Three-dimensional Reconstruction Method Of Case Scene Based On Structured-light

In the field of public security,the scene investigation is an important part of the case investigation.However,affected by factors such as the complexity of the site and the volatile environment,the case scene is prone to damage and ignore.New technical methods are urgently needed to improve the efficiency of scene investigations.The structured light-based 3D reconstruction method is a new type of 3D imaging technology that combines computer vision and information optics.It has the advantages of low cost,high precision,fast speed,noncontact,easy implementation,and dense imaging,which can meet the needs of current scene investigations.In order to improve the adaptability of 3D reconstruction of the case scene,this paper uses the fringe projection measurement method to accurately reconstruct the key evidence.The depth data obtained by depth camera is used to reconstruct the case scene in real time.The algorithms are optimized for these two methods respectively.The specific contributions are as follows:First,a feature point detection method based on concentric circular grid is proposed.This method uses concentric circles grid as the camera calibration pattern,and the local sub-pixel edge detection algorithm and the common self-polar triangle of concentric circles are leveraged to accurately locate the sub-pixel coordinates of the feature points.In order to improve the robustness of the detection,each concentric circle is composed of multiple circles with gradually decreasing radius.This multi-level geometric structure establishes a feature point voting scheme to solve the problem of numerical instability caused by the calculation of the projection center.Simulations and experiments have demonstrated that the performance of feature point detection using concentric circular grids is better than the currently popular grid patterns under the influence of image noise,lens distortion,and resolution.Second,a high-frequency sinusoidal fringe pattern that can suppress surface reflection is designed,and a system calibration method based on a rational function distortion model is proposed.This method brings the rational distortion model into the out-of-plane height expression,and effectively eliminates the influence of the lens distortion of the camera and the projector on the measurement results.In the case of arbitrarily placed measurement components,the three-dimensional model of the key evidence surface can be retrieved accurately without the precise instructors and projector calibration.Finally,a depth data preprocessing method based on guided filtering is used.This method can perform depth data processing while preserving depth edge information.The 3D reconstruction method based on the depth camera performs real-time 3D reconstruction of the real scene through four steps of depth data preprocessing,3D model fusion,back-projection model and poses estimation.Using this method to reconstruct the freiburg1＿room data in the TUM dataset,the absolute maximum error between the measured trajectory and the real trajectory accumulated within 1351 frames is only 0.207 m.