Research On Identification Method Of Sow Body Size Based On TOF Depth Data

The sow’s growth status,body size traits,body traits and reproductive traits are closely related and changes in body size parameters directly reflect the growth status of the sow.Changes in body size parameters of pregnant sows play a key role in the number of litters,the weight of piglets at birth,and the weight of weaned piglets.Direct contact measurement of sows using conventional methods of obtaining body size parameters can cause stress to the sows and affect their healthy growth.With monocular vision technology that the three-dimensional body size parameters of sow pigs cannot be obtained and use the multi-vision vision technology,the body size parameters of the sow can be obtained without stress.However,there are restrictions on conditions such as camera erection,lighting for acquiring data and factors such as low efficiency of 3D(3 Dimensional)data reconstruction will affect the execution efficiency of sow body size acquisition.Using a depth camera to obtain 3D data on the body surface of the sow and to explore the key technology of deep data processing,can lay a good foundation for fast,continuous,real-time access to sow body size information.This article takes Rongchang pregnant sows and Landrace pregnant sows as research objects and use the TOF(Time of Flight)depth camera to obtain 3D point cloud data from the back of the sow pig.After data preprocessing,the outline and envelope of the projection data of the back of the sow are obtained and the algorithms to extract body length,height,chest width and bust are designed.The recognition algorithm of the body length is that based on the envelope,the two-dimensional coordinates of the ear root and the tail root of the sow are identified.Then,the three-dimensional coordinates of the points on the dorsal ridge between the midpoint of the ear root and the midpoint of the tail root are identified,and the distance between the adjacent two points is accumulated to obtain the body length;The recognition algorithm of height is that selecting the midpoint of the widest point of the pig body armor as the recognition point of the body height,identifying the height measurement identification point in the three-dimensional depth data and calculating the distance between the Z coordinate and the ground data,recorded as height;The method of recognizing the chest width is that the identification point of the chest width position is a method of utilizing the proportion of the pig body and take the proportion of this position in the pig body.Through a lot of data calculation and experimental verification,the chest width recognition position is at the 1/6 position between the center of gravity of the data and the midpoint of the ear root and calculate the distance within the Y-axis range,which is the chest width;The bust depth data identification algorithm is that projecting the acquired chest width data into the YOZ plane and insert a point on the underside of the measurement position of the pig chest by interpolation method.Analyze each point mathematically which fit the bust outline with a fan shape and an ellipse to obtain its bust.The algorithm was verified by 30 sets of data from Rongchang sow and Landrace sow and the results shows that the average relative errors of body length,height,bust and chest width of Landrace sows were 2.94%,2.58%,1.49%,4.12%,respectively.The average relative errors of body length,height,bust and chest width of Rongchang sows were 4.99%,4.39%,2.18%,and 6.93%,respectively.Extracting the sow’s body size parameters with depth data and the algorithm can be applied to the measurement of body size parameters of different sows of Landrace sow and Rongchang sow with high accuracy and strong robustness.Using this algorithm to process the depth data reduces the complexity of the calculation and improves the execution efficiency of the algorithm.In this paper,the non-stress method is used to analyze the point cloud data of the breeding pigs and obtain the body size parameters and mastering the real-time growth status of pregnant sows without affecting their normal growth provides a new method for sow welfare farming.