摘要
车位增强算法作为自动泊车的重要组成部分,其增强结果直接影响车位线提取效果。基于此,引入了暗通道作为低频分量进行自适应对比度增强,基于多组低对比度车位图像数据,讨论了多种低对比度增强算法的适用性,针对增强算法部分区域产生的车位增强不足与过曝现象导致降低车位线提取的完整性与精度下降的问题,提出峰值直方图均衡化的快速增强算法,结合了PSNR、结构相似性、平均亮度和信息熵等作为客观评价指标,利用了霍夫直线检测统计算法增强结果的车位提取精度,并进行了验证。研究结果表明:本算法能够减少环境信息干扰,保留更多纹理细节,提升全局图像亮度与对比度,其在低照度环境下仍然具有出色的鲁棒性。本算法车位线提取精度超过90%,算法运行时间仅为37.18 ms,能够为低对比度场景下的自动泊车系统提供方法指导。
The parking space enhancement algorithm is an important part of automatic parking,and its enhancement result directly affects the extraction effect of the parking space line.Based on this,this paper introduces the dark channel as a low-frequency component for adaptive contrast enhancement.Based on multiple sets of low-contrast parking space image data,the applicability of various low-contrast enhancement algorithms is discussed.The exposure phenomenon leads to the problem of reducing the integrity and accuracy of parking space line extraction.A fast enhancement algorithm for peak histogram equalization is proposed,which combines PSNR,structural similarity,average brightness and information entropy as objective evaluation indicators,and uses Hough The linear detection statistical algorithm enhances the accuracy of the parking space extraction results and is verified.The research results show that the algorithm in this paper can reduce the interference of environmental information,retain more texture details,improve the brightness and contrast of the global image,and it still has excellent robustness in low-light environments.The algorithm in this paper has a parking space line extraction accuracy of more than 90%,and the algorithm running time is only 37.18 ms,which can provide method guidance for automatic parking systems in low-contrast scenes.
作者
苗作华
刘代文
尹东
李诒雯
陈澳光
MIAO Zuohua;LIU Daiwen;YIN Dong;LI Yiwen;CHEN Aoguang(Wuhan University of Science and Technology,School of Resources and Environmental Engineering,Wuhan 430081,China;Hubei Key Laboratory of Ef icient Utilization and Blocking of Metallurgical Mineral Resources,Wuhan 430081,China;Wuhan Kotei Informatics Co.,Ltd.,Wuhan 430000,China)
出处
《激光杂志》
CAS
北大核心
2024年第7期168-173,共6页
Laser Journal
基金
国家自然科学基金(No.41971237、41071242)
教育部产学合作协同育人项目(No.202102136008)。
