Research On Image Measurement Technology For Cutting Tool Geometric Parameter

Tool size measurement package was presented through automatic identification that applied to the initial tool image or interaction in order to complete the measurement, which was the typical workflow of cutting tool presetters.But it was more recommended to extract and process or infer the original information by automatic or intelligent method rather than interaction for improving the performance and quality of the system, which was also one of the most important research direction in tool detection technology.Firstly,Combining the conception and strategy of tool information management system with the characteristics of the cutting tool structure,a new method was proposed that using tool imaging area projection profile information and nominal size of the tool diameter and nose radius and blade number type-ahead as the searching source information of tool identity.Then Unified the mixture information that contained text and image information to a string and different extents hierarchical code was weighted to realize the retrieval of different kinds of tools.As the core section of the code, the contour segment string is composed of some elements of quadrant position, the number of arcs and edges, the combination form of arcs and edges, contour convexity, the angle type of main and auxiliary blades. These elements are from the graph of normalization of interest region in the image.Secondly,according to the information design such as tool nominal parameters and quadrant,dynamic Crosshair is applied to automatically cut the interest area and to prevent information redundancy caused by the span of tool sizes. In order to prevent excessive segmentation of the traditional polygon approximation algorithm, the fusion algorithm combining primitive recognition and contour segments is adopted to fuse and identify those contours.For the purpose of restraining the outlier influence,Weighted Least Squares is utilized to synthesize the contour data to geometric primitives;and the improved algorithm is verified by experiments that it’s adaptable and advantageous in assuring efficiency,multiplying contours and accurately positioning segmentation points.Finally,we create an experimental platform for tool retrieval and set the pattern for tool spectrum,choosing six kinds of tools which are different in size and structure to perform the experiment.The validity of retrieval results is generally over 92%;because a kind of tool noseradius is too small and the system has set the limit threshold, its accuracy is 83%.Research shows that:both the way of encoding which can turn tool information to texts and the retrieval method can be used to search or identify tools;The contour collection and recognition in interest areas of tool images is the key to identify various tools,and the factors like nominal size can ensure the recognition process in progress;The primitive recognition of image contour,as the article has mentioned,is of certain values for application and reference.