Operational discharges of oil from vessels, whether accidental or deliberate, are a growing concern as the levels of maritime traffic increase. Oil tankers and other kinds of ships are among the suspected offenders of illegal discharges. The international legislation contains minor and well-defined exceptions related to ocean areas (internal waters, marine protected areas, MARPOL aÌ22́Ơ¿3specialaÌ22́Ơ℗+ areas, territorial seas or exclusive economic zones). These areas often determine whether an action is considered legal or not and define the rights and obligations, including law enforcement obligations. Synthetic aperture radar (SAR) is the most used remote sensing tool for monitoring oil pollution over vast ocean areas. SAR is an active microwave RS sensor capable of taking measurements day or night and almost independently from atmospheric conditions. Manual oil spill detection in a SAR image is ordinarily done by a trained human interpreter who visually inspects SAR images for any possible spills. However, manual inspection can be time-consuming, biased, inconsistent and subjective. A faster and more robust alternative is to use automated image processing and machine learning methods. The current automated oil detection methods, however, are still not ideal and there is still a need for improvement. Also, data costs have resulted in limited studies on oil spill detection in African oceans. The launch of several Sentinel missions with SAR sensors has considerably improved coverage and accessibility of data over African oceans. The goal of the study is to develop an automated detection of oil spill discharges from vessels in African seas using the freely available Sentinel SAR data. A novel oil spill detection framework that can detect possible oil spill candidates and remove unwanted detections (i.e., false positives) was proposed. The framework used a novel linear dark spot detection algorithm and an improved oil spill discrimination process. The linear detection process used a segmentation-based algorithm to isolate linear dark spots (potential oil spills) from other features in the image. The process involved a more efficient feature selection and classification process. The proposed linear detection algorithm was evaluated for detection accuracy and compared to other segmentation-based oil spill detection algorithms, including state-of-the-art oil spill detection methods. The results demonstrated the proposed approach to be a more efficient and robust linear dark spot detection method. An improved discrimination process was presented to reduce false detections from a segmentation-based algorithm. The selection of relevant oil spill features depends on many factors which could influence the accuracy of the classification task. Automated features selection methods were thus considered to improve the discrimination process. Using feature selection, the most significant oil spill features with minimum variations were determined. The significant features were used as input vectors to classify oil spill events from moving vessels. An optimised Gradient Boosting Tree Classifier (GBT) was used for the classification task. The proposed novel framework showed promising results for monitoring oil spill from moving vessels using SAR in African oceans on a regular basis. Future work includes adding a confidence measure and alert level estimation. The system will incorporate ancillary information such as the oil spill source and the sensitivity of the polluted area to measure environmental impact.