Trajectory Data Mining in Distributed Sensor Networks

The prevalence of distributed sensor networks on the Internet has allowed the field to become a relevant and active research area, attracting professionals and researchers from a variety of fields and disciplines. By incorporating trajectory information, we can bring distributed sensor networks back to the physical world, conveniently sharing our real-life experiences in the virtual world. By mining trajectory patterns or predicting locations from distributed sensor networks, people cannot only track and share location-based information with each other via mobile devices, desktop computers, or sensors, but also benefit from the collective knowledge learned from this content [1].

Wireless sensor networks, global positioning systems, and mobile computing techniques are developing rapidly, such that it is now possible to systematically track and predict the mobility of objects which accumulate a huge collection of mobile data. Accordingly, there is an ever-increasing interest in performing data analysis over trajectory data [2]. By processing and analyzing the historical trajectory data using data mining or machine learning techniques, it is relatively easy to predict accurate information associated with the position of moving objects. This practice is known as trajectory data mining (TDM for short) [3, 4], and as our daily lives become even more embedded with these technologies, the importance of TDM is even more important. For example, consider the case when a GPS enabled device is taken beyond the working proximity of satellites and is rendered inoperable. A possible solution to this is to employ TDM software which can provide an intelligent navigation service.

Arguably, a trajectory is one of the most fundamental properties in human life, and the research on TDM in distributed sensor networks works to bridge the gap between the virtual and physical world. This research of trajectory data mining in distributed sensor networks has the potential to change the way we live, such as enabling applications for better path planning [5, 6] and restaurant/business recommendations. Research can also be conducted to advance human mobility modeling and user activity analyses [6–9], which can have broad impacts on social science and engineering.

This special issue covers a wide range of research work which can contribute to addressing and solving the challenges faced in the location-based distributed sensor network research domain, for example, spatial and spatiotemporal data mining in distributed sensor networks, moving object tracking, indexing, and retrieval in distributed sensor networks, and activity recognition and sensing for distributed sensor networks.

The readers of TDM in distributed sensor networks can find in this special issue not only state-of-the-art research findings and updated reviews on the common techniques in TDM, but also important questions to be resolved, that is, user behavior modeling using physical sensor data and mobile and ubiquitous computing for distributed sensor networks.

Currently, the problem of security in distributed sensor network has attracted a lot of attention from researchers. And cloud data storage and retrieval have become popular for efficient data management in distributed sensor networks. In this special issue, X. Zhang et al. proposed an efficient pairing-free auditing scheme for data storage in distributed sensor networks. They employed a third party auditor (TPA) to verify the integrity of sensor data without retrieving the entire data information. In addition, they designed the homomorphic message authentication codes to reduce the space storage of the verification information. They also employed the random masking technique to guarantee that the TPA cannot recover the primitive data blocks in distributed sensor networks. Finally, they adapted the proposed scheme for supporting batch auditing so that the TPA can efficiently perform multiple auditing tasks.

Multisensor information fusion has garnered wide support in a variety of applications and is gradually becoming an active research area. Bad shunting of track circuit is one of the major risks for railway traffic safety. The occupancy of track cannot be correctly detected due to bad shunting, which could severely degrade the efficiency of dispatching train commands. In order to improve the efficiency of track occupation detection, Z. Hua et al. proposed a multisensor track occupancy detection model based on chaotic neural networks. This model used the detection results of track occupancy collected by multiple sensors as the fundamental data and then calculated the weights using the chaotic neural networks and performs data fusion, in order to determine whether the track is occupied. Extensive experimental results demonstrate that the proposed model can detect the track occupancy in an effective and efficient fashion.

With the rapid development of space technology, asteroid exploration becomes an active research field in deep space exploration. How to find the global optimum flight program is a key problem in TDM of the deep space exploration. Aiming to handle this problem, M. Wang et al. proposed an approach to design the optimal trajectory by differential evolution (DE) algorithm for asteroid exploration based on mixed coding, while the celestial sequence and the time sequence are coded together into the chromosomes of DE and optimized simultaneously. The proposed algorithm can utilize the characteristics of the high efficiency and global optimization ability of differential evolution, as well as avoid the problem of high complexity in the branch-and-bound algorithm and the problem of local optimal solutions in the greedy algorithm. The proposed approach can be used to solve the Fourth Contest of National Space Orbit Design in China, and the result shows that both the computational efficiency and the performance of the algorithm are superior.

Nowadays, a large number of network thermal imaging cameras have worked over distributed sensor networks, offering the capability of online remote intelligent video surveillance. Z. Li et al. proposed a new intrusion detection and cooperative tracking approach applied for PTZ (Pan/Tilt/Zoom) network thermal imagers. First, the intrusion detection module eliminates the offset between the current frame and the prior frame via FOV (Field-of-View) matching and then handles intrusion detection by motion detection in the preset surveillance zone. The cooperative tracking module shifts the priority of tracking by imager pose estimation, which is based on FOV matching, and can avoid transferring the local features from one imager to another one. In addition, another work relevant to this research is addressed. J. Yang et al. presented an original particle filter tracking algorithm named labeled particle filter which describes each image patch with a binary label. They used a one-bit binary label, that is, positive or negative, to describe the attribute of image patch. Therefore, the candidate target template is established only if the label of the candidate targets matches the label of the reference target, and the computational complexity can be reduced. Experimental results show that the proposed algorithm can handle the real-time object tracking with less time cost as well as maintaining high tracking accuracy.

We hope that this special issue will spark your interest in the young yet fast-evolving field of trajectory data mining in distributed sensor networks. The techniques and algorithms presented are of practical utility, rather than selecting algorithms that perform well on small “toy” sensor data sets. The research works described in this special issue are geared for the discovery of user behavior and mobility in real sensor data.