- 學位論文
VEHICLE DETECTION CONTROL TO SUPPORT ADAPTIVE TRAFFIC LIGHT CONTROL
VEHICLE DETECTION CONTROL TO SUPPORT ADAPTIVE TRAFFIC LIGHT CONTROL
摘要
ABSTRACT Because of the high demand many urban road facilities are frequently oversaturated and congested respectively. Through congestion the capacity of the road infrastructure is in fact reduced and the performance deteriorates considerably during rush hours when the maximum capacity is most urgently needed. Given the demand, the most efficient means to alleviate congestions and its negative impacts on travel times, as well as on environmental pollutions and on traffic safety is to balance the traffic saturation on the different parts of the road network. In urban road networks this goal is to be achieved primarily by the means of adaptive traffic signal light control. This can be achieved by deploying various kinds of sensors for traffic light control. For example: pressure sensors; which reacted to masses that are passing on top of them; magnetic field sensor which responds to objects within its range; videos and cameras for capturing images of both moving cars or stills that can then processed using specific software to be utilized to manage traffic congestion. In this study we explored more on utilizing still images captured by digital camera as sensor. We will use MATLAB software as tool for image processing to control traffic congestion as a mean to support the adaptive traffic light control. Two techniques for extraction of information from traffic images were developed – the first technique is based on histogram readings to measure the queue length, and second technique by counting cars requiring that the presence of vehicles in any part of the image be labeled and distinguished from the background scene. These approaches improve traffic efficiency of preinstalled traffic light controller through adaptive dynamic timing according to traffic condition. We found that noises are present during dusk, and reduces accuracy. We believe that such problem could be reduced by technical refinement of our syntaxes i.e. during the threshold step. Simulations results show that using the described approaches improve measurable traffic efficiency. We suggest that these approaches should be developed in term of technical refinement i.e. automated threshold technique. Simulation results show that these approaches have a very good prospect in the future, and should be further tested and developed for real traffic light operation. DGE Nokia – Siemens in particular has expressed their interest in implementing them by customizing their traffic light microcontroller to fit for these approaches.
關鍵字
HASH(0xb5fdf00)並列摘要
ABSTRACT Because of the high demand many urban road facilities are frequently oversaturated and congested respectively. Through congestion the capacity of the road infrastructure is in fact reduced and the performance deteriorates considerably during rush hours when the maximum capacity is most urgently needed. Given the demand, the most efficient means to alleviate congestions and its negative impacts on travel times, as well as on environmental pollutions and on traffic safety is to balance the traffic saturation on the different parts of the road network. In urban road networks this goal is to be achieved primarily by the means of adaptive traffic signal light control. This can be achieved by deploying various kinds of sensors for traffic light control. For example: pressure sensors; which reacted to masses that are passing on top of them; magnetic field sensor which responds to objects within its range; videos and cameras for capturing images of both moving cars or stills that can then processed using specific software to be utilized to manage traffic congestion. In this study we explored more on utilizing still images captured by digital camera as sensor. We will use MATLAB software as tool for image processing to control traffic congestion as a mean to support the adaptive traffic light control. Two techniques for extraction of information from traffic images were developed – the first technique is based on histogram readings to measure the queue length, and second technique by counting cars requiring that the presence of vehicles in any part of the image be labeled and distinguished from the background scene. These approaches improve traffic efficiency of preinstalled traffic light controller through adaptive dynamic timing according to traffic condition. We found that noises are present during dusk, and reduces accuracy. We believe that such problem could be reduced by technical refinement of our syntaxes i.e. during the threshold step. Simulations results show that using the described approaches improve measurable traffic efficiency. We suggest that these approaches should be developed in term of technical refinement i.e. automated threshold technique. Simulation results show that these approaches have a very good prospect in the future, and should be further tested and developed for real traffic light operation. DGE Nokia – Siemens in particular has expressed their interest in implementing them by customizing their traffic light microcontroller to fit for these approaches.