【摘要】 信息技术与通讯技术的发展促进了智能交通系统的变化,一方面可以有效地辅助驾驶员驾驶,另一方面可以促进无人驾驶系统的发展。智能系统的安全系数成为了考量发展技术水平的关键。提出了一种基于RGB空间与曲率的交通灯检测技术。首先为了抵抗噪声对于颜色检测结果的影响,将RGB空间转换到归一化的RGB空间提取候选区域;然后利用交通灯等规则的形状,根据曲率值提取圆形,得到最后的交通灯位置和颜色信息;最后,通过语音输出将有效的信息反馈给驾驶员或无人驾驶系统中的信息处理块。实验部分选择Hough变换(HT)、Radon变换(RT)、自组织模糊规则(SOFRS)为对比算法。对50幅交通灯图像的检测结果表明:HT准确率66%、RT准确率80%、SOFRS准确率76%,而新算法准确率高达96%。运行时间上,HT耗时0.561 s、RT耗时0.983 s,SOFRS耗时0.793 s而新算法仅耗时0.073 2 s。因此,新算法有效且快速。更多还原

【Abstract】 the development of information technology and communication technology provide good support for the intelligent traffic system.On the first hand,it can assist the divers’ driving;on the other hand,it can help the development of Non-driver system.The security of the intelligent traffic system is a crucial component as it decides whether it is possible to make it practice.A method is proposed based on RGB value and curvature.First,the traditional RGB value is converted into normalized RGB value to avoid the effect of the noises,and then the regular shape of the traffic light is used to detect the position based on the compactness and finally the position and color information is got.As the finale,all the position and color information will be output to the users,either the diver or the non-driver system.Experiments chose Hough Transform(HT),Radon Transform(RT),and Self-Organized Fuzzy Rule System(SOFRS) as the comparative algorithms.The results on fifty traffic light images show that for detection accuracy,HT obtains 66%,RT 80%,SOFRS 76%,and the new algorithm as high as 96%;for computation time,HT costs 0.561 s,RT0.983 s,SOFRS 0.793 s,and the new algorithm merely 0.073 2 s.Therefore, our algorithm is effective and fast.更多还原