Abstract: In order to perform the studies on the precision feeding and behavioral monitoring of dairy cows, an automatic feeder of dairy cows was designed, which accomplished the functions of cows automatic identification, automatic feeding data (feed intake time and amount) acquisition and data analysis simultaneously. The automatic feeder was composed of mechanical device system, electric identification system, weighing system, central control system, live data collection and storage system, and remote feeding data extraction and analysis system. The mechanical device system was constituted of feeding bin, brackets, railing and blocking apron. The electric identification system included reading antenna and pneumatic switch for discharging. The weighting system was made up of brackets and embedded weight sensor (L6G, technical parameters: Weighing range ≤ 200 kg, error less than 0.002 kg). Central control system was composed of microprocessor (LPC1766, technical parameters: Operating temperature of from -40 to 105 ℃; operating voltage of 2.0-3.6 V; flash memory of 256 K, low power consumption), watchdog reset circuit, card reader circuit, weighing data collection circuit, data communication circuit, data transceiver circuit, and external regulator circuit. The reader circuit adopted multi-channel R232 interface and chips (Model: MAX232E), and low-pass filter and 24 bit conversion chip (Model: ADS1232, Dezhou) were used in the weighing data collection circuit. For transceiver circuit, according to the standard ISO 11898, the universal CAN (controller area network) transceiver chip (Model: CTM8251A) with isolation function was adopted, which had 110 nodes at the most and the transmission rate increased to 1 M/s. The ferroelectric memory and serial peripheral interface were adopted in the circuit of data caching system, and the cable data transmission rate could reach 15 MB/s. The live data collection and storage system received signals from the central control system in each feeder, the preset record number in storage system could reach 14 000, and the form of stack data was applied in system. The feeding data could be managed and analyzed in real time by data process system in PC (personal computer) terminal. The feeding experiment showed that the cognition rate for low frequency RFID (radio frequency identification) (134 kHz) ear tag by the automatic feeder reached 100%, the range of weighing was 0.01-200 kg, the precise was 10 g, and the weighting error was below 0.15%, which could meet the requirement of cows' precise feeding intake record. The performance test of control system showed that individual cow's feeding behaviors, including feeding frequency, intake time, and feed intake, were different significantly (P<0.05). The average feeding frequency was 10-13 times per day, and the average intake time was 5.38 h per day, which were consistent with cow feeding characteristics. However, the deviation between average daily feed intake and predicted intake value by NRC (National Research Council) model was -4.76%-7.83%, which may be caused by the low applicability of NRC model. In conclusion, the automatic feeder developed in our study can meet the requirement of precise feeding in cows' production, and supply an online and intelligent data automatic record and analysis platform for cow feeding behavior research.