Study On Visual Detection Method For Weight Assessment Of Layer Chicken

In recent years,intelligent breeding has drawn wide attention of the research field of poultry breeding.It is well known that body weight makes a big difference on the performance of chicken meat and egg production,which is one of the important indicators reflecting the situation of raising chicken.This paper,taking green shell laying hens as the research subject,examined the establishment of the body weight evaluation system,the segmentation algorithm of the background image of laying hens,the separation algorithm of attached hens,the feature optimization algorithm,the establishment of the body weight estimation model and the compilation of the body weight evaluation software of laying hens in the captive environment.Details are as follows:(1)A visual detection system was established for weight assessment of laying hens.It mainly selects the type of surveillance camera,calibrates the camera,corrects the image and sets up the detection system.(2)The method of image segmentation based on texture features is put forward.By calculating the gray level co-occurrence matrix under the image mask,it focused on four texture features-energy,entropy,contrast and correlation.To realize the background segmentation of laying hens,it adopted the segmentation threshold of target and background after analyzing the texture features.Then by dealing with 200 images of laying hens with different feather colors and light intensity,it concluded that the average segmentation error of this method is 3.6%and the maximum segmentation error is 8.7%.(3)A new method based on concave point analysis was proposed to segment the chicken body with adhesions.By analyzing the convex defect contour of the adhesive area,the concave point position was determined by using a square template,and then the correct concave point matching method was determined by random matching of the concave points,and finally the segmentation of the adhesive chicken body was achieved.The results concluded that the average segmentation accuracy was 92.8%and the average running time was 2.817s.(4)Seven image were screened out,featuring with the following characters day age,projected area,perimeter,maximum radius of inner tangent circle,minimum length and width of outer rectangle,as well as long and short axis of ellipse fitting.Aiming at the influence of the irregular swing of the head and tail on the image feature stability of layers with different shapes,an algorithm based on ellipse fitting was proposed to remove the head and tail of layers.Firstly,the contour of laying hens was elliptically fitted,and then the vertices of the head and tail of laying hens were finalized by calculating the distance between the contour pixels and the boundary of the ellipse.Finally,mask elimination was used to remove the head and tail of laying hens.The results show that the relative range of the image features is reduced by at least14.41%,which indicates that the proposed algorithm can effectively improve the stability of the image features under different shapes of laying hens.(5)To solve the problem of collinearity between features,a prediction model of laying hens'body weight based on ridge regression analysis was proposed.For the prediction of the average weight of the single laying hens,it showed that the determination coefficient R~2was0.964,the mean relative error was 3.56%,and the mean absolute error was 24.35g,while for the prediction of the average weight of multiple laying hens,the average deviation was 4.13%and the maximum deviation was 7.9%.(6)The software of weight assessment system for laying hens based on Matlab was written.The software fulfilled the process of individual recognition,feature extraction and weight prediction of laying hens.