Research On High-resolution Image Mosaic Algorithms For Industrial Automatic Optical Inspection

Vision is an important way for humans to understand the world,and image processing is an important modern way for humans to process visual information.With the continuous progress and development of society and computer science,digital image processing technology has also developed rapidly,which is used in many important fieldsImage stitching technology is an important research direction in the field of digital image processing.As an effective method to resolve the contradiction between large field of view and high resolution,it has been occupied in many important research positions,such as medical image synthesis,panoramic autonomous driving scene construction,etc.A major application scenario for image stitching is industrial optical inspection.In the production and manufacturing process of the product,there will inevitably be various defects on the surface of the product,such as scratches and short circuits on the printed circuit board.Therefore,it is very important to add automated online detection during the production process to improve the efficiency of industrial production.Automatic Optical Inspection System(AOI)is an automated science and technology that integrates optical sensing technology,motion control technology,signal processing technology,and performs tasks such as measurement,detection,identification,and guidance in the industrial production process,which is also an important filed of image stitching technology applicationsThis paper mainly aims to improve the real-time and accuracy of image stitching in AOI industrial inspection.First,the paper introduces the current development status of AOI and image stitching technology.Then,by analyzing the detection target and purpose,the paper selects the appropriate image acquisition equipment,the appropriate image registration algorithm and image fusion algorithm for improvement.Finally,the paper makes subjective and objective evaluation of the stitching results.In terms of feature extraction,considering the purpose of improving real-time performance,hardware conditions,and targeted equipment construction,this article selects the ORB algorithm with faster detection speed for improvement,which mainly includes the improvement of FAST feature point extraction method,optimization of extraction speed,and improvement of the algorithm for filtering mismatch points.In terms of image fusion,in order to maintain the integrity of the target and avoid bad situations such as ghosting and blurring,the line-finding method of the best stitching algorithm was improved,and a combination of dynamic fusion width of fade-in algorithm was proposed.The subjective and objective evaluation of the stitching result is compared with the traditional algorithm,and the experimental results are given.The experimental analysis shows that the method can meet the real-time requirements of industry,meet the needs of industrial inspection in terms of details retention,and has high comprehensive performance.