Abstract: Combined with image processing technology, a pipeline integrity detection method for oil and gas station is proposed. In the research, the pipeline image of oil and gas station is collected by pipeline robot. In order to improve the image quality, the collected original image is processed. For the preprocessed image, the image defect area was marked and the following five geometric feature parameters were extracted, which were the defect area, defect perimeter, defect roundness, the ratio of defect long axis to short axis, and defect gradient. Taking these five geometric characteristic parameters as inputs, the classifier is constructed by DBN (depth confidence network) to detect the pipeline integrity of oil and gas stations. The results show that the quality index of defect identification is greater than 8. 0, which shows that the integrity detection result of this method is more accurate and can be used in the integrity detection of pipelines in actual oil and gas stations.