Research On3D Acceleration Sensor Recognition For Goats’ Behavior Recognition

In recent years, the healthy culture, safe production, quality control andquality traceability requirements of goat herd become more stringent with thedevelopment of information technology in China. However, the judgment andmonitor of behavior characteristics of the goat herd still have to rely on intuitionand experience of the farm yard worker. This measure not only waste a lot ofmanpower, but also is inefficient. It is no longer suitable for large-scale,intensive modern livestock industry.In order to judge the relationship between animal behavior and diseasediagnosis in the most fast and accurate way, the researchers introduced thebehavior characteristics recognition technology such as the machine visiontechnology, the video or image processing technology, the acoustic detectiontechnology, and so on. But recording animal especially the free grazing livestockanimal behavior data continuously must rely on the sensor informationacquisition and processing technology. For this, the key of this paper is toresearch and recognize goats’ typical daily behavior recognition on3Dacceleration sensor, the main contents include:(1)The goat’s daily behavior data were collected and analyzed targetedly.This paper used the method of the K-means clustering algorithm or SVMalgorithm to iterative training the3D acceleration data sets of the goats throughanalyzing and comparing of the effect of these algorithms for accuracy of theclassification model. At the last time, this paper used the video monitorinstallation and the time of the action to identify the goats’typical daily behavior corresponding to the acceleration data model. The research results showed thatthe accuracy of K-means clustering algorithm on typical daily behavior (thelying, standing or walking, eating, and jumping) is more than87.76%, and therate increased nearly6percent points through the optimization of SVMalgorithm. The research indicates that the result of recognition and classificationof the goats’ typical daily behavior characteristics can achieve higherclassification accuracy based on this algorithm. This algorithm has highersearching efficiency and better application value in the field of behaviorrecognition.(2)This paper concluded the best location of3D acceleration sensor byanalyzing the typical daily behaviors of goats and the relationship between thedeployment location of3D acceleration sensor and the experiment results.(3)This paper also studied recording intervals time of the3D accelerationsensor influence on the classification accuracy and data processing. The resultsshow that this method reduces the amount of data processing and improves thespeed of data processing without affecting the accuracy of classification whenrecording intervals time of the3D acceleration sensor was selected as2s.(4)This paper established the acceleration data model of the goats’ typicaldaily behaviors so as to judge the goats’ abnormal behaviors and improve animalwelfare management of the goats’ production system. This data model providedthe basis for creating Goat Disease Prediction Model and judging therelationship between the typical goats’ daily behavior and disease types.