Oil pipeline leakage detection has been an important and ongoing problem in industrial development throughout the ages. Data based on SAR-captured images has been used to detect oil spills. Then, intelligent models are compared with each other using existing performance metrics (accuracy, sensitivity, precision, and F1 score). The purpose of this paper is to explore the feasibility of using AI, such as deep learning convolutional neural networks, in oil spill detection and monitoring on land. This study investigates the effectiveness of pre-trained deep learning models, which were trained on big data in ImageNet. Three pre-trained models, DenesNet-201, ResNet-50, and Xception were selected for the experiments based on their structure and features. The results showed that the highest accuracy was obtained by Xception model compared to the DenseNet-201 and ResNet-50 algorithms, where Xception algorithm gave an accuracy of 99.2%, which is much higher compared to DenseNet-201 and ResNet-50, algorithms which gave an accuracy of (96.5%, and 92.8%) respectively. Thus, the Xception model has better performance in oil leak detection.