Research On 3D Reconstruction Of Dairy Goat Based On Single Image

3D reconstruction is an important research direction in the field of computer vision,which has important research significance in animal husbandry,medical treatment and digital archaeology.Dairy goat is an important dairy breed goat in China,and with the increasing scale of dairy goat farming,the monitoring of individual dairy goats and the efficient management of goat farms have become new issues.The 3D reconstruction of dairy goats can change the measurement method of dairy goat body size from manual contact to non-contact,and obtain the dairy goat body size information in real time,which helps to realize the efficient management of dairy goat breeding.Therefore,this paper uses the single dairy goat image collected from the livestock teaching experiment base of Northwest A&F University as the research object.Based on the 3D morphable model of dairy goats,the dairy goat is reconstructed in 3D by establishing the objective optimization function and using gradient descent and golden segmentation optimization algorithms,respectively.The results of the reconstruction were evaluated.The main research contents of this paper are:(1)Research on 3D Morphable Model of Dairy Goat.For the problem of mesh tearing and topology change during the 3D reconstruction of dairy goats,This research implements a parametric representation of a 3D model of a dairy goat,By performing linear blend skin binding on the 3D model of the dairy goat,the parametric representation of the shape and pose of the 3D model of the dairy goat is realized by using principal component analysis and bone motion respectively.The dairy goat 3D model proposed in this paper can effectively solve the problems such as mesh tearing occurring in dairy goat deformation,and has fewer parameters,which is convenient for the regression of subsequent objective optimization algorithms.(2)Research on gradient descent based 3D reconstruction algorithm for dairy goats.In response to the discomforting problems in the 3D reconstruction of dairy goats,In this paper based on the feature points and contour segmentation maps of dairy goat images,the establishing the reprojection objective optimization function,after aligning the dairy goat 3D model and the image,the dairy goat 3D model and the dairy goat in the image are fitted using SGD,BGD and MBGD gradient descent algorithms,realize the 3D reconstruction of dairy goats,solving the problem of discomfort in 3D reconstruction of dairy goats.The experimental results showed that the 3D model of dairy goats reconstructed using MBGD method PCK@0.1accuracy reached 73.42%,which has high reconstruction accuracy.(3)Research on 3D reconstruction algorithm of dairy goats based on golden section search.For the problem of slow speed of gradient descent based optimization algorithm in 3D reconstruction of dairy goats,we propose a golden section search based on Improved objective optimization algorithm.Replace the gradient descent algorithm,which requires constant derivation for iteration,with the golden section search algorithm,based on the skeletal hierarchy of dairy goats,the parameters are continuously optimized so that the control dairy goat3 D model area will be continuously reduced,enhancing the speed of reconstruction.Experiments show that the dairy goat 3D reconstruction algorithm based on golden section search can guarantee PCK@0.1 accuracy while reduce reconstruction time.Finally,the dairy goat3 D reconstruction method based on the golden segmentation search is implemented as a dairy goat 3D reconstruction platform.In summary,in this paper,by research on the 3D morphable model of dairy goat,establishing the objective optimization function,and fitting the dairy goat 3D model and pictures based on gradient descent and golden search algorithm,and the results of the reconstruction were evaluated,which had a positive effect on improving the efficiency of dairy goat husbandry.