摘要
针对地面作战军用机器人非结构化道路识别,提出一种改进的具有生物学背景的交叉视觉皮质模型(ICM)方法,将非结构化道路分分割为道路区域和非道路区域。首先将非结构化道路图像从RGB颜色空间变换到HSI空间模型,并对HIS三个分量图像进行了快速中值滤波及直方图均衡化等预处理。然后对预处理后的图像通过改进的交叉视觉皮质模型(ICM)进行道路分割。并采用二维最小Tsallis交叉熵作为分割准则,可自动地确定分割阈值以及循环迭代次数,避免了人为干预,提高了分割速度。通过实验分析,上述方法具有与传统的图像分割算法相比,具有精度高、适应性强、分割速度快、分割质量接近最佳的特点。
In unstructured lane identification for ground combat robot, an improved Intersecting Cortical Model(ICM) in the biologic back ground is proposed. The preprocessed road images were segmented into the road and non-road regions by using improved ICM. Firstly, we converted unstructured road images into HSI color space, then H, S, and I components were preprocessed through histogram equalization and fast median filter. Then, the preprocessed road images were segmented by using the Intersecting Cortical Model(ICM).The decision mechanism of the 2-D minimum Tsallis cross-entropy was introduced to determine the cyclic iterative times and the best segmentation threshold automatically in order to avoid human intervention and improve the segmentation speed. Based on the experimental results, the proposed algorithm, compared with the traditional segmentation algorithms, has the characteristics of high precision, strong adaptability, high segmentation speed and nearly optimal segmentation quality.
作者
李郁峰
李魁武
潘玉田
郭保全
LI Yu-feng;LI Kui-wu;PAN Yu-tian;GUO Bao-quan(Institute for Civ-Mil Integration&Collaborative Innovation,North University of China,Taiyuan Shanxi 030051,China)
出处
《计算机仿真》
北大核心
2020年第10期307-311,共5页
Computer Simulation
关键词
非结构化道路
交叉视觉皮质模型
道路分割
unstructured lane
intersecting cortical model(ICM)
lane segmentation
