Moving-objects Detection and Background Modeling

Time-varying phenomenon, such as ripples on water, trees waving in the wind and illumination changes, produces false motions, which significantly compromises the performance of an outdoor-surveillance system. In this project, we propose a corner-based background model to effectively detect moving-objects in challenging dynamic scenes. Specifically, the method follows a three-step process. First, we detect feature points using a Harris corner detector and represent them as SIFT-like descriptors. Second, we dynamically learn a background model and classify each extracted feature as either a background or a foreground feature. Last, a ‘Lucas-Kanade’ feature tracker is integrated into this framework to differentiate motion consistent foreground objects from background objects with random or repetitive motion. The key insight of our work is that a collection of SIFT-like features can effectively represent the environment and account for variations caused by natural effects with dynamic movements. Features that do not correspond to the background must therefore correspond to foreground moving objects. Our method is computational efficient and works in real-time. Experiments on challenging video clips demonstrate that the proposed method achieves a higher accuracy in detecting the foreground objects than the existing methods.

We detail the learning and classification in Algorithm 1. In our algorithm, instead of blindly updating the background model, we choose the selective update scheme, i.e. only the background features classified in Step 3 are used to update model entry. Moreover, rather than using a direct mapping between position of model entry and feature location, a local window (e.g. 5 × 5) is defined to search the best model entry accounting for the newly detected feature. This shift matching design results in a very sparse representation of the learned model because an existing model entry will prohibit generating new entries within its neighborhood. Therefore, features subject to small shifts will be mapped onto the same model entry as time proceeds.

We detail the usage of a ‘Lucas-Kanade’ tracker in Algorithm 2, which mainly consists of three individual modules. The first module takes charge of generating a new tracker for each newly identified foreground feature. The second module deals with the optic-flow calculation for each tracker in the list once a new frame becomes available. A number of rules are designed for tracker deletion and merger, which result in a significant reduction of the total tracker count. In addition, we pass the trackers with consistent trajectories to the third module, where a cluster of similar motion trajectories is confirmed as a real moving-object. The misclassified features, which are from the dynamic background, usually result in repetitive or random motion. Therefore, they can be removed in this step.

• moving-objects detection in several typical surveillance videos (download)