基于RGB熵和改进区域生长的非结构化道路识别方法

doi:10.13229/j.cnki.jdxbgxb20171257

摘要/Abstract

摘要：

非结构化道路通常没有车道线等引导标线，且边界模糊，周围干扰因素较多，在这种道路上自动驾驶以及车辆辅助驾驶的车道保持功能将不能可靠工作。通过生成道路RGB图像的熵图像，并计算此熵图像直方图的最小差值，以此差值作为阈值初步分割道路图像并使用改进区域生长方法提取出道路区域。使用实时性较好的二次曲线建立车道模型，并使用改进的最小二乘拟合方法可有效避开道路区域边缘杂点并提高边缘拟合速度。试验结果表明，改进的方法可以快速并较好地提取出非结构化道路图像中的车道并拟合出车道线，有利于实现基于视觉的自动驾驶和车辆辅助驾驶系统在非结构化道路上的车道保持。

关键词: 交通信息工程及控制, 非结构化道路检测, RGB图像熵, 区域生长方法, 二次曲线模型

Abstract:

Unstructured roads usually do not have leading lines such as lanes, and the road boundary is vague and there exist many interference factors around the roads. Therefore it is not reliable to use the lane keeping system for autopilot and driver assistance system in this kind of roads. By generating the entropy image from original image the minimum difference of histogram of the entropy image is computed, and the difference is used as the segmental threshold to preliminarily segment the image. Then, the improved region growing method is used to extract the lane from the preliminarily segmented image. The real?time quadratic curve is used to establish the lane model, and the improved least square fitting method is applied to effectively avoid noisy points on the edges of lane region and promote the fitting speed. The experimental results show that the improved methods can be used to rapidly extract the lane from road image and fit out the lane line, therefore it helps to achieve visual based lane keeping on unstructured road for autopilot and driver assistance system.

Key words: traffic information engineering and control, unstructured road detection, RGB image entropy, region growing method, quadratic curve model

中图分类号:

U495

吴骅跃,段里仁. 基于RGB熵和改进区域生长的非结构化道路识别方法[J]. 吉林大学学报(工学版), 2019, 49(3): 727-735.

Hua⁃yue WU,Li⁃ren DUAN. Unstructured road detection method based on RGB entropy and improved region growing[J]. Journal of Jilin University(Engineering and Technology Edition), 2019, 49(3): 727-735.

