| There were some very important indicators to evaluate the growth status of pigs and its breeding. They were the body size, weight and shape. Body size, body type and other growth parameters are generally obtained by manual direct-contact measurement in the traditional way. This approach is not only a heavy workload, but also for porcine stress, leading to reduced pig welfare. For the real problem, in this paper, we use live pigs and pig specimens as the object of study. The use of networking technology, mathematical and statistical modeling techniques, machine vision, image processing technology, big data and visualization technology, we studied the method to extract the pig body size, body size and other parameters and to reconstruct the pig’s three-dimensional. The main research contents and results are as follows:(1) Study on the method for estimation model of pig body weight.There is collinearity between the independent variables for estimation of body weight, such as body length, width, height, hip width, hip height, chest, back fat thickness, loin eye area and so on. We carried out correlation analysis, and then constructed the pig body mass estimation models by using the methods such as canonical correlation analysis, partial least squares and artificial neural networks. We found that the body weight estimation model was very good to predict the body weight of Landrace based on RBF neural network, its goodness of fit R2=0.977, the relative error is1.34%, forecast better than the linear regression model. We builded the pig body weight linear prediction model based on PLS, R2is0.945, the relative error is2.7%, and has a strong operationa.Used RBF neural network and PLS to build model, was an effective method to build the pig body weight estimation model. This method eliminates the linear regression analysis of the original independent variables collinearity.(2) Study on pigs body measurements point extraction algorithms and its optimization in a complex context based on computer vision.For pig’s natural standing posture, the use of machine vision to extract pig body size measuring point, there is a lower rate measuring point identification problem. In this paper, a new method is proposed which can be applied to calculate pig body width and body length with no stress, and further to estimate the pig body weight. First, In light of the site conditions of pig house, the background information is removed via arithmetic operation based on gray process of the pig body image and the background image. And image noise is removed by the median filtering method, and detailed porcine somatic information is obtained. Then, the pig body image’s segmentation threshold is determined using dynamic threshold method, and the binary image is acquired; After calculating the number and the area of connected regions, the other pigs which may exist in the imaging area get removed via maximum ordinate-area method. Finally, the individual pig contour is extracted through the Canny edge detection algorithm. Because of the interference the body size extraction, the identification algorithm aiming at removing head and tail regions in the image is designed, and the data envelopment analysis is used in the algorithm, based on the distance between the body contour and the envelope line, the pig body contour with the head and tail removed is finally obtained. The algorithm stability and the extraction accuracy of the verification were tested using9different pig images. The result shows that the detection accuracy of the pig body size is high, and the average relative error of the detection values and measured values is2.26%; the average relative errors of weight is5.11%. Overall, the algorithm has a better detection effect for pig body size, and it is stable with better robustness.(3) Study on the method for three-dimensional reconstruction of pig body based on binocular stereo vision research.For two-dimensional images of pigs, how to reconstruct three-dimensional model in pigs, pig body to extract three-dimensional information such problems, this paper carried out a parallel binocular stereo vision based on the three-dimensional reconstruction method pigs, raised pigs under complex background three-dimensional reconstruction method. This method is right around the camera calibration, the intrinsic and extrinsic parameters for the camera; remove pigs background information, to improve the efficiency of stereo matching gray template; based on triangulated irregular network, reconstruction pig surface information for visual analysis. Experimental results surface:stereo matching efficiency is doubled; through machine vision to build three-dimensional model body size detection accuracy is2.09%. And weight detection accuracy is3.38%. Proved by experiments by binocular stereo vision, three-dimensional reconstruction of the pig, extracting body size, body type and other parameters, the method is feasible and effective.(4) Study on the method of Active3D reconstruction for pigTo verify the three-dimensional reconstruction method based on binocular vision body size extraction accuracy, we propose a three-dimensional reconstruction of the pig and the shape parameter extraction method based on three-dimensional laser scanner. In this paper, we got the point cloud data of pig through the three-dimensional laser scanner. By Polygon Editing Tool Vel.2.40software, carried on the point cloud data preprocessing, and then based on triangulated irregular network, the three-dimensional surface model of pig body was reconstructed. Then the parameters were extracted for pig’s body length, body width, hip width, height, hip height, chest measurement, body surface area, volume and so on. The results indicate that the272,021pigs point data was acquired by three-dimensional laser scanner, and reconstructed pig’s three-dimensional surface by544,042polygons; By Error test method, compared to the detection values by the three-dimensional model and measured values of the body parameter. The regression analysis showed that there is high detection precision in its body size. The results showed that the laser scanning three-dimensional reconstruction of body size detection accuracy is higher, its relative error is only0.17%; the weight average relative error is2.7%. while the three-dimensional reconstruction of weight binocular vision detection accuracy is slightly lower, the relative error is3.38%. Relative to the laser three-dimensional reconstruction model, based on binocular stereo vision technology and three-dimensional reconstruction of the pig weight parameter extraction relative error is2.7%. Through binocular stereo vision technology to build three-dimensional pig, pig body size can be extracted parameters for the prediction model to provide data to support the weight. |
PDF Full Text Request