Abstract: In view of the lack of comprehensive management and precise feeding of individual pigs in the shelter pig house on the current market, a multi-sensor based pig feeding behavior system for the shelter pig house is designed. First of all, The pig files are established based on RFID technology to identify the identity information of the pigs, then the pig movement discrimination algorithm based on the improved adaptive Gaussian mixture model is used to judge whether the pigs enter the feeding areas, finally, the feeding data of pigs are obtained by weighing sensors Through the above-mentioned system, the identity, feeding time, feed intake and feed intake amount of the feeding pigs can be accurately locked. In order to verify the feasibility of the system, a four-factor and three-level experiment is designed to explore the effects of reader power, tag approach speed, radio frequency, and tag approach direction on the RFID recognition effect. The test results show that the reader power is 27 dBm, the tag approach speed is slow, the radio frequency is 920.75 MHz, and the tag approach direction is horizontal, which will make the electronic tag identify the best within the range of 0.5 m from the antenna. At the same time, a pig individual feeding test is carried out, and the test shows that the error of the system is 0.08%-0.1%, which meets the requirements of the pig farm.