Abnormal Area Detection Method For Images In Texture Reconstruction

Nowadays,3D reconstruction technology has been widely applied into industrial design,mechanical manufacturing,cultural relic protection,forensic research etc.However,existing texture reconstruction techniques are susceptible to complex lighting environments.In order to improve the robustness of texture reconstruction technology,this paper has carried out a series of researches on the method of image anomaly detection in texture reconstruction.Firstly,this paper introduces the white light scanning system,data acquisition method and the spatial relationship in texture reconstruction.After that,by analyzing the shortcomings in the process of texture reconstruction technology and the types of abnormal regions that may appear in the texture image,it summarizes the causes and characteristics of various image anomaly regions.As a result,an image anomaly detection method based on visual consistency of multi-view images is proposed and a technical flow chart of abnormal region detection is designed.Secondly,this paper introduces the visibility judgment method based on the projection relationship of the collinear equation,and analyzes the back face misjudgment and patch occlusion that may occur.Then introduces the back face culling algorithms and classic occlusion detection algorithms.The visibility judgment method is improved by simplifying the principle of occlusion detection algorithm and optimizing the operation structure.Experiments show that the improved visibility judgment method greatly improves the computational efficiency while ensuring accuracy.Moreover,it studies the spectral features,texture features,gradient strength features and optical geometric features of texture images,and uses these feature descriptors to describe the projection area of the triangular patch in the multi-view image.In addition,the effectiveness of each feature to distinguish various image anomaly regions is derived from feature validity analysis.According to the analysis results of feature validity,the detection methods of each type of anomaly regions in the image are determined.After that,this paper compares and analyzes a variety of classification methods.Combined with the actual needs of this paper,the Mean Shift clustering method based on Gaussian kernel function is selected to classify the images in multi-feature space.Based on the principle of visual consistency of multi-view images,the classification results are marked,and the break lines are extracted and marked at the same time,thus the synchronous detection of multiple image anomalies in texture reconstruction is realized.Experiments show that the image anomaly detection method proposed in this paper can detect the multiple anomalous regions in the image in a complete and effective manner.Finally,the image anomaly detection method proposed in this paper is applied to texture reconstruction.Experiments show that this method can effectively improve the adaptability of texture reconstruction technology in complex lighting environment and enhances the practicability of texture reconstruction technology.