algorithm, test results of some methods, this paper puts forward the design scheme of color-based recognition.5. According to the comparing the algorithm, test results, putting forward the design scheme of texture-based recognition.6. According to the comparing the algorithm, test results, putting forward the design scheme of morphology-based recognition.7. Putting forward the building and upholding methods of flaw-image database that applies the serializing method and can put such as reading image's characteristic into effect.8. Adopting the operation platform of Windows NT by Visual C++6.0, and this software has the trait of friendly interface, convenient application, easily upholding, safe reliability.