Research On Vision-based Geometric Shape Detection Method And System Implementation

With the development of micro-nano technology,more and more precise micro devices are produced and manufactured.For these precise and vulnerable micro devices,it is more and more difficult to detect the errors in processing and manufacturing.The original manual detection of micro-device errors can not meet the modern development requirements.Therefore,the error detection for precision micro devices has become the focus of attention in the industry.In this paper,the following studies were conducted for visual detection:1.The overall design of the visual inspection system,the reliability for the combination of hardware and software,the functions of each module in the software and the cooperative relationship between them are studied.The method for identification and detection of the visual inspection system is proposed,and then the entire visual inspection system is generally analyzed.2.In order to achieve higher recognition and feature extraction accuracy,the system collects images and preprocessing modules for the visual inspection system,and a method based on template area dichotomy to automatically select thresholds is put forward.Compared with the traditional automatic threshold selection method,higher recognition and feature extraction accuracy can be achieved,and high-definition part images are obtained.3.The anisotropic bilateral filtering is used to complete the smoothing of the target part.The target region is extracted by the combination of threshold segmentation and area denoising.Meanwhile,the NMI feature target matching and shape template matching algorithm are added to identify the part image.These two algorithms are aimed for image recognition of different types of parts.4.C++and library functions are used to programme a visual inspection system,which mainly includes image preprocessing module,image recognition module and measurement error module.The effect of high precision measurement error is achieved through the mutual assistance of the three modules.5.A large number of testing experiments have been carried out based on the geometric characteristics of small parts and the detection accuracy of small part errors.Through the analysis of the experimental results,the error of the tiny parts is obtained,and the accuracy meets the requirements of the subject.