基于曲率滤波和反向P-M电动车充电孔检测方法  

Detection method for electric vehicle charging hole based on curvature filter and inverse P-M diffusion

作　　者：张辉[1,2] 金侠挺 

机构地区：[1]长沙理工大学电气与信息工程学院,长沙410012 [2]湖南大学电气与信息工程学院,长沙410082

出　　处：《仪器仪表学报》2016年第7期1626-1638,共13页Chinese Journal of Scientific Instrument

基　　金：国家自然科学基金(61401046);国家科技支撑计划(2015BAF11B01);湖南省自然科学基金(13JJ4058);湖南省教育厅科学研究青年项目(13B135);图像测量与视觉导航湖南省重点实验室开放课题(TXCL-KF2013-001);长沙市科技计划项目(K1404019-11)资助

摘　　要：为了解决电动车人工充电作业的效率低下、场合限制、漏电隐患等难题,研究了基于曲率滤波和反向P-M扩散的电动车充电孔检测与定位方法,该方法能够从充电座图像中精确高效地提取目标物充电孔的特征,实现机器人的自动化充电。针对图像信号受强电磁干扰,采用了具有噪点识别能力和边缘保持特性的曲率滤波法进行图像除噪;由于充电座图像背景复杂、亮度不均以及干扰项多,导致常规分割方法失效,故研究了反向P-M扩散和BP神经网络(BPNN)相结合的分割方法。预处理图像经反向P-M扩散、差分运算、滤波及空洞填充得到P-M精确分割的连通域,再利用形态学击中算法,在P-M连通域中搜寻由BPNN粗略提取的连通域,从而得到精确的充电孔目标区域。最后,对所提出的方法进行验证测试,实验结果表明该方法能有效识别充电孔,并且检测速度、定位精度很好地满足充电机器人作业要求。A new method based on curvature filter and inverse P-M diffusion is proposed for electric vehicle charging hole detection and location. As a result,the automatic charging for robots can be achieved in order to improve the low efficiency,limits of application and possible leakage for manual charging. The method extracts the features efficiently and accurately from the images of charging hole.Firstly,the curvature filter is adopted for noise-filtering and detail-preserving to overcome the electromagnetic interference to image signal. Then,the charging socket image shows illumination variation,reflection inequality and interference factor,which brings great challenge to normal segmentation methods. Therefore,a segmentation scheme combined inverse P-M diffusion with BP neural network（ BPNN）） is studied. The pre-processed image is processed with the inverse P-M diffusion,difference operation,filtering and hole filling,to obtain the P-M precisely segmented connected domain. Using morphological hit algorithm,the target can be extracted accurately with the rough BPNN connected domain. Finally,the experimental results indicate that the method can recognize charging hole and satisfy the requirements of robot-based recharge both in detection speed and location accuracy.

关 键 词：新能源电动车 机器人充电 充电孔视觉检测与定位 曲率滤波 反向P-M扩散 BP神经网络 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术] TH89[自动化与计算机技术—计算机科学与技术]