Yak Weight Estimation Methods Based On Machine Vision

The yak's body size,weight and other growth indicators are very important for evaluating yak growth,reproduction and selection of breeding cattle.For traditional yak body measurements and weight measurement,manual measurement is usually used.The traditional manual measurement method has the disadvantages of producing physiological stimulus to the yak and heavy workload.Based on the above considerations,this article uses the method of machine vision to estimate the body size and weight of yak.The main contents and results of this study are as follows:(1)Research on yak weight estimation model.This article uses linear regression to estimate yak weight.There are autocorrelation and collinearity problems among the independent variables such as yak body height,body oblique length,chest circumference,cross section height,and tube circumference.Based on correlation analysis,this paper uses linear regression,Gaussian process,neural network,and support vector machine to construct a yak weight estimation model using variable regression and multivariate regression.After testing,the yak weight estimation model based on multiple linear regression has an R2 of 0.8001 and an average absolute error of 14.1916%.The prediction effect is better than the Gaussian process,neural network and support vector machine.This method is an effective method to construct a pig body weight estimation model.(2)Research and optimization of yak image foreground recognition algorithm under complex background.In the natural standing posture of yak,using machine vision to extract the yak foreground,there are problems such as low foreground recognition rate.This paper proposes a method for yak foreground extraction under complex backgrounds,and designs a yak foreground extraction algorithm based on SLIC(Simple Linear Iterativeclustering)and Sobel edge detection operator under complex backgrounds.This method realizes the foreground recognition of yak images under complex background.Experimental results show that the accuracy of the algorithm in identifying yak images is 95.15%.(3)Research on recognition algorithm and optimization of yak foreground image measurement points.For the yak in the natural standing posture,using machine vision to extract the yak body measurement points,there are problems such as a low recognition rate of the points.This paper proposes a method for extracting yak body measurement points under complex backgrounds,and designs algorithms for yak individual extraction under complex backgrounds,and yak body measurement point extraction based on curvature analysis.This method realizes the coordinate extraction of yak body height,body oblique length,chest depth,tube circumference,cross section height and other 10 individual ruler measurement points.The experimental results show that the body size measurement point extraction algorithm is stable,has strong robustness and stability,and the recognition accuracy of the algorithm is 96.94%.(4)Estimation of yak body size in the image.By performing a proportional analysis on each yak's body size,a data mining algorithm is used to obtain the yak's body size parameters,and the yak's weight in the image is calculated by the yak weight estimation model.From the test results,the method of using data mining to estimate the yak body size parameters and using the yak weight estimation model to calculate the weight of the yak is more accurate.It is feasible to use machine vision to estimate the yak body size and weight.