Abstract: Due to the demand for safe obstacle avoidance in the autonomous navigation of Mars rover in complex terrain and the double constraints of computational resources and energy supply of the onboard platform, a lightweight detection model, YOLOv8-LMD, was constructed, aiming at realizing the requirements of high precision and lightweight characteristics of the rock detection algorithm on the surface of Mars. First, the lightweight backbone network was reconstructed based on the HGNetv2 architecture to realize the preliminary compression of model parameters. Secondly, a multi-scale feature fusion network structure was designed, and the neck network was reconstructed by integrating slim-neck and ASF-YOLO to effectively improve the feature characterization of rock targets at different scales. In addition, a lightweight detection head was designed by using the convolutional sharing strategy, which reduced the computational complexity and enhanced the classification and localization accuracy at the same time. Finally, a pruning algorithm was used to prune the model parameter redundancy to further compress the model, and the knowledge distillation technique was used to achieve the compensation and optimization of the accuracy. Through experiments, it is found that compared with YOLOv8n, YOLOv8-LMD accuracy was improved by 1.7%, the computational amount was reduced by 68%, the parameter amount was reduced by 77%, and the model size was reduced by 75%. Therefore, it can be concluded that the model proposed in this paper is more suitable for the task of rock detection on the surface of Mars.