Identification And Measurement Of Signs Of Live Animals Without Contact

This paper takes live pigs as the representative research object,identifies the individual identity information of live pigs,measures the physical sign information of live pigs accurately,and correlates the identity information and physical sign information of live pigs effectively,so as to realize the refined,intelligent and efficient management of the physiological health and growth of live pigs.The measurement of physical signs of live animals can provide efficient and high-precision individual animal information support for the tracking of their growth trend and the scientific adjustment of breeding programs.Firstly,the cameras were used to automatically capture and scan each pig,and the individual identity photos of each pig were obtained and the identity information of each pig was manually marked,so as to establish the pig identity authentication information database.A video camera was used to capture images of pigs during their growth.The multi-object image segmentation method proposed in this paper was used to obtain individual pig images.In this paper,the convolutional neural network algorithm was used to recognize the characteristics of individual pig images by using the pre-established pig identity authentication information database as the sample database,and then the individual pig identity information was determined to realize the accurate identification of pig identity in the breeding industry.Secondly,two two-dimensional images(overall top view image and side image)of the pig at a certain moment were collected by two wide-angle cameras.The multi-objective image segmentation method proposed in this paper was used to obtain individual pig images from different angles,and the edge detection algorithm was used to extract the back contour of the pig.The pig body size feature points were detected and coordinates were extracted from the contour of individual pig images with Open CV.Finally,the body sign parameters(body height,body length and body width)of pigs were calculated by parameter calibration and body size calculation formula.Finally,the accuracy and effectiveness of the pig identification and physical sign measurement algorithm were verified.Based on experienced zoology experts,the identification of individual pigs in the sample set to be identified was used as the real signs.The image of the sample set to be identified was taken as the verification sample set.The identity of the verification sample set can be speculated by the model.The prediction accuracy of the pig identification algorithm was quantitatively analyzed by using the AUC(area under curve)of the ROC curve with the real value and predicted value.The physical sign parameters of the target pigs were measured by manual measurement method,and the accuracy and robustness of the proposed pig physical sign measurement algorithm were analyzed by comparing the physical sign parameters measured by manual measurement with those calculated by algorithm.Results show that the prediction accuracy of pig identity identification model is ROC＿AUC: 0.951,which indicates that the pig identity identification model constructed in this study has high identification prediction accuracy.In terms of the measurement of physical sign parameters,the average relative deviation of body length,body width and body height calculated by the measurement method proposed in this study were2.63%,6.63% and 3.27% respectively,indicating that the accuracy of measurement method of physical sign parameters based on this study is high.