Learning Spatio-Temporal Information for Multi-Object Tracking

ABSTRACT:

The robust multi-object tracking problem is a challenging issue in the field of computer vision. In this paper, we propose a multi-object tracking algorithm with temporal-spatial information and trajectory of confidence. The whole process is divided into local and global association. Trajectories with high confidence are associated with the detection result of the current frame during local association, whereas trajectories with low confidence are associated with the detection results of the current frame are not matched during global association. We determine the association results using a combined model. By utilizing the information of spatial-temporal correlation, the model is more robust and can deal with missed detection. In addition, we measure the reliability of the spatial information by the confidence map smoothing constraint and the peak sidelobe ratio criterion. We conduct experiments using a challenging public data set, and the results show that our proposed algorithm is superior to many other popular algorithms when dealing with problems, such as missed detection and poor tracker robustness.

EXISTING SYSTEM:

• In recent years, the existed methods in the field of multi-object tracking are mostly based on Kalman filters and particle filters. These methods are effective for forecasting states that have a short duration rather than in complex scenes. These data association methods, such as the joint probabilistic data association filter (JPDAF), multiple hypothesis tracking (MHT), and Markov chain Monte Carlo sampling techniques (MCMCDA) can solve the tracking problem in complex scenarios.

• This method associates all detection results obtained by detection to generate the trajectories of objects and can be roughly divided into steps, namely batch processing and online processing.

DISADVANTAGES OF EXISTING SYSTEM:

• The global association is critical during the progress.

• Many global association methods have been proposed, but they still cannot treat cases in which target objects are blocked for a long time.

• The confidence of the trajectories may decrease when objects are blocked, and detection is lost during detection; hence, global association is still needed to connect these fragments of low confidence to form complete trajectories.

PROPOSED SYSTEM:

• In this article, we apply spatial and temporal information to the combined model. For the problem of tracking, temporal information refers to all of the objective information of prior frames, while spatial information refers to local target and surrounding regions in the background.

• Most local spatial information remains unchanged between two adjacent frames owing to the small neighboring time intervals. As a consequence, there is a strong spatial and temporal relationship between continuous frames.

• The proposed combined model is more robust because of the strong correlation of temporal information, and can address issues such as incorrect tracking more easily.

• In addition, in this paper, we propose to measure the reliability of spatio-temporal information using the confidence map smoothing constraint and the peak sidelobe ratio criterion. The proposed multi-object tracking algorithm can effectively deal with the problems of missed detection, and improve the robustness of the target tracker.

ADVANTAGES OF PROPOSED SYSTEM:

• In this paper, we used the CLEAR evaluation, which is composed mainly of two parts: multiple object tracking precision (MOTP), which reflects the accuracy when determining the targets location, and multiple object tracking accuracy (MOTA), which reflects the accuracy when determining the number of goals and the related properties. Both of them jointly measure the ability of the algorithm to perform continuous tracking.

• In this paper, we consider the spatial information continuity between adjacent frames to solve the problem of missed detection.

• The improved model is more robust because of the correlation relationship of the spatial-temporal information, and it helps to solve the problem of missed detection.

• In addition, we measure the reliability of the spatial-temporal information using the principles of SCCM and PSR.

• We use a publicly available dataset to perform experiments using our proposed algorithm, and our results show that our algorithm can deal with problems such as missed detection, simultaneously improving the robustness of the track detector, hence, it is superior to many other popular algorithms.

SYSTEM REQUIREMENTS:

HARDWARE REQUIREMENTS:

• System           : Pentium Dual Core.

• Hard Disk      : 120 GB.

• Monitor          : 15’’ LED

• Input Devices : Keyboard, Mouse

• Ram                : 1GB.

SOFTWARE REQUIREMENTS:

• Operating system : Windows 7.

• Coding Language : MATLAB

• Tool                       : MATLAB R2013A

REFERENCE:

JIAN WEI, MEI YANG, AND FENG LIU, “Learning Spatio-Temporal Information for Multi-Object Tracking”, IEEE Access 2017.