Novel Approach for Detection and Tracking of Explosive Carrying Mobile Humans with Odor-Sensor Based Multisensor Networks

In this study, a distributed system which could be used to detect and track an explosive carrying mobile human is designed. As a novel approach to existing odor detection and tracking methods, a technique which combines odor and sound detection is proposed for increased accuracy. To this aim, a MEMS sensor network including a pressure sensor, an accelerometer, a microphone, an odor sensor and electronic unit is proposed. For the electronic unit of this system a sensor network and least square estimation based sensor fusion algorithm is developed and simulated. The results are evaluated with respect to various paths of two humans, one carrying the explosive. The results show that odor sensing alone is not sufficient for the accurate determination of the odor track, and that by adding other conventional tracking methods, accuracy could be increased. Key-Words: Odor, Tracking, Sensor networks, Sensor fusion, Least squares, Explosive 1 Introduction Although there is sufficient technology to detect metal weapons, detection of plastic explosives such as C-4 with methods that are harmless to humans, still remains to be a challenge. Metal detectors or X-ray facilities commonly used at airports and customhouses cannot find plastic explosives. Y-ray detection using irradiation of neutron beams is recognized as an effective method for C-4 detection at this point, but few actually use this method due to doubts about the safety of using nuclear radiation on the public. Hence, methods must be sought to carry out C-4 inspection safely, nondestructively, and without physical contact. One possible method is the use of THz radiation, which is claimed not to harm the human body in [1]. Detection of explosives using odor sensors is already in use for the detection of land mines and there has been ongoing research on odor sensor based detection of plastic explosives [2][3],[4],[5], which mostly develop ways to detect the different odorants in a specific odor. There are numerous studies in the literature for odor source localization; some studies try to locate the odor source by triangulation [6], while some studies try stochastic estimation techniques like Least Square Estimation Method [6] and Maximum Likelihood Method [7], or Bayesian based methods like PQS (Process Query System) [8]. A common characteristic in all these algorithms is their low accuracy and sensitivity to environmental effects like the wind. The major contribution of this study is the development of a method that could be used with odorsensor based microelectromechanical (MEM) sensor networks for the detection and tracking of a human carrying a plastic explosive while moving among several other mobile, unarmed humans. In this study, it is also demonstrated that combining odor detection with sound detection will provide higher accuracy than finding the odor source directly, with the use of odor detection alone. In the developed algorithm, commonly used passive target tracking techniques such as acoustic, seismic and pressure based techniques are fused to accurately detect and track the target trajectory. These targets are then matched with the odor detection results, which are obtained via the conventional triangulation method. The resulting performance provides a higher accuracy in comparison to other odor source localization techniques reported in the literature Another contribution of the study is the detection and tracking of mobile odor sources (i.e explosives in this case) with the use of stationary sensors, unlike most studies in which stationary targets are pursued using mobile sensors [9]. This approach will also increase the chances of tracking and disarming the intruder without his/her awareness and at an appropriate moment and will significantly limit the harm caused to innocent bystanders and security officers. The use of odor sensors will further contribute to optimize energy consumption in the network in that all sensor groups in the module will be kept on standby, unless a command is received from the odor sensor. WSEAS TRANSACTIONS on SYSTEMS and CONTROL Ahmet Kuzu, Metin Gokasan, Seta Bogosyan ISSN: 1991-8763 63 Issue 2, Volume 3, February 2008 This study will combine these approaches for the aim of accurate detection and tracking of a human carrying an explosive in a public place. To this aim, a multisensor network (MSN) will be designed based on n-MEM sensor modules; each MEM sensor module will include low-cost, low-power passive sensors, namely odor sensors for the detection of the explosive, and a microphone, accelerometer, passive infrared sensors (PIRs) and a pressure sensor for the tracking of the human carrying the explosive. Each MEM module will also include circuitry for the data fusion of the odor sensors, signal conditioning circuitry as well as RF communication and power supply circuitry in the same case. Signal conditioners will convert sensor outputs in each module to appropriate levels, which are then processed via analog-to-digital converters (ADCs). Data from the odor sensors will also undergo a fusion process. The fused data will then be transmitted from each module, along with the outputs of the multiple sensors to the hub unit, via RF transmission. This unit is the major processing unit located in each subdivision of the network and performs fusion and decision tasks, while also conducting communication with each sensor module, other hub units in the network and the base computer for higher level data fusion and decisionmaking. This process will be further discussed in the “Proposed multisensor networking and sensor/data fusion approaches” section. A functional block diagram of the process is given in Figure 1. Figure 1. Functional block diagram of each MEM sensor module and hub unit Four different sensor types are considered for each MEM sensor module; namely, odor sensor, microphone, accelerometer [10], and pressure sensor. Structural diagram of this mechatronic module is shown in Figure 1. Odor sensors detect the explosives while microphones serve as passive radars, which detect the location of mobile objects from their vibrations in the air. Accelerometers, which detect seismic waves, also serve the same purpose. Finally, pressure sensors are activated upon direct contact, in that sense; they produce highest accuracy and hence, are also used to calibrate the other sensors. The Zigbee Network protocol [11], in RF Communication Block, is used to collect the sensor data, which are then used in fusion and decision algorithms [12]. The Rf Communication block also supplies sensor localization by using trianglization method [13,14]. Figure 2: Structural diagram of this mechatronic module. The odor sensor is comprised of a sensor array [15] to measure the density of a variety of odorants constituting an odor, a classifier to fuse and classify the collected data [16,17]; This sensor detects the odor of weapon, and estimate the source of odor. [18]. Although the resolution of odor sensors is lower than the other sensors, their use in combination with the odor sensor for the approximate localization of the target proves to be very effective as demonstrated by simulations in this study. The following sensors is designed individually and then, combined in a single case to constitute the above mentioned MEM sensor module: Sound sensors (MEM microphones): A microphone is a sensor, which converts acoustic pressure to voltage by using a thin diaphragm. When voice pressure strikes a diaphragm, stress and depletion occurs. This stress can be easily measured by piezoresistors, or depletion can be detected by capacitance change. Figure 3 depicts the architecture of a piezoresistive microphone. Signals output by the microphone provide some information on the position of the intruder based on the duration of time required for the intruder-transmitted sound waves to arrive at the microphone. This data component will be further fused with data produced by other sensor modules at different locations to yield the position of the intruder. The need for other sensors in addition to the HUB UNIT SENSOR FUSION Microphone