In conventional computer vision systems, high image quality and long target exposure requirements are required. In this thesis, two algorithms to overcome such limitations of current computer vision systems have been proposed. The Pixel Exclusion Double Difference Algorithm (PEDDA) algorithm is a novel object detection algorithm that is able to detect fast moving objects in noisy images and suppress interference from large, low speed moving objects. The State-based “Observation, Analysis and Prediction” Target Election and Tracking Algorithm (SOAPtet) algorithm uses a deterministic state machine to guide the SOAPtet algorithm predictions. A novel stochastic based approach is also implemented in this algorithm to elect the target of interest from its candidates that are usually triggered by noise. A real time experimental system is developed based on the two algorithms. The experiment results show that this system detects up to 92.3% of moving objects in noisy environment and the tracking accuracy is up to 97.42%.