Research On Key Algorithms For Visual Positioning And Recognition Of Workpieces

In recent years,machine vision technology has developed rapidly and is constantly being used in various industries.For industrial sewing machine sewing material positioning,the traditional way is that is manually position by holding the template of the sewing material.But this way is low efficiency,poor precision,high cost of making templates.This process is done manually,which has ineffective,poor precision and high cost of making templates.In order to solve these problems,it is necessary to introduce machine vision technology.However,current algorithms of feature extraction in machine vision mostly have many disadvantages including large computational complexity,low noise immunity,and poor robustness,such as SIFT,SURF,FAST and so on.These algorithms are difficult to meet the requirements of fast and precise positioning in the industry.In order to ensure the real-time and stability of positioning in the industry,this paper constructs a real-time positioning platform for workpieces,then studies and improves the key algorithms in the visual positioning of sewing materials.This paper mainly completes the following aspects: The construction of the visual platform includes a mechanical motion positioning platform,a visual positioning platform and a control platform.The mechanical motion positioning platform is a four-axis motion platform that realizes XYZ linear motion and rotational motion of the material table.The visual positioning platform is a monocular vision system consisting of industrial cameras,industrial lenses and light sources.The specific research and analysis of its design and selection are made in the paper.The control platform is composed of an image processing system and a servo system.The former drives the latter to complete positioning by calculating image position information in real time.The research of preprocessing algorithms: including background subtraction,threshold segmentation,edge extraction,image smoothing,etc..And then And optimizing the algorithms for actual sewing material image processing in the industry.Finally the algorithms are applied to the system and achieve good results.Improing the SIFT feature extraction algorithm : an invariant feature algorithm is proposed by using Harris and Gaussian scale causality,and its features are used for rough image matching.And then using the rotated template image to precisely match the captured image.Finally,the algorithm proves that it has high efficiency,strong stability and high matching precision.It can meet the real-time attitude extraction of workpieces in the industry to complete the accurate positioning of the workpiece.The experiment and analysis of the positioning system,using the four kinds of sewing materials in the industry as the object of the experiment to obtain the image processing positioning error and the actual positioning error,and then analyzing to summarize the reasons for the error.Finally,proposing a strategy to reduce the error.