Research On Geometric Analysis Algorithms Design Of Image Data And Its Applications

In the field of computer vision,pattern recognition and classification of specific scenes and specific objects has always been a hot spot and difficulty in the field of applied research,and its industrial applications are extensive,involving dynamic target recognition,security scenes,biometric recognition,industrial robots,automatic driving,land resources hyperspectral remote sensing monitoring and many other aspects,which is an important component of artificial intelligence and its application fields.Exploring new feature extraction algorithms for image data and promoting exploratory attempts to analyze and characterize texture images are undoubtedly of great scientific significance and practical application value.Based on the existing research results,this paper carries out the creation of a new paradigm for image data representation and application research from the perspective of complex network analysis,attempts to propose a novel geometric analysis algorithm for image data,and implements it into related industrial application scenarios,and achieves good results.Specifically,the main work of this paper is reflected in the following aspects:(1)A multi-layered pyramid-like network model is constructed based on the geometric analysis method of local global structural features of complex networks.This method is not only suitable for grayscale images,but also for multi-channel color images with different color space representations,and is a general image associated complex network modeling method.(2)A classification framework model is designed.Based on the local global correlation multi-layer class pyramid network model,the framework extracts network topology features for characterizing association images,and constructs simple graph and hypergraph feature network models respectively for the extracted feature matrix.Finally,the classification task of image data is converted into the classification of complex network model nodes.This framework integrates the current mainstream deep learning and graph neural network technology,and focuses on the application of complex network theory,which is an effective end-to-end framework model with strong application value.(3)The local global network model and the classification method based on complex network are applied to the application scenarios of industrial surface defect identification and detection.By constructing an associated complex network model for industrial surface defect images,high-dimensional explanatory vectors are generated and feature network models are constructed,and geometric digital features of image data are extracted to achieve the purpose of classification and recognition.The related algorithm has achieved good experimental results on the newly released industrial surface defect dataset.