A team of researchers at IIT Roorkee has developed an efficient architecture and algorithm to enable better driving experience and lower the risks of accidents in the low visibility scenario. This research has been published in the IEEE Transactions on Intelligent Transportation Systems.
The presence of fog decreases the visibility distance exponentially, thus making it one of the most dangerous weather conditions for driving. The drivers tend to overestimate the visibility distance while travelling in fog conditions and drive with excessive speed. Also, different components of an automatic driver assistance system (ADAS) such as blind-spot detection, lane departure warning, and collision warning require clear image data. Therefore, every year hundreds of vehicle accidents are reported that claim thousands of lives.
“The objective of this research was to design a system for real-time defogging that produces a clear image stream from the input foggy frames. In addition to that, a high frame rate is essential in transportation to avoid frame lag or drop. If a vehicle is running at 110 km/h and 5 frames per second ADAS is adopted, the vehicle will travel 21 feet distance before the system reacts (processing/thinking time), whereas for 60 frames per second ADAS the reaction distance reduces to 2 feet. To achieve a high frame rate at high resolution, dedicated video defogging hardware is required. However, effective mapping of an algorithm to the dedicated hardware for real-time processing is non-trivial,” said Prof Brajesh Kumar Kaushik – Department of Electronics and Communication Engineering.
“Operations such as exponential function, floating-point multiplication and division, full image buffer, data transactions between the processor and dynamic random access memory (DRAM), deteriorate the performance. To circumvent these challenges, they developed an efficient method and architecture for real-time video defogging that are co-designed to attain high performance and image restoration quality while reducing the power and memory requirements,” he added.
For conducting this research, the team used several standard foggy datasets containing a variety of fogs from light to dense. They devised an approach for atmospheric light and transmission map estimation that is suitable for high-speed parallel hardware such as field-programmable gate array (FPGA). They also utilized a flicker reduction technique to ensure temporal consistency among the video frames. To validate the design, they have used the Xilinx FPGA development kit and FMC card.
“The collision of vehicles due to poor visibility caused by fog leads to numerous fatalities annually. This advanced defogging system will aid drivers by providing real-time information and minimize the risks of road and train mishaps due to fog,” said Prof Ajit K Chaturvedi, Director, IIT Roorkee.
The research team comprised three members namely Prof Brajesh Kumar Kaushik, Department of Electronics and Communication, IIT Roorkee, Prof Balasubramanian Raman, Department of Computer Science, IIT Roorkee, and Rahul Kumar, Department of Electronics and Communication, IIT Roorkee.
The team is also working on prototyping and commercializing this technology.