Abstract: In order to evaluate the abnormal situation of pigs comprehensively, the grey-evidential fusion evaluation model combined with DS evidence theory was adopted in this experiment to evaluate the porcine abnormalities based on the abnormal probabilities of pigs detected by three monitoring methods including machine vision, sound recognition and ultrasonic technology. Aiming at the problem that original DS evidence theory could not effectively integrate high conflict evidence, an improved DS evidence theory was proposed. The rules of evidence combination were modified by introducing the reasonable distribution conflict of evidence credibility. The results of abnormal behavior fusion evaluation of pigs under multi-source and multi-period were analyzed by concrete examples. The results showed that the evaluation model proposed in this experiment assigned a value of 0.822 3 to the normal combination basic probability under the condition of low conflict evidence. In the case of high conflict evidence, the combination basic probability of first-order anomalies was assigned to 0.649 7, which could accurately evaluate the anomaly level of pigs. The results indicated that the multi-source and multi-period grey evidence fusion evaluation model was reasonable and reliable for the evaluation of pigs’ abnormal behaviors, which could provide the feeders with effective abnormal information of pigs and improve the detection efficiency.