Abstract: Oil producing safety is the key to ensure the stable supply of energy. Using robots to detect the state of oil producing can effectively prevent accidents and reduce risks. In this paper, we propose a method for detecting the leakage state of oil pipeline. Firstly, a feature enhancement module based on the residual pyramid is designed to capture image details and improve the discrimination of leakage state features and complex background by using its multi-scale analysis capability. Then, a model compression strategy based on information entropy metrics is proposed to improve the detection efficiency of oil pipeline leakage state under the deployment of edge equipment. Experimental results on a real data set demonstrate the effectiveness of the proposed method. Compared to the baseline model, the compression rate of the model parameters is increased to 56.43%, GFLOPs and parameters decrease 44.29% and 28.60% respectively, with only a 0.02% decrease in term of detection accuracy.