摘要
钢轨表面区域提取方法存在需要预先设定轨面宽度的问题,且容易受到光照不均和噪声的干扰,因此,该文提出一种基于HSV色彩空间S分量的轨面区域提取方法。进行自适应加权中值滤波操作,消除杂乱而细小的噪声;将去噪后的RGB轨道图像转化为HSV色彩空间轨道图像,将S空间分量图像单独提取出来,减少光照不均带来的干扰;采用改进的线性函数对S空间分量图像进行增强,提高不同区域的对比度差异;通过绘制灰度投影曲线,搜索列曲线的最大值和次大值,确定轨面区域左右边界,完成轨面区域的提取。实验结果表明:该文算法能够有效地提取钢轨表面区域,提取精度较高,准确率达94%,平均提取时间为22.8 ms,可以满足实时性的需要,具有一定的实用价值。
Existing rail surface area extraction methods have to preset the rail width and are susceptible to uneven lighting and noise.This paper presents a rail surface area extraction method based on the S component of HSV color space.This paper performs adaptive weighted median filtering to eliminate messy and small noise.By converting the denoised RGB rail image into HSV color space rail image,this paper extracts the S component image separately and reduces the interference caused by uneven lighting.This paper uses an improved linear function to enhance the S component image for a high contrast difference of different areas.By drawing the grayscale projection curve and searching for the maximum and secondary maximum of the column curve,this paper determines the left and right boundaries of the rail and extracts it.The simulation experiment result shows that the proposed method demonstrates highly competitive rail surface area extraction performance,achieving 92%precision and 22.8 ms extracting time on average.
作者
曹义亲
丁要男
Cao Yiqin;Ding Yaonan(School of Software,East China Jiaotong University,Nanchang 330013,China)
出处
《南京理工大学学报》
CAS
CSCD
北大核心
2021年第4期464-471,共8页
Journal of Nanjing University of Science and Technology
基金
国家自然科学基金(61663009)
江西省科技支撑计划重点项目(20161BBE50081)
江西省研究生创新专项资金项目(YC2020-S354)。
作者简介
曹义亲(1964-),男,教授,主要研究方向:图像处理与模式识别,E-mail:yqcao@ecjtu.edu.cn;通讯作者:丁要男(1996-),男,硕士生,主要研究方向:图像处理,E-mail:1357598539@qq.com。
