Color and Height Characteristic of Surrogate Grass for the Evaluation of Vehicle Road Departure Mitigation System 2019-01-1026
New vehicles equipped with roadway departure mitigation (RDM) systems have been introduced to the market in recent years. An earlier study showed that 56% of road edges in the U.S are grass. To support the standard objective evaluation of RDM system performance, the development of surrogate road edge grass with the representative property of real grass from the view of automotive sensors is needed. This paper is to describe the process for determining the color and height requirements of surrogate grass for a camera sensor.
824,957 road locations in the U.S. were randomly sampled. Then a stratified sub-sampling was conducted with the balancing consideration, which includes the geographic locations, road levels, and population densities. Finally, 901 randomly selected Google street-view images with grass road edges but without road markings or with poor road markings were gathered and analyzed. Based on the observation of these 901 images, four parameters were identified to describe the visual characteristics of the grass. To determine the grass color and color evenness, a k-means clustering was performed in LUV color space. 6 representative colors with 3 levels of yellow/brown and 3 levels of green were obtained as the reference green and yellow colors. Then, the perspective transformation was adopted to remove perspective distortion. Based on the portion of 6 representative colors in images and Pearson chi-square statistics, the color and color evenness of each image were determined. The height and the height evenness of the grass were determined by manual clicking of the referencing of gathered images with known grass height and height evenness. The estimation results were cross-validated by manual check. With the estimated values of color, color evenness, height, and height evenness, 901 sampled Google images were grouped in multiple levels of abstractions. Eventually, two representative grass types are suggested as required for the surrogate grass development. They are (1) mixed yellow and green color and large random uneven patches grass, and (2) green and even grass.
This paper describes the determination of the representative height and color characteristics of grass on road edges. It is the vision sensor related part of the first step in developing surrogate grass, which is needed for standard performance evaluation of RDM for all levels of vehicle automation.
Qiang Yi, Dan SHEN, Jun Lin, Stanley Chien, Lingxi Li, Yaobin Chen, Rini Sherony
Indiana Univ Purdue Univ Indianapolis, Toyota Motor Corp