Pig Behavior Tracking And Recognition System Research And Achievement Based On Embedded System

The process of agricultural science and technology is speeding up,and the demand for automation of modern agricultural equipment is increasing.After summarizing the study of intensive pig raising at home and abroad,proposes a pig behavior tracking and recognition device based on embedded,use multi target tracking technology for pig position determination,automatic recognition and detection of abnormal behavior of pigs,give an alarm of pigs abnormal behavior in time,provide a technical scheme for the control of the epidemic situation in the intensive farming area.In this paper,the software and hardware platform of the system is designed with ARM framework,and the improved particle filter algorithm is used as the target tracking recognition algorithm.The method of combination of color feature and gradient feature is used to extract the feature of the target,so as to solve the accuracy of the algorithm tracking.The Mean Shift method is combined with the particle filter algorithm to improve the real-time performance of multi target tracking.The state transfer model of the system,the multi feature observation model and the correlation strategy of the particle state and the observation value are established.According to the tracked position of pigs and combined with gravity sensor data to determine whether the pigs eating,drinking and excretion or not.And record eating,drinking time and weight of pigs,and time and number of excretion.By comparison with the standard value of normal pigs to determine whether exist abnormal behavior or not.And alarm the abnormal behavior in time.The video images of three pigs are used to verify the performance recognition performance of the algorithm and of system.Firstly,with different particle number for variables to analysis the accuracy of algorithm before and after improvement.Analysis of timeliness of algorithms,then the performance of the algorithm is analyzed in the case of adherence and occlusion between the tracking targets.Finally,validate the accuracy of identification of pig behavior.The experiment proves that the improved algorithm can keep tracking performance under the greatly reducing the quantity requirement of particles of multiple target tracking algorithm,and the algorithm can meet the real-time requirements of tracking and identification of pig behavior,and abnormal behavior can be identified accurately.It has great practical application prospect.Not only to overcome the waste of time and manpower in traditional human observation and the effect of human subjective factors,but also to solve the stress response of traditional earcon recognition,provide a new method for the general behavior of pigs.