Decomposing Of Cows'body Parts Based On Skeleton Feature

Large-scale farming,which need a higher demand on the feeding and management is a development trend of our country's aquaculture.Through the application of information technology to improve the efficiency of farming and health management level is becoming an inevitable trend.Video analysis technology can automatically monitor and understand the behavior of animals,It is an important means to improve the level of information of feeding management,and more and more used in the Intelligent feeding of cow.Domestic and foreign scholars have done a lot of research work in this area.But most studies are conducted for the whole cow to analyze.As a multi-joint large animals,cows have head,neck,forelimbs,hind legs and tail divided through the joint limbs.Through the body parts of cow can get more accurate details of the movement of cows,it is the basis of cow posture detection,behavioral analysis and understanding.In order to realize the accurate decomposition of the head,neck,trunk,forelimb,hind limbs and tail of dairy cow,this paper proposes a body decomposition method of cow based on skeleton.Using the Kinect to obtain the depth image data,the paper studies the extraction method of cow based on depth image,the extraction method of cow skeleton and the body decomposition of cow based on skeleton.The major work and conclusions of this thesis are as follows:(1)A method which used depth threshold,image morphology and median filter to extract cow's target is proposed.Because the cow in the farm with the complex background and the impact of light are difficult to accurately extract.So considering the layout environment of equipment,cost-effective and other factors,the Kinect was used to get the depth image data of the cow.First,depth data from Kinect was converted to text data of txt in order to facilitate deal with the data.The target cow was effectively extracted from the complex background by using the depth threshold segmentation and image morphological transformation,and the median filter to denoise the image.The experimental results show that the overlap rate of cows extracted by this this paper and artificial is 95.62%.(2)The cow skeleton was extracted based on the skeleton point judgment criterion under Choi skeleton width constraint theory,then less processing branch.Through the analysis of the main skeleton extraction methods and the connectivity,single pixel,high efficiency and other factors of skeleton used skeleton point judgment criterion under skeleton width constraint theory of Choi to extract the skeleton.The discrete curve evolution model pruned the skeleton to obtain a simplified skeleton.This skeleton can reflect the complete contour characteristics of the cow.(3)A method to decompose the cows based on skeletal characteristics is proposed.The bifurcation point of the skeleton with important position information was extracted from the skeleton.The candidate segmentation line was generated by the bifurcation point of the skeleton according to the constraints set in the paper.Using the shape visual significance and the Split line priority to optimize the candidate segmentation line realized the body decomposition of the cow.The result show that when the significance threshold is 2.5,the average correct rate is 95.09% and the correct rate of tail,which is most difficult to segment,is 95.51%.Using the method of paper to decompose the cow different body posture,such as upward,normal walking,micro bow,and bow get the average correct rate,which is 95.18%?95.00%,94.85% and 96.23% respectively,achieving the high precision body decomposition of cow of different body.