• No results found

LITERATURE REVIEW

2.3 WIRELESS VISUAL SENSOR NETWORK

In recent years, inexpensive hardware such as CMOS cameras are widely available in the market, which allows images to be captured from the environment in the WSN [15]. With this development, it gives rise to a new form of network known as Wireless Visual Sensor Networks (WVSNs). Each of these sensor nodes, it can process captured image data locally and extract the relevant information to be sent to the base-station. Besides, the sensor nodes are able to collaborate with other nodes on the application-specific task that is able to provide the user with information rich in description for a particular captured event [3]. The WVSN offers a wide range of applications, from remote and distributed video-based surveillance system to ambient assisted living and personal care applications. Users can remotely visit interesting locations through virtual reality using WVSN [5].

The WVSN offers many new applications compared to the WSN that uses only scalar sensors. Nevertheless, WVSN does face new problems such as a huge amount of data produced from the camera sensors [5]. Processing such large amount of data under constrained conditions, where there are limited amount of energy source, low bandwidth resources and limited processing power [3], is a challenge for generally low-powered sensor nodes. In the literature [7], it is stated that the amount of energy that can be used to process the data is much lower compared to the energy for use in transmitting the data across the wireless network. The transmitting cost (in terms of energy) of 1kb data is comparable to the same amount of energy use by a general-purpose processor that executes 3 million instructions [11]. As a result, large amount of data that is produced by the visual sensor nodes can be locally compressed, such that it reduces the large amount of data transmitting through the wireless network [9].

Since the sensor nodes have limited energy constraints and also limited processing power, the image compression techniques/encoders developed for use in WVSNs need to be low in power consumption [5]. Traditional image compression algorithms are not suitable for use in WVSNs [9] as they are mainly designed for multicasting/broadcasting applications [5], which is shown in Figure 8. The emphasis is to design a low-complexity decoder with the tradeoffs that the encoder bears the computational burden in the process of transmitting the information. However, for the

21

WVSNs, the complexity requirements are reversed due to their mostly many-to-one information flow.

For this research, the focus is on sensor node that captures still images and transfers the image data to the sink (base-station) across the resource-constrained WVSNs. The purpose of developing the visual sensor node is to provide surveillance for the military, especially to determine the number of enemy soldiers beyond the enemy line. Various existing WVSN platforms that were previously developed are reviewed in the following Section 2.2.1 to show the differences between these platforms and the developed joint image compression, encryption and error correction processing framework for WVSNs.

Figure 8 Information flows in traditional broadcasting application [5].

2.3.1 Existing WVSN Platforms

In [51], a wireless sensor device or “mote” known as Telos was introduced by the researchers from the University of California, Berkeley. The Telos was designed to sleep for major of time, wake up quickly on an event, process the information and return to sleep. The Telos platform was controlled using the Texas Instrument (TI) MSP430 microcontroller with 48kB of Programme Memory (FLASH memory) and 10kB of RAM buffer (SRAM). For the communication radio, the Telos uses the Chipcon CC2420 radio that operates in IEEE 802.15.4 standard [52] with transmission frequency of 2.4GHz. The Telos platform was developed such that it provides the capability to incorporate sensor into the platform. For the microcontroller

22

unit to write data into flash memory, the power consumed by the Telos platform is 27.18mW at operating voltage of 1.8V [51]. However, the developed Telos platform was designed to be a basic WVSN platform that transmitted scalar data to the base-station without performing any data processing.

Later on in the literature [53], Cyclops platform was developed to perform hand posture recognition onto the captured images. Cyclops is an electronic interface that connects a camera module and a lightweight wireless host together. The Cyclops module is made up of a Xilinx XC2C256 CoolRunner Complex Programmable Logic Device (CPLD), 64kB of SRAM, 512kB of FLASH memory storage, an Atmel ATmega128L Micro-Controller Unit (MCU) running at 4MHz and an Agilent CMOS camera module (ADCM-1700). For the maximum power consumption, the worst case scenarios is considered when the Cyclops platform performs a write data operation with the permanent memory access. As reported, the maximum power consumed by the Cyclops platform operating at 3.0V is 64.8mW [53].

Next, an Intel Mote platform for industrial monitoring that measures vibration was developed [54]. The developed Intel Mote platform incorporates an Zeevo integrated wireless microcontroller module, industrial vibration sensor and a surface-mount 2.4GHz antenna together as a complete platform. The Zeevo module used for the Intel Mote consists of an ARM7TDMI architecture core, 64kB of SRAM, 512kB FLASH memory and a CMOS Bluetooth radio. TinyOS operating system was ported into this ARM architecture and this leaves about 11kB of free SRAM available to be used by any written applications in the platform [54].

In [55], the Panoptes video sensor platform was developed by integrating Intel StrongARM 206MHz embedded processor, a Logitech 3000 USB video camera, 64MB of memory, Linux 2.4.19 operating system kernel and an 802.11-based networking card together in a Bitsy board. For the proposed Panoptes video sensor platform, the power consumed by Computer Processing Unit (CPU) alone is 2.287W.

The total power consumption required for Panoptes platform to capture, process and transmit the video is 5.268W. Since the proposed Panoptes platform is intended for use with a wind-powered generator that has unlimited energy source. Therefore, the high power consumption by the Panoptes platform is suitable in this applications [55].

The author in [56] had proposed an image sensor mote for use in Wireless Image Sensor Networks (WISNs). The microcontroller unit used in this sensor mote is the Atmel AT91SAM7S128 microcontroller based on the ARM7TDMI architecture

23

core. The operating frequency of the microcontroller was set at 48MHz with available memory storages: 32kB of RAM and 128kB of Flash memory. Meanwhile, the proposed sensor mote used the Chipcon CC2420 based on IEEE 802.15.4 communication radio. The sensor mote operating voltage is at 3.3V and the power consumption for the microcontroller itself is 99mW. The developed image sensor mote was used to detect and determine the direction of pedestrian movement in a narrow pathway [56].

A CRITIC wireless camera mote was developed for the Heterogeneous Sensor Networks (HSNs) [57]. The CRITIC platform consists a 1.3 Megapixel OmniVision OV9655 CMOS sensor, Intel XScale PXA270 fixed-point processor (256kB internal SRAM), 64MB of external SRAM, 16MB Flash memory and Chipcon CC2420 radio.

In the Idle mode (with no active processes), the power consumption of the Intel XScale processor was between 428mW - 478mW. The proposed CRITIC platform was developed to perform multiple target tracking and sending low amount of image data across the low bandwidth WSNs. This was done by processing the captured images locally on the camera board by using background subtraction for single target tracking and camera localization for multiple target tracking. Then transmitting only compressed low-dimensional image features to the sink (base-station), which routes the information to various clients for further processing and visualisation [57].

2.3.2 Summary

Most of the existing sensor nodes mentioned earlier were operating in the IEEE 802.15.4 standard that only consists of an error detection (CRC-16) [52]. In the IEEE 802.15.4 standard, there is no security protection is applied onto the data. The available data security protections is only applicable for the Waspmote (eg. Digi XBee, which is a proprietary Radio Frequency transceiver) that operates in IEEE 802.15.4 standard. As such, the CRS coding scheme was used in this research to provide data security. The CRS coding scheme has similar security level as the Advanced Encryption System (AES), which is the standard use for encrypting data by the US National Institute of Standards and Technology (NIST) [58]. At the same time, the CRS coding scheme also offers error protection capability as offered by traditional Reed Solomon coding scheme [59].