Abstract: Abstract: When agricultural greenhouse collects temperature and humidity data, cable network transmit data lead to wiring complexity and difficulties of moving nodes. Zigbee wireless sensor network transmit data can effectively solve the above problems. Zigbee has the virtue of ad hoc network, low power consumption, equipment layout flexibility, good stability and so on. The core components in greenhouse are consisted of terminal nodes, route nodes and coordinator node. Terminal nodes placed in greenhouse continuously send out request signals to join the network. All the terminal nodes receiving a response signal from the coordinator node can join the network. Terminal nodes collect temperature and humidity data and transmit these data to coordinator node. Coordinator node is the core of Zigbee and is placed in the geometric center of the greenhouse. It is responsible for the establishment of network and can communicate with computer in the control center. Coordinator node sends commands to terminal nodes and sends temperature and humidity data monitored by terminal nodes to the control center. Affected by unfavorable factors such as electromagnetic interference and transmission channel congestion, the collected data might include error data. This paper applied an improved Chauvenet's Criterion to eliminate the error data in greenhouse data collection system software. Traditional Chauvenet's Criterion eliminates one error data per cycle of operation, which leads to low convergence rate. Amount of error data cost a long time to process error data, which result in that data information cannot be obtained in a timely manner. Improved Chauvenet's Criterion is designed to solve this problem by applying the interquartile deviation method before data began cycle operation. According to the distribution of data, the data is sorted from smallest to largest so that improved Chauvenet's Criterion can find the upper quartile, median and lower quartile quickly. Quantile is used to determine a numerical interval. Large deviation error data will be eliminated. Improved Chauvenet's Criterion reduce the number of data participated in each cycle operation in order to save computer memory. Using the standard deviation to determine condition, each cycle of operation can eliminate multiple error data. This can improve data convergence rate, save processing time and meet requirements for real-time tracking data in greenhouse. Comparative analysis of the time complexity of traditional Chauvenet's Criterion and improved Chauvenet's Criterion, it fully proved that the improved Chauvenet's Criterion saves computing time. Greenhouse environment data collection system was tested in the Northeast Agricultural University Horticulture Station. The tomato plants were growing in order, greenhouse is convenient for arrangement of monitoring points. Monitoring nodes extracted a data per second. A group of 32 data was transmitted to the data processing software. From the system sending out commands to system display data, traditional Chauvenet's Criterion use 40s, improved Chauvenet's Criterion use 35s, and improved Chauvenet's Criterion saves 5s. It was found that data distribution area where data is processed by the improved Chauvenet's Criterion is more concentrated. And it is convenient for system and experts to make a decision and take the necessary control measures according to data distribution.