Though GPS collar has been widely used to study the activity rhythm of giant panda (Ailuropoda melanoleuca), the lack of researching on the reclassified value of the collar data leads to a deviation result. To determine the threshold, we carried out the research on the two giant panda (Table 1) in Chengdu Research Base of Giant Panda Breeding Dujiangyan Field Research Center for Giant Pandas. Based on the strategy of maximum positive rate. We compared the behavior data obtained from the video monitor (Table 2) with the motion sensor data of GPS collars in the period of May 23rd, 2015 to June 10th, 2015. The results showed there was no significant difference between rest motion index and active motion index of the two giant panda (Fig. 1). The threshold was determined as 32 while 0 was used as the threshold in prior researches. The percentage of correctly classified rest behavior and active behavior was 98.23% and 94.48%, respectively. Whereas, based on the threshold of 0, the percentage of correctly classified rest behavior and active behavior was 100% and 77.34%, respectively (Fig. 2). We identified the resting and activity behaviors of the two giant pandas in 19 days based on the value of 32 and 0. The result showed that the daily average activity rate of giant pandas based on the threshold of 0 (“Qian Qian” was 59%, “He Sheng” was 70%) was significantly higher than the daily average activity rate based on the threshold of 32 (“Qian Qian” was 54%, “He Sheng” was 50%) (Fig. 3). It suggested that the activity rates of giant pandas could over estimated based on the value of 0.