Reference test system for machine vision used for ADAS functions 2020-01-0096
LDW and LKA systems have the potential to prevent or mitigate 483,000 crashes in the United States every year which includes 87,000 nonfatal injury crashes and 10,345 fatal crashes. Studies have shown that fatalities due to unintentional roadway departures can be significantly reduced if Lane Departure Warning (LDW) and Lane Keep Assist (LKA) systems are used effectively. While LDW and LKA technologies are available, there has been low customer acceptance and penetration of these technologies. These deficiencies can be traced to the inability of many of the perception systems to consistently recognize lane markings and localize the vehicle with respect to the lane marking in a real-world environment of variable markings, changing weather and occlusions. These challenges translate to (i) inconsistent lane detection; (ii) misidentification of lane markings; and (iii) the inability of the systems to locate lane markings in some conditions. Currently, there is no available standard or benchmark to evaluate the quality of either the lane markings or the perception algorithms. This project seeks to establish a reference test system that could be used by transportation agencies to evaluate the quality of their markings to support ADAS functions that rely on pavement markings. The test system can also be used by system designers as a benchmark for their proprietary systems. The reference test system is comprised of a set of test scenarios, defined by roadway and environmental features, as well as pavement marking presence and luminance variables. To support the development of the system, an extensive video dataset was collected at different times of day and weather conditions on various roads in Central Texas. The videos were evaluated on different state-of-the-art lane detection algorithms and their performance was ranked based on a set of metrics specifically developed for evaluating the effectiveness of lane estimation system. A systems approach is presented by correlating the algorithm performance data to the type, marking color, marking material, and the retroreflectivity of pavement markings. Using the results obtained, a reference Lane Detection (LD) system is proposed to benchmark and rank new perception algorithms, sensors, and lane markings that constitute a reference lane system.
Abhishek Nayak, Sivakumar Rathinam, Adam Pike, Swaminathan Gopalswamy