Research On Algorithm Of 3D Reconstruction Of Material Pile Surface

The three-dimensional reconstruction of the material pile surface has always been one of the hot research topics in the field of measurement.The basic steps of 3D reconstruction are image acquisition,camera calibration,feature extraction,stereo matching,and 3D reconstruction.Aiming at the disadvantages of complex 3D reconstruction methods,high cost and low precision,this paper proposes to use the active stereo vision method to mark objects with laser grid and establish a prediction model to realize the mathematical mapping between pixel coordinates and world coordinates.Material surface model.The specific work and innovation of the article are as follows:In the process of laser marking image feature extraction under natural illumination background,this paper proposes an adaptive clustering laser marker image extraction algorithm based on information entropy for the automatic extraction of laser features and the flaws of color image edge detection and overdetection.The algorithm uses the information entropy evaluation function and the cluster number as the clustering number optimization index of K-means clustering,and automatically determines the optimal clustering number of K-means clustering.Experimental results show that the algorithm can extract laser markings automatically from complex backgrounds more accurately.Aiming at the defects of traditional calibration,such as many parameters,poor operability and low calibration accuracy,an improved artificial bee colony optimization least squares support vector regression machine(ABC-LSSVM)prediction model is proposed.The artificial bee colony algorithm is easy to fall into the local optimal solution and has low convergence precision.The Rayleigh distribution function is added to the neighborhood search of the following bees to enhance the randomness of the search.Verifying the four benchmark functions from different dimensions and improving the artificial bee colony algorithm verification results improve the convergence accuracy and reduce the number of times the local optimal solution is trapped.Compared with PSO-LSSVM,GA-LSSVM and standard ABC-LSSVM,the improved algorithm has improved the MA,Y and Z model MAPE of ABC-LSSVM to 0.061,0.125 and 0.055,and the RMSE is 0.3793,0.4241 and 0.309,which improves the prediction performance.According to the experimental model data of the improved artificial bee colony optimization LSSVM prediction model input matching,the spatial point coordinate values are obtained.The average errors of the models X,Y and Z are 1.22%,2.25% and 1.1%,respectively,the effects are better.To a certain extent,the measurement method of material pile surface reconstruction of this subject has realized the automation of material surface measurement,and verified the correctness of the algorithm through laboratory modeling test.The method has strong operability,low cost and is suitable for general promotion.