Body Size Measurement Of Dairy Cows Based On 3D Point Cloud Data

The body size parameters of dairy cows can not only reflect the growth and development status of dairy cows,but also reflect the performance indicators such as reproductive ability,milk production ability,and economic value of dairy cows,which have important application value for modern dairy cattle breeding.At present,the body size parameters of dairy cows are mainly measured by manual contact,and the measurement indicators are heavily dependent on the subjective judgment of the measurer,and the manual measurement workload is large,which affects the living habits of dairy cows and cannot meet the development requirements of modern dairy farming.With the development of computer graphics and three-dimensional imaging technology,the application of obtaining geometric parameters of objects from threedimensional models is more and more extensive,and it has great application potential and prospects in the measurement of cow body measurements.In order to achieve non-contact automatic and objective measurement of cow body size parameters,this paper proposes a cow body size measurement method based on 3D point cloud data.The main research contents of the paper are as follows:Analyze the performance characteristics of different depth cameras,determine the selection of the camera,use the camera calibration to obtain the internal and external parameters of the camera,and realize the conversion of the depth image to the 3D point cloud.According to the limited working field of the camera and the size of the cow,a cow point cloud data acquisition platform is designed and built,and the cow point cloud image is collected.The processing methods of cow background wall point cloud and outlier point cloud filtering,ground point cloud segmentation,etc.are studied to remove the interference of factors such as ground,background,and uneven illumination.For the background wall point cloud,it can be effectively removed by the method of straight-through filtering;for the outliers around the cow point cloud,the two methods of radius filtering and statistical filtering are analyzed and compared,and it is verified that the statistical filtering is more effective than the radius filtering.At the same time,it can effectively retain the edge information of cow point cloud data;for the segmentation of ground point clouds,this paper analyzes the shortcomings of Euclidean clustering algorithm and random sampling consistency algorithm,and proposes an improved random sampling consistency algorithm based on PCA adaptive threshold,which verifies that the improved algorithm can effectively extract ground points.It has strong robustness and high stability,and can obtain cow point clouds that exclude external environmental interference.The research proposes an automatic measurement method of body size parameters based on 3D point cloud of dairy cows.The position relationship between the body size feature points and the contour,shoulder,abdomen and tail of dairy cows was analyzed,and the target region of body size feature points was determined.The maximum curvature of the target region and the attributes of the feature points were used to realize the location of the body size feature points.The Euclidean distance measurement method was used to convert the corresponding body size parameters between the body size characteristic points,so as to realize automatic measurement of seven body size parameters including height,depth,straight length,oblique length,shoulder width,belly width and waist width of dairy cows.In order to verify the accuracy of the automatic measurement of cow body size parameters,the cow body type appraiser used a tape measure and other tools to manually measure seven kinds of cow body size parameters,and compared and analyzed the results under the two measurement methods.The average absolute error of the seven body size parameters automatically measured in this paper was 2.33 cm,and the average relative error was 2.73%.Among them,the mean absolute error of the cow’s rib width parameter is the smallest,which is 0.98cm;the mean absolute error of body height,body depth,shoulder width and abdomen width are all smaller than the mean absolute error,which are 1.38 cm,1.14 cm,2.21 cm and1.93 cm in sequence;The average absolute errors of body straight length and body oblique length are 3.94 cm and 4.69 cm,respectively,so it is feasible to propose a method for measuring the body size of dairy cows based on 3D point cloud data.