Research On Body Size Measurement And Weight Estimation Of Cow Based On Three-dimensional Reconstruction

With the improvement of living standards,people’s demand for dairy products and quality requirements are also increasing.In the process of dairy cow feeding,in order to improve milk production and milk quality,the monitoring of the health of the cow is particularly important,and the health of the cow is closely related to the measurement of body size and weight.However,most of the current farms use traditional manual measurement methods to measure the body weight of dairy cows.This method is not only easy to cause the stress response of dairy cows,but also has low efficiency and heavy workload.In response to this situation,this study proposed a method for body size detection and weight estimation of dairy cows based on three-dimensional reconstruction.The main content and conclusions are as follows:(1)A cow depth video acquisition platform was built.After the depth image frame was converted into three-dimensional point clouds,the designed algorithm realized the three-dimensional reconstruction of the cow body surface point clouds.Aiming at the problem that the point clouds of side view was missing due to fence occlusion,a video-based different frame completion method was proposed.This method used the difference in the horizontal direction of the two side view point clouds to complete the point clouds.A method for selecting a symmetry plane based on the spine of a cow was proposed.This method was used to mirror and symmetry single-view side-view point clouds of a cow to obtain double-view side-view point clouds,and completed the three-dimensional reconstruction of the point clouds on the surface of the cow.The results showed that the success rate of three-dimensional reconstruction of the point clouds on the body surface of dairy cows was relatively high.(2)The cow body size was automatically detected.After reducing the point clouds to a two-dimensional image,a method for extracting cow’s body measurement points based on the peak of the projection curve was proposed.The measured values of cow’s body straight length,shoulder width,abdomen width,and hip width were calculated.The reconstructed point clouds extracted the body height measurement points and calculated the size.The developed interactive dairy cow body size detection system was calibrated,and the results showed that the average relative error of the linear body size measurement was less than 2.562%.A projection curve method based on point clouds slices was proposed to detect the position of the cow’s bust,and the bust curve was fitted with an improved moving least square method to approximate the bust size.The results showed that the absolute error of the bust measurement points detected by this method was between-2.25 and+2.80 cm,the average relative error was 3.30%,and the success rate of bust curve fitting was high.(3)The weight of the cow was automatically estimated.A method for predicting the weight of dairy cows based on surface reconstruction was proposed.This method reconstructed the point clouds of dairy cows based on the envelope of the point clouds.The volume and surface area parameters of the reconstructed surface model were used to construct a dairy cow weight estimation model.The accuracy of the model was tested using cow test samples.The results showed that the overall surface area,the volume of parts that removed limbs and head of the cow’s curved surface model were significantly positively correlated with cow’s weight.The absolute error of the weight estimation model was from-24.287 to+17.242 kg.The average relative error between was 1.382%,and the weight estimation model has a high accuracy.