Study On Digital Sows Body Condition Scoring Method Based On Image Recognition Technology

The reproductive capacity of sows is not only related to the economic benefit of pig farms,but also affects the survival of pig farms.As an indirect indicator of sows’ reproductive ability,body condition score is the best way to evaluate the reproductive ability of sows in recent decades,at present,the scoring method of sows’ body condition is mainly based on artificial observation of buttock shape or measurement of Backfat,which has the disadvantages of large error and low efficiency,and can not meet the actual production demand.Therefore,this paper presents a method of digitalizing sows’ body condition score,using digital image processing technology and statistical means to explore the method of digitalizing score from the characteristics of sows’ buttocks,the main work and results are as follows:(1)In the standardized facilities in the pig farm to start the data acquisition required for this study.The images of sows were collected by smart phone from preparturient and weaned landrace sows.Measurement of p 2 point backfat in sows by type A ultrasonic apparatus.The total litter size,live litter size,stillbirth number,litter weight and average birth weight of each experimental sow were obtained directly on the Internet.(2)Extracting the buttock region of sow image.Aiming at the high resolution and high noise of sows image taken by smart phone,this paper uses U-Net network model to segment sows image semantically,and improves the U-Net coding block and Model Evaluation Index.The experimental results show that this model can locate the information of sow’s buttock region automatically and segment the buttock region accurately,and the effect is better than other traditional image segmentation methods.(3)For the semantic segmentation of the image of sow buttocks,combined with the content observed by visual scoring method,several contour feature extraction algorithms and self-design Algorithms provided in OPENCV are used to extract the features of sow buttock:roundness,ovality,eccentricity of ellipse,rectangle,radius of curvature,spherical shape and leaf shape.SPSS24.0 was used to describe the parameters of various morphological features,principal component analysis and cluster analysis,three parameters,roundness,leaf shape and radius of curvature,were selected to represent the shape characteristics of sow buttocks.(4)Body condition scoring method and result evaluation.The reproductive performance of sows with different body conditions at the breeding stage was analyzed by one-way ANOVA.The ability of the body condition scoring method established under roundness,leafiness,radius of curvature and Backfat index to distinguish the reproductive performance was compared.The results showed that the roundness body condition scoring method had the best discrimination effect and was more consistent with the results of the standardized backfat scoring method.The correlation analysis and linear fitting between sow body condition score and back fat in breeding stage were carried out.The results showed that the roundness scoring method could well distinguish the back fat value under different body condition scores,and finally put forward the scoring standard of sow body condition hip roundness.The results show that the method can effectively meet the concept requirements of body condition score and the practical production needs of sows,and realize the method of evaluating the body condition of sows from the body appearance characteristics of sows.The method presented in this paper is more intuitive,clear and understandable than the previous scoring method,and solves the problems of subjective and contact scoring in the current scoring process,it improves the efficiency of grading and protects the welfare of animals,and provides reference for the accurate management of breeding sows and the adjustment of feeding plan in pig farm.