| Abstract:When it comes to the production detecting automatically based on machine vision in industry, image processing is closely related to the timeliness and accuracy of the detecting system. And the extraction of region of interest (ROI) plays an important role and lays the foundation for the subsequent target detection and extraction, reducing computation cost and the complexity of the analysis process. At present, research on the extraction of ROI for the visual inspection of industrial production is insufficient, so it has an important engineering application value and significance for the related research. Therefore, a new technology of the accurate extraction of ROI based on modeling and matching is presented in the thesis. The technology includes modeling for interested area and accurate positioning for the graphics.In the thesis, a modeling technology of the extraction of ROI based on configuration method is put forward through analyzing the characteristics and composition of ROI in the process of defect detecting for mobile phone. That is to model for basic shape primarily, then to model for complex shape, to complete the model for ROI based on configuration method level by level. In addition, for the accurate positioning for the graphics, a classification by a tripartite attribution for composition of ROI is brought forward.Regarding the problem that there may exists larger offset between the image to be detected and template image, this paper proposed a method that uses hierarchical positioning in the ROI. Firstly, this method applies fast template matching method based on gray-scale projection to get the horizontal and vertical offsets of the image to be detected and template image, to get the coarse positioning area by adding the offsets to the template. Then, do precise positioning in the coarse positioning area of each graph. Here, the paper put forward a quadrilateral positioning method through finding precise edges based on block projection, a circular positioning method based on adapted threshold Canny circular edge detection and a method of irregular shapes positioning based on plane curve rotating matching. Finally, it fulfils the precise positioning of complex graphics by geometric positional relationship between the graphics based on the precise positioning of fundamental graphics.Finally, the technology of the extraction of ROI is applied to the surface detection system for mobile phone. Through the real-time online detection and analysis, the technology of the extraction of ROI put forward in the thesis can meet the requirement of accuracy and real-time efficiency in the industrial production. |
PDF Full Text Request