Methodology and experiments on spiral removal of pineapple spines based on 3D reconstruction
[Objective]The cultivation of pineapples in China boasts a long history,and the cultivating area and the yield both account for approximately 7% of the total global cultivation.In 2018 alone,the annual yield of pineapples in China reached 1.64 million tons,30% of which underwent further processing.Although there have been significant advances in the research on the removal of pineapple peel,the removal of pineapple spines remains unexplored.To realize the automatic removal of pineapple spines,a trajectory search algorithm for the spiral removal of pineapple spines is proposed with the aim of improving the efficiency and accuracy of the removal and satisfying the requirements of the automatic operation.[Methods]First,the depth camera is fixed on the spine removal stand,and the camera is adjusted such that the center of the picture targets the central axis of the pineapple's rotation.The pineapple fixed on the fixture rotates at a constant speed,and pictures are taken at every 3.6° of rotation.After the pineapple accomplishes one revolution,a set of sequential images is obtained.Then,procedures such as bilateral filtering,region interception,and threshold segmentation are conducted on the obtained RGB images and depth maps to extract the outlines of the pineapple and the reference disk.The internal parameters of the camera are determined through the Zhang Zhengyou calibration,and the external parameters are calculated based on the motion equation of the pineapple constructed from the analysis of its motion.Further,the preprocessed images are fused into a three-dimensional point cloud based on the conversion relationship between each coordinate of the camera imaging system.Then,the matrix of rotation and translation is calculated along with the scaling by fitting the equation of the plane where the reference disk is located to correct and scale the initial point cloud of the pineapple.Then,the Otsu method is utilized to determine the optimal binarization threshold for the point cloud and extract the point cloud of pineapple spines.The DBSCAN clustering algorithm is integrated to group the point cloud of pineapple spines and calculate the central coordinate of each spine.A trajectory search algorithm is proposed to solve the problems arising from the missed and deviated spines in the trajectory search by calculating the matching costs.Finally,the terminal path of the removal is formulated based on the trajectory found to realize the spiral removal of pineapple spines.[Results]Experiments were conducted on 15 Chinese Guangdong Xuwen pineapples.The experimental outcomes showed that the complete removal rate of spines averaged 98%,while the incomplete removal rate was 1.39% .The removal rate exceeded 95%,and the average time consumption was 41.7 s.The removal yielded high efficiency and favorable effects.Among these,a small number of spines were missed in the detection due to the malformation of their growing position.However,due to the high matching costs in the trajectory search for the removal,these spines were ignored by the algorithm,which exerted no impact on the whole removal trajectory.The average number of missed pineapple spines was 0.53,and each pineapple yielded 8-9 removal trajectories.Further,the average spine number of each strip was 11.1,indicating that the trajectory search for the removal of the spines boasted a rather high quality.For the eccentric distance of the spines,the mean eccentric distance of the spines averaged 0.57 mm,the maximum eccentric distance averaged 1.01 mm,and the root-mean-square error of the eccentric distance averaged 0.62 mm.For the overcutting depth of the spine H,the mean overcutting depth of the spine averaged 2.63 mm,the maximum overcutting depth averaged 2.9 mm,and the root-mean-square error of the overcutting depth averaged 1.53 mm,indicating that the overall error was rather small.[Conclusions]The precision of the spine removal meets the requirements of the automatic spine removal operation,indicating its ability to provide technical support for the automatic removal of pineapple spines and significantly improve its efficiency and quality.
pineapplespiral removal of pineapple spinesthree-dimensional reconstructionpoint cloud processingtrajectory search
万方数据
维普
