光学图像和激光雷达在建筑物分类中的应用  

Application of optical image and laser radar in classification of buildings

作　　者：石东[1] 

机构地区：[1]西安石油大学机械工程学院,西安710065

出　　处：《激光杂志》2016年第11期145-149,共5页Laser Journal

基　　金：陕西自然科学基金(9201444)

摘　　要：通过对激光雷达成像的光学图像分析,实现地面建筑物的识别和分类,面临抗噪性能差,特征可变性强等问题。提出一种基于自适应局部噪声滤波和仿射不变矩特征的建筑物激光雷达光学成像分类识别算法,首先构建激光雷达光学成像的采集模型,采用横向扫描、块扫描方法进行图像的特征点扫描,根据自适应局部噪声滤波算法进行光学图像的降噪提纯处理,提取滤波输出图像的仿射不变矩特征,采用BP神经网络分类器进行特征分类,实现建筑物激光雷达光学成像的分类识别。仿真结果表明,采用该方法进行光学图像的降噪滤波以及特征提取的准确性较好,输出图像的峰值信噪比较高,对建筑物分类的精度较高,计算开销小于传统方法,误分率等各项指标性能优越。Analysis of optical images of laser radar imaging, recognition ground, will have a good application value in the construction planning and and classification of the buildings on the geographical exploration etc.. The traditional method using laser remote sensing feature fusion method for building the feature extraction and analysis, when the optical image acquisition of the interference is large, and the SNR is relatively low, effect of construction effect is not good, classification accuracy is not high. A based on local adaptive noise filtering and affine moment invariant fea- ture extraction buildings laser radar optical imaging classification and recognition algorithm is proposed. Firstly, we build the imaging laser radar optical acquisition model, with Hilbert scanning method for the feature point of image scanning, optical image de - noising and purified by using local adaptive noise filtering algorithm, extraction filtering the output image affine invariant moment features, feature classification using BP neural network classifier, building laser radar optical imaging classification and recognition. Simulation results show that using the method of optical image denoising and feature extraction of good accuracy, high output image peak signal to noise, high precision of building classification, the computational overhead is smaller than that of the traditional method, the indicators of performance advantages.

关 键 词：光学图像 建筑物 激光雷达 降噪滤波 仿射不变矩 

分 类 号：TN27[电子电信—物理电子学]