Abstract: In order to improve the driving safety factor of tractor drivers and improve the problem that tractor drivers often turn their heads to watch the operation of rear farm machinery and tools, the use of HUD in tractor cab was studied and analyzed. Firstly, the images of tractor drivers observing the rear agricultural machinery and tools were collected and the important information was marked by the combination of DeepLabv3+ image semantic segmentation method and eye movement tracking technology. In the experimental part, Pytorch deep learning framework is used to build the required model network structure and complete the corresponding code development, and finally get a relatively clear segmentation image. Then the original image and the segmented image are verified by eye movement tracking experiment, and the eye movement track of the subject is extracted. After watching, it is found that the segmented image can be better recognized by the viewer. The results show that using DeepLabV3+ image segmentation enhancement, applied to the head-up display technology can effectively improve the safety and efficiency of tractor drivers.