Evaluation Methods For The Quality Of Direct Rendering Of 3D Point Cloud

With the widespread application of point cloud data,objective evaluation of the quality of point cloud data has become a research hotspot.In recent years,many scholars have conducted a lot of researches in the denoising,simplification,registration and 3D reconstruction of point cloud data,and have made technical breakthroughs.However,research on the evaluation of point cloud data quality is still in its infancy.The traditional single-point accuracy evaluation method cannot accurately describe the overall quality of point cloud data.To this end,some researchers have proposed that by evaluating the image quality obtained by directly drawing point cloud data,the quality of point cloud data,as well as the quality of point cloud processing and rendering algorithms,can be evaluated indirectly.For this purpose,this thesis conducts a study on the image quality evaluation method of direct point cloud rendering,and applies it to the evaluation of the quality of point cloud processing algorithms.The main work and innovations of this thesis include:This thesis mainly studied the method of evaluating the quality of the image directly drawn by the three-dimensional point cloud,and indirectly evaluated the quality of the point cloud processing algorithms.The main work and contributions of this thesis are as follows:(1)Research on the quality evaluation algorithm of point cloud direct drawing.Since the projected image after the point cloud is directly drawn and it is a two-dimensional image,this thesis improves the image quality evaluation algorithm based on visual saliency for the direct drawing of the three-dimensional point cloud.The first innovation of this paper is visual salience quality evaluation algorithm combining global and local information.The algorithm extracts the global information of the image through fractional calculus,uses the visually significant view as the weight to fuse the local parts,performs local fusion on the visually significant evaluation and gradient features,and fuses the global information and local information similarly Quality evaluation results.The proposed algorithm is tested in four selected public databases.The test results show that the optimized algorithm achieves the improvement effect on the image quality evaluation algorithm based on visual saliency.(2)Research on quality evaluation method of point cloud direct drawing.At present,there are relatively few studies on point cloud direct drawing quality evaluation methods.There is a lack of comprehensive evaluation schemes,and there is no comprehensive analysis of various factors that affect the evaluation results.Drawing on the quality evaluation method of two-dimensional images and the three-dimensional model quality evaluation method,the second innovation of this thesis is to propose a more accurate and comprehensive quality evaluation method that combines visual saliency,image structure information,and image pixel information.By comparing the evaluation results of the evaluation method with other algorithms and subjective quality evaluation results,the experimental results and comparative analysis show that the evaluation results of the quality evaluation method proposed in this thesis are basically consistent with the results of the comparison algorithm,but closer to the subjective evaluation results.The quality evaluation method proposed in this thesis is more comprehensive.(3)Quality evaluation of the preprocessing algorithm for direct rendering of point clouds.When acquiring actual point cloud data,due to the occlusion of the object itself and uneven illumination,the 3D scanning device can easily scan some areas of complex-shaped objects as visual blind spots,and some objects with large sizes that cannot be measured at one time.Multiple scan measurements must be performed.The scan results are often multiple pieces of point cloud data with different coordinate systems and noise,which cannot meet the requirements for the authenticity of the digital model in the later period,so the point cloud data needs to be denoised,simplified,and matched.This thesis uses the quality evaluation method to evaluate the quality of the point cloud data after the point cloud data processed by the point cloud denoising algorithm,point cloud simplification algorithm and point cloud registration algorithm.The experimental results prove that the quality evaluation method proposed in this thesis is suitable for evaluating the effects of three preprocessing algorithms for point cloud denoising,simplification and registration.