The body surface temperature of pigs is an objective reflection of their own physiological conditions. Testing the body surface temperature of individuals and groups is an important way to achieve fine and efficient production of pigs. In order to realize the online monitoring of the body surface temperature of pigs in a top view, a method of detecting and tracking the head and trunk of pigs based on thermal infrared video was proposed. Firstly, the pig channel in the collected thermal infrared frame was intercepted, and the overall skeleton of the pig was extracted after pretreatment in this area. Then the key points at the front end of the skeleton were scanned to detect the head skeleton. After that, the trunk detection was realized by calculating the position of the key point of the torso tracking frame based on the position of the head and the spatial position tracking frame of the body. Finally, head and body detection were performed on each frame to achieve head and torso tracking. Using the collected 50 pig videos, the proposed algorithm was tested on the Matlab R2014a platform and compared with the compressive tracking (CT), kernel correlation filter (KCF) and fast discriminative scale space tracking (FDSST). The results showed that the tracking precision was 0.6752 (threshold value was 20 pixels), which were 9.41, 7.09 and 2.72 percentage points higher than those of CT, KCF and FDSST, respectively. The proposed algorithm can effectively solve the problem of automatic detection and tracking of the head and trunk of the thermal infrared video of the pig in the top view, and can provide more accurate regional information for the body temperature extraction of the head and the trunk. The average tracking frame rate was 31.63f/s, which can meet the requirements of online monitoring of farms. This technology provided technical support for further intelligent monitoring equipment for postgraduate pig body surface temperatures.