Optical Camera-Based Indoor and Vehicular Communication

Recently, optical wireless communication (OWC) emerges as a complementary solution to conventional radio frequency (RF) communication, which can offer high data rate, secure communication and immunity to RF interference. With the development of lightemitting diode (LED)-based lighting infrastructure, as well as the wide availability of cameras on mobile phones and other devices (e.g., vehicles), Optical camera communication (OCC) has been proposed as a type of practical and cost effective OWC technology. In OCC systems, LEDs can serve as both illumination sources and communication transmitters, and ubiquitous cameras on different smart devices can be conveniently used as receivers without significant change to hardware. OCC is also a key component within the IEEE 802.15.7 standard, which focuses on short range wireless optical communication. The aspects such as modulation techniques, data rates, synchronization methods and error correction specific to OCC, are also included in this revised standard. It is worth noting that an accurate and reliable model is an essential piece of a physical layer communication analysis and simulation. However, current theoretical model for OCC is mostly based on general OWC channel model while ignoring the characteristic of image receivers. Therefore, in this research, a dedicated OCC model has been extensively investigated to solve this problem. Firstly, a comprehensive OCC system model at the pixel level incorporating the imaging receiver and considering both line-of sight (LOS) and non-line-of-sight (NLOS) links has been established in Chapter 3. The characteristics of camera, such as camera distortion and receiver noise have been explored and incorporated into the designed camera model. In addition, we adopt stereo vision and P3P algorithms to verify the precision of our proposed model, especially whether the captured image can accurately represent the relationship of spatial position between the LEDs and the receiver camera. After that, an outdoor OCC based vehicular communication model has been established by incorporating solar radiation noise model for outdoor wireless channel in Chapter 5. Results show that solar radiation noise has great impact on the outdoor vehicular communication system, leading to an increase BER in the received signal. Consequently, the transmission distance of outdoor OCC systems is significantly reduced. In addition, the impacts of modulation scheme and modulation depth are also investigated. In addition to the communication capability, the OCC can also provide the indoor localization function, which is also highly demanded recently, due to their widely applications in navigation, Internet of Things (IoT) and wireless communications. Conventionally, radio frequency (RF) based methods are widely used for the indoor localization, but they normally suffer from signal interference and multipath effects, which highly limits the achievable accuracy. Thus, OCC based indoor positioning approach has been proposed to overcome these limitations. There are several traditional positioning algorithms commonly used in OCC system, including received signal strength (RSS), time-of-arrival (ToA) and angle-of arrival (AOA). Besides, machine learning based image processing algorithm has also attracted intensive interests to solve this problem. In OCC system, the accuracy of the indoor positioning system is significantly influenced by the orientation of the receiver, as it affects how the transmitter is imaged within the field of view (FOV). However, current OCC based localization schemes mostly assume the receiver and transmitter planes are parallel, which ignore the critical tilt angle of the receiver in practical scenarios. Furthermore, previous OCC-based positioning techniques usually employed the central location of the LED image, ignoring the shape of the LED image, which can provide information about both orientation and distance. This is due to the difficulty of theoretically modelling and analyzing the shape information. These observations mentioned above have motivated the proposal of a novel indoor localization scheme, which can achieve accurate estimation for both position and orientation. In Chapter 4, a novel OCC based indoor localization scheme has been proposed. Results show that our proposed method can realize an average Euclidean distance of 0.0571m and an average quaternion angle error of 0.5038°, thus accurate simultaneous position and orientation estimation has been achieved. Furthermore, the impacts of pretrained convolutional neural network (CNN) models and dataset size on the performance of the proposed method are also investigated. Results show that the GoogLeNet based CNN model with the dataset size of 40000 has the best performance in the localization system. In addition, the proposed scheme has been verified can effectively reduce the number of LEDs needed to be captured, thus can improve the working area of the localization system.