图/表 11

图1

图2

图3

图4

图5

图6

图7

图8

表1

表2

表3

参考文献 16

1	熊思, 李磊民, 黄玉清. 基于小波变换和K⁃means的非结构化道路检测[J]. 计算机工程, 2014, 40(2): 158⁃161.
	XiongSi, LiLei⁃min, HuangYu⁃qing. Unstructured road detection based on wavelet transform and K⁃means[J]. Computer Engineering, 2014, 40(2): 158⁃161.
2	郭秋梅, 黄玉清. 基于二维熵和轮廓特征的非结构化道路检测[J]. 计算机应用, 2013, 33(7): 2005⁃2008.
	GuoQiu⁃mei, HuangYu⁃qing. Unstructured road detection based on two⁃dimensional entropy and contour features[J]. Journal of Computer Applications, 2013, 33(7): 2005⁃2008.
3	WangY, TeohE K, ShenD G. Lane detection and tracking using B⁃Snake[J]. Image and Vision Computing, 2004, 22(4): 269⁃280.
4	JeongH, OhY, ParkJ H, et al. Vision⁃based adaptive and recursive tracking of unpaved roads[J]. Pattern Recognition Letters, 2002, 23(1⁃3): 73⁃82.
5	周植宇, 杨明, 薛林继, 等. 一种基于高斯核支持向量机的非结构化道路环境植被检测方法[J]. 机器人, 2015, 37(6): 702⁃707.
	ZhouZhi⁃yu, YangMing, XueLin⁃ji, et al. Vegetation detection approach based on gaussian kernel support vector machine in unstructured road environment[J]. Robot, 2015, 37(6): 702⁃707.
6	柏猛, 李敏花. 基于图模型的道路检测方法[J]. 模式识别与人工智能, 2014, 27(7): 655⁃662.
	BaiMeng, LiMin⁃hua. Road detection method based on graph model[J]. Pattern Recognition and Artificial Intelligence, 2014, 27(7): 655⁃662.
7	龚建伟, 叶春兰, 姜岩, 等. 多层感知器自监督在线学习非结构化道路识别[J]. 北京理工大学学报, 2014, 34(3): 261⁃266.
	GongJian⁃wei, YeChun⁃lan, JiangYan, et al. Unstructured road recognition using self⁃supervised multilayer perceptron online learning algorithm[J]. Transactions of Beijing Institute of Technology, 2014, 34(3): 261⁃266.
8	ZhouS Y, GongJ W, XiongG M, et al. Road detection using support vector machine based on online learning and evaluation[C]∥IEEE Intelligent Vehicles Symposium, San Diego, CA, USA, 2010: 256⁃261.
9	兰丽, 胡晓辉, 孙苗强. 一种基于灰关联和Sobel算子的完全非结构化道路边缘检测方法[J]. 湖南工程学院学报, 2013, 23(1): 42⁃44.
	LanLi, HuXiao⁃hui, SunMiao⁃qiang. Completely unstructured road edge detection based on grey relation and sobel operator[J]. Journal of Hunan Institute of Engineering, 2013, 23(1): 42⁃44.
10	吕艳鹏, 潘玉田. 基于MATLAB图像处理的非结构化道路识别[J]. 机械工程与自动化, 2014, 43(2): 32⁃33.
	LvYan⁃peng, PanYu⁃tian. Unstructured road detection based on MATLAB image processing[J]. Mechanical Engineering & Automation, 2014, 43(2): 32⁃33.
11	李进, 陈杰平, 易克传, 等. 基于特征融合的视觉导航智能车辆的车道识别[J]. 汽车工程, 2015, 37(5): 587⁃592.
	LiJin, ChenJie⁃ping, YiKe⁃chuan, et al. Lane identification of vision⁃guided intelligent vehicle based on feature fusion[J]. Automotive Engineering, 2015, 37(5): 587⁃592.
12	许华荣, 王晓栋, 方遒. 基于B样条曲线模型的结构化道路检测算法[J]. 自动化学报, 2011, 37(3): 270⁃275.
	XuHua⁃rong, WangXiao⁃dong, FangQiu. Structure road detection algorithm based on B⁃spline curve model[J]. Acta Automatica Sinica, 2011, 37(3): 270⁃275.
13	ChiuK Y, LinS F. Lane detection using color⁃based segmentation[C]∥IEEE Intelligent Vehicles Symposium, Las Vegas, USA, 2005: 706⁃711.
14	颜飞,周长久,田彦涛. 用于目标定位的图像边缘点检测算法[J]. 吉林大学学报：工学版,2016,46(6):2103-2110.
	Yanfei,zhouchang-jiu,tianyan-tao. Image edge points detection algorithm for object localization[J]. Journal of Jilin University(Engineering and Technology Edition), 2016,46(6):2103-2110.
15	肖明尧,李雄飞,张小利,等. 基于多尺度的区域生长的图像分割算法[J]. 吉林大学学报：工学版,2017,47(5):1591-1597.
	XiaoMing-yao,LiXiong-fei,ZhangXiao-li,et al. Medical image segmentation algorithm based on multi-scale region growing[J]. Journal of Jilin University(Engineering and Technology Edition), 2017,47(5):1591-1597.
16	武历颖, 余强. 一种快速准确非结构化道路检测方法研究[J]. 计算机仿真, 2016, 33(9): 174⁃178.
	WuLi⁃ying, YuQiang. A fast and accurate detection method of unstructured road[J]. Computer Simulation, 2016, 33(9): 174⁃178.

Metrics

Viewed

Full text

Abstract

Cited

Shared

Discussed

帧率/(帧·s^?1)	总帧数	总生成时间/ms	每帧平均时间/ms
20	6000	144 215	24
	12000	287 547	23
	36000	904 671	25
25	7500	157 614	21
	15000	344 752	22
	45000	1080 418	24
30	9000	198 406	22
	18000	414 057	23
	54000	1243 259	23
35	10500	252 467	24
	21000	483 197	23
	63000	1574 945	25

帧率/(帧·s^?1)	总帧数	总识别时间/ms	每帧平均时间/ms
20	6 000	150 347	25
	12 000	299 679	25
	36 000	934 249	25
25	7 500	165 207	22
	15 000	359 814	24
	45 000	1125 574	25
30	9 000	207 581	23
	18 000	426 903	23
	54 000	1297 433	24
35	10 500	253 108	24
	21 000	504 268	24
	63 000	1633 869	25

车速/(km·h^?1)	帧率/(帧·s^?1)	总帧数	完成帧数	完成率 /%	识别间隔距离/(m·帧^?1)
20	15	18 000	18 000	100	0.37
	20	24 000	24 000	100	0.27
	25	30 000	30 000	100	0.22
40	20	24 000	24 000	100	0.55
	25	30 000	30 000	100	0.44
	30	36 000	36 000	100	0.37
60	25	30 000	30 000	100	0.66
	30	36 000	36 000	100	0.55
	35	42 000	41 271	98	0.47
80	30	36 000	36 000	100	0.74
	35	42 000	40 834	97	0.63
	40	48 000	45 273	94	0.55

[1]	陶涛，徐洪泽. 高速列车浸入与不变自适应容错控制方法[J]. 吉林大学学报(工学版), 2015, 45(2): 554-561.
[2]	陈强，李江，吴想，闫松申. 轮胎印痕识别算法及实例分析[J]. 吉林大学学报(工学版), 2005, 35(01): 39-0043.