Exploring Local Context for Multi-target Tracking in Wide Area Aerial Surveillance

Abstract

Abstract. Tracking many vehicles in wide coverage aerial imagery is crucial for understanding events in a large field of view. Most approaches aim to associate detections from frame differencing into tracks. However, slow or stopped vehicles result in long-term missing detections and further cause tracking discontinuities. Relying merely on appearance clue to recover missing detections is difficult as targets are extremely small and in grayscale. In this paper, we address the limitations of detection association methods by coupling it with a local context tracker (LCT), which does not rely on motion detections. On one hand, our LCT learns neighboring spatial relation and tracks each target in consecutive frames using graph optimization. It takes the advantage of context constraints to avoid drifting to nearby targets. We generate hypotheses from sparse and dense flow efficiently to keep solutions tractable. On the other hand, we use detection association strategy to extract short tracks in batch processing. We explicitly handle merged detections by generating additional hypotheses from them. Our evaluation on wide area aerial imagery sequences shows significant improvement over state-of-the-art methods.

Comparison of different detection methods for persistent multiple hypothesis tracking in wide area motion imagery

“…For the exemplary configurations, best performance is achieved by Median BG + N for q = 0.0036 with an f-score of 0.773 after tracking and a MOTA value of 0.611. In comparison to a recent survey on trackers for persistent tracking in WAMI data [2], where the trackers proposed by Chen et al [2], Reilly et al [6], and Prokaj et al [1] achieved best results with MOTA values of 0.599, 0.522, and 0.493, respectively, this demonstrates state of the art performance of the proposed tracking framework. A more detailed performance comparison, which is based on the same evaluation sequence and code also used in [2] to ensure comparability of results, can be found in [3].…”

“…As trackers solely relying on moving object detections fail when targets slow down or become stationary, developing methods for persistent tracking has recently become a focus of research interest in the WAMI community. To handle long-term missing motion detections, Prokaj et al [1] and Chen et al [2] combine a detection-based tracker with a regression tracker and with a local context tracker, respectively. In [3] it is demonstrated that state of the art results can also be achieved with a MHT-based single-tracker framework that avoids the additional complexity introduced by frameworks based on two trackers and recovers missing motion detections with a classifier-based detector.…”

“…Many trackers are evaluated using only one specific detection method with fixed settings (e.g. [2,8,9]); in [1] two detector configuration are evaluated in order to test the robustness of the tracking approach. To the best of our knowledge, the impact of systematically varying detection methods and detector settings on the output tracking quality have not yet been analyzed.…”

“…Recent approaches for persistent tracking, i.e. tracking vehicles even if they become stationary, compensate for missing motion detections by combining a detection-based tracker with a second tracker based on appearance or local context [1,2]. In order to avoid the additional complexity introduced by combining two trackers, we employ an alternative framework based on multiple hypothesis tracking (MHT) that requires only a single tracker and enables persistent tracking in WAMI data by recovering missing motion detections with a classifier-based detector [3].…”

Detecting and Tracking Small Moving Objects in Wide Area Motion Imagery (WAMI) Using Convolutional Neural Networks (CNNs)

“…For the areas with heavy traffic, a lightweight regression network is trained to predict the centre of multiple moving objects from individual detections. This approach of combining the CNNs' and background subtraction's outputs makes it possible for the shape of moving objects to be obtained: this can be useful information that can be exploited by appearancebased tracking systems (e.g., [17], [18]). Finally, in addition to comparing the detections with the ground-truth, we applied a Gaussian Mixture Probabilistic Hypothesis Density (GM-PHD) filter to the detections to directly utilise the product of the proposed algorithm.…”