Study On Sow Body Size And Weight Detection And Evaluation System

Based on the relevant technical problems in modern breeding sows,this study aims to solve the problem of non-contact measurement of sows related body size indicators during sow growth and development,to estimate the difficulty of sow body weight parameters,and to evaluate the performance of gilts.Wait.According to the physiological characteristics of sows,in view of the strong herd behavior of sows,the deployment of cameras for sow body position and image collection is limited by feeding machines;the sow weight estimation model and the prediction of the sows are constructed by the sow manual measurement of body size data.Model;using machine vision technology,image acquisition technology,image analysis technology and other methods to achieve non-contact acquisition of sow body length,body width,circumference and other seven body size data and use the construction model to estimate.The main contents and conclusions are as follows:(1)Construct a sow weight estimation model and a prediction model for the circumference.Through statistical analysis,the weight of the sow has a significant correlation with some of its body size.This paper first uses OLS and SMLR to construct a sow weight estimation model.Through analysis,although the two models have a high degree of fit,there is a serious collinearity problem that cannot be eliminated.For this reason,the PLS method is used to construct the sow weight estimation model,and the corresponding R2 is 0.945.This model eliminates linear regression.The independent variable collinearity problem in the analysis is a more effective estimation model for the weight of the sow.At the same time,the sow perinatal development prediction model was constructed by artificially collecting the data of pregnant sows,which provided a theoretical basis for enriching sow body size research.(2)Image acquisition and calibration of the camera.In order to eliminate the framing error caused by the sow’s deviation from the framing center axis,the data is complemented by the left and right cameras.In order to obtain the sow circumference,this article sets up a rear view camera to obtain the sow rear view image.Finally,the internal and external parameters of the camera are calibrated by using the camera calibration method based on 2D planar targets and the camera calibration method based on the same ratio.(3)Sow image extraction.In order to obtain accurate sow body size parameters,the quality and integrity of the sow image must be guaranteed.In this paper,the image of the collected sow video is pre-processed to reduce the image noise to ensure the quality of foreground extraction.The sow image and background are segmented by the PBAS foreground extraction method.The extracted sow images were processed through a series of morphological processes,smoothing the edges to optimize the sow image,and laying the foundation for sow contour extraction.(4)Body size acquisition.Aiming at the physiological structure characteristics of sows,this paper designed a sow body size extraction algorithm.Including: sow head and tail removal method,body length acquisition method based on sow spine line,body height based on sow back feature point,hip height acquisition method,geometric primitive based segmentation and fitting method for obtaining sow circumference.These methods have achieved seven body sizes such as sow body length and body width.(5)Accuracy test and fitting correction.By comparing the sow body size,body weight manual measurements and system estimates.The relative error between the two is analyzed,and the average relative error of the eight parameters of the sow body size and body weight is less than 3.30%.From the comparison results,the detection algorithm of the sow weight detection system is better.In order to improve the accuracy of the system,the error correction of the two parameters with relatively high error of circumference and weight is made.By correcting the sow weight detection system,the estimation results of sow body size and body weight parameters are more accurate,and the system is more robust.This paper achieves the non-contact sow body size,body weight detection,and the prediction of sow circumference development.It provides a model basis for animal body shape index calculation and body shape linear assessment,and lays a foundation for body behavior research based on body state.