Tracking by Deblatting

Objects moving at high speed along complex trajectories often appear in videos, especially videos of sports. Such objects travel a considerable distance during exposure time of a single frame, and therefore, their position in the frame is not well defined. They appear as semi-transparent streaks due to the motion blur and cannot be reliably tracked by general trackers. We propose a novel approach called Tracking by Deblatting based on the observation that motion blur is directly related to the intra-frame trajectory of an object. Blur is estimated by solving two intertwined inverse problems, blind deblurring and image matting, which we call deblatting. By postprocessing, non-causal Tracking by Deblatting estimates continuous, complete, and accurate object trajectories for the whole sequence. Tracked objects are precisely localized with higher temporal resolution than by conventional trackers. Energy minimization by dynamic programming is used to detect abrupt changes of motion, called bounces. High-order polynomials are then fitted to smooth trajectory segments between bounces. The output is a continuous trajectory function that assigns location for every real-valued time stamp from zero to the number of frames. The proposed algorithm was evaluated on a newly created dataset of videos from a high-speed camera using a novel Trajectory-IoU metric that generalizes the traditional Intersection over Union and measures the accuracy of the intra-frame trajectory. The proposed method outperforms the baselines both in recall and trajectory accuracy. Additionally, we show that from the trajectory function precise physical calculations are possible, such as radius, gravity, and sub-frame object velocity. Velocity estimation is compared to the high-speed camera measurements and radars. Results show high performance of the proposed method in terms of Trajectory-IoU, recall, and velocity estimation.

Real-time data mining for multimedia streaming

Developing a Web Video Player connected to a security surveillance camera for collecting the video streams is the main objective of this study. The Developed Web Application tracks the target object through the sequences of video frames and generates the object trajectories. The video frames are analyzed, and the object trajectories are fed into a classifier or clustering method for training and movement detection purposes. In this thesis, several machine learning techniques are applied and implemented in Batch and in Real-Time mode including SVM, J48 Decision Tree, PART, Decision Table, Decision Stump, Multilayer Perceptron, and K-Means clustering by using two customized datasets. The object tracking, and movement detection are based on a simplified HSV color space model. The developed Web Application and proposed architecture are implemented on a local area network with in-house Server as well as a single computer and can detect the trajectories of the moving objects effectively.

Real-time data mining for multimedia streaming

Developing a Web Video Player connected to a security surveillance camera for collecting the video streams is the main objective of this study. The Developed Web Application tracks the target object through the sequences of video frames and generates the object trajectories. The video frames are analyzed, and the object trajectories are fed into a classifier or clustering method for training and movement detection purposes. In this thesis, several machine learning techniques are applied and implemented in Batch and in Real-Time mode including SVM, J48 Decision Tree, PART, Decision Table, Decision Stump, Multilayer Perceptron, and K-Means clustering by using two customized datasets. The object tracking, and movement detection are based on a simplified HSV color space model. The developed Web Application and proposed architecture are implemented on a local area network with in-house Server as well as a single computer and can detect the trajectories of the moving objects effectively.

Synthesis of an Algorithm for Processing the Trajectories of Moving Objects Using the Methods of Data Clustering Theory

Introduction. Requirements for the quality of information about the trajectory of moving objects provided by sensor networks are increasingly becoming more stringent. For Information and Data Processing Centers (DPC) at control and management command posts, the issue of information mapping and forming the true trajectories of moving objects in the area of intersection of network detection zones is of particular importance. The use of conventional approaches to solving this problem involves issues related to ensuring the efficient provision of users with complete and reliable information about trajectories in real time. In this article, wee propose a new approach to solving this problem using data mining theory, in particular, the methods of data clustering theory. Based on an analysis of the process of processing radar data in a DPC and its similarity with that of data clustering, we synthesized an algorithm for processing the trajectories of moving objects. The algorithm was verified by modelling and experimental research.Aim. To develop a generalized scheme for processing object trajectories (TP) in a DPC and to synthesized a TP algorithm using the methods of data clustering theory.Materials and methods. Data Clustering theory, Systems Engineering theory, Radar Data processing theory (RD), methods of mathematical modelling and experimental research.Results. Based on an analysis of the essence of radar data processing (RD) in a DPC and its similarity with the process of data clustering, an algorithm for processing the trajectories of moving objects was synthesized and verified by modelling and experimental research. A generalized scheme for processing the trajectories of moving objects in a DPC and a TP algorithm for a DPC were synthesized.Conclusions. An algorithm for processing object trajectories was proposed based on a new approach of data clustering theory. A generalized scheme and an algorithm for processing object trajectories (TP) in a DPC were suggested. These developments can be effectively applied in various models, e.g. centralized, hierarchical and decentralized. The synthesized algorithm can provide output information about the true identified trajectories in terms of various indicators of data processing systems (DPS).

Clustering Moving Object Trajectories: Integration in CROSS-CPP Analytic Toolbox

Mobile devices equipped with sensors are generating an amount of geo-spatial related data that, properly analyzed can be used for future applications. In particular, being able to establish similar trajectories is crucial to analyze events on common points in the trajectories. CROSS-CPP is a European project whose main aim is to provide tools to store data in a data market and to have a toolbox to analyze the data. As part of these analytic tools, a set of functionalities has been developed to cluster trajectories. Based on previous work on clustering algorithms we present in this paper a Quickbundels algorithm adaptation to trajectory clustering . Experiments using different distance measures show that Quickbundles outperforms spectral clustering, with the WS84 geodesic distance being the one that provides the best results.

Video trajectory analysis using unsupervised clustering and multi-criteria ranking

Surveillance camera usage has increased significantly for visual surveillance. Manual analysis of large video data recorded by cameras may not be feasible on a larger scale. In various applications, deep learning-guided supervised systems are used to track and identify unusual patterns. However, such systems depend on learning which may not be possible. Unsupervised methods relay on suitable features and demand cluster analysis by experts. In this paper, we propose an unsupervised trajectory clustering method referred to as t-Cluster. Our proposed method prepares indexes of object trajectories by fusing high-level interpretable features such as origin, destination, path, and deviation. Next, the clusters are fused using multi-criteria decision making and trajectories are ranked accordingly. The method is able to place abnormal patterns on the top of the list. We have evaluated our algorithm and compared it against competent baseline trajectory clustering methods applied to videos taken from publicly available benchmark datasets. We have obtained higher clustering accuracies on public datasets with significantly lesser computation overhead.