摘要
激光雷达图像的目标识别依赖于图像本身所反映的目标信息,由于原始激光雷达图像中存在着一定的噪声,导致图像不能清晰反映边缘特征,为了获取准确的目标特征信息,研究基于中值滤波的激光雷达图像边缘特征提取方法。定标处理激光雷达图像,避免出现干涉相位解缠绕的现象;基于中值滤波消除图像噪声,保留边缘特征信息;通过激光雷达图像边缘检测,获取图像中每个位置中心像素的梯度;提取图像边缘特征,融合探测器算子,完成特征提取。经实验论证分析,以PCA+SVM方法、NFM+HMM方法和LDA+SVM方法作为对照,本方法能够有效排除噪声干扰,在降噪处理和边缘特征保留之间取得了较好的平衡;在不同噪声条件下的特征定位误差均未超过0.10像素,精度较高,同时在边缘特征提取时间中的优势最为明显,证明本方法具有可行性。
The target recognition of lidar image relies on the target information reflected by the image itself,due to the existence of certain noise in the original lidar image,resulting in the image cannot clearly reflect the edge features,in order to obtain accurate target feature information,the edge feature extraction method of lidar image based on medi-an filtering is studied.The laser radar image is calibrated to avoid interference phase unwinding.Based on median fil-tering,image noise is eliminated and edge feature information is retained.Through LiDAR image edge detection,the gradient of the center pixel of each position in the image is obtained.Image edge features are extracted,and detector operators are fused to complete feature extraction.Experimental analysis shows that PCA+SVM,NFM+HMM and LDA+SVM can effectively eliminate noise interference and achieve a good balance between noise reduction and edge feature retention.Under different noise conditions,the feature location error is less than 0.10 pixel,with high accuracy,and the edge feature extraction time is the most obvious advantage,which proves the feasibility of the proposed method.
作者
陈初侠
丁勇
CHEN Chuxia;DING Yong(School of Electronic Engineering,Chaohu University,Hefei 238000,China;School of Micro-Nano Electronics,Zhejiang University,Hangzhou 310000,China)
出处
《激光杂志》
CAS
北大核心
2023年第8期94-98,共5页
Laser Journal
基金
国家重点研发计划(No.2019YFA0110400)。
关键词
中值滤波
定标处理
激光雷达图像
边缘检测
边缘特征提取
图像噪声
median filtering
calibration processing
lidar image
edge detection
edge feature extraction
image noise
