电子学报 ›› 2019, Vol. 47 ›› Issue (2): 322-330.DOI