This paper presents a machine learning approach to corner detection, which significantly improves both the repeatability and efficiency of existing corner detectors. The proposed method is designed to process live PAL video using less than 5% of the available processing time, outperforming other detectors like Harris and SIFT in terms of speed. The detector is also generalized to optimize for repeatability with minimal loss of efficiency. A rigorous comparison of corner detectors based on repeatability in 3D scenes shows that the proposed method significantly outperforms existing detectors, even though it was primarily designed for speed. The comparison demonstrates that machine learning improves repeatability, resulting in a detector that is both fast and high quality. The paper reviews previous work on corner detectors, including edge-based, grey-level derivative-based, and direct grey-level detectors. It also discusses machine learning-based detectors, which use classifiers trained on example corners to detect features in images. The paper evaluates the performance of various corner detectors using three categories of tests: corner detection as object recognition, system performance, and repeatability. The results show that the proposed method achieves high repeatability and efficiency, making it suitable for real-world applications. The FAST detector is introduced as a high-speed corner detector that uses a segment test criterion to identify corners. It is based on a decision tree classifier trained using machine learning. The FAST-ER detector is an enhanced version of FAST that is optimized for repeatability. The paper also discusses the use of simulated annealing to optimize the decision tree for repeatability and efficiency. The results show that the FAST-ER detector achieves high repeatability and efficiency, making it suitable for real-world applications. The paper concludes that the proposed method is a significant improvement over existing corner detectors, offering a balance between speed and repeatability.This paper presents a machine learning approach to corner detection, which significantly improves both the repeatability and efficiency of existing corner detectors. The proposed method is designed to process live PAL video using less than 5% of the available processing time, outperforming other detectors like Harris and SIFT in terms of speed. The detector is also generalized to optimize for repeatability with minimal loss of efficiency. A rigorous comparison of corner detectors based on repeatability in 3D scenes shows that the proposed method significantly outperforms existing detectors, even though it was primarily designed for speed. The comparison demonstrates that machine learning improves repeatability, resulting in a detector that is both fast and high quality. The paper reviews previous work on corner detectors, including edge-based, grey-level derivative-based, and direct grey-level detectors. It also discusses machine learning-based detectors, which use classifiers trained on example corners to detect features in images. The paper evaluates the performance of various corner detectors using three categories of tests: corner detection as object recognition, system performance, and repeatability. The results show that the proposed method achieves high repeatability and efficiency, making it suitable for real-world applications. The FAST detector is introduced as a high-speed corner detector that uses a segment test criterion to identify corners. It is based on a decision tree classifier trained using machine learning. The FAST-ER detector is an enhanced version of FAST that is optimized for repeatability. The paper also discusses the use of simulated annealing to optimize the decision tree for repeatability and efficiency. The results show that the FAST-ER detector achieves high repeatability and efficiency, making it suitable for real-world applications. The paper concludes that the proposed method is a significant improvement over existing corner detectors, offering a balance between speed and repeatability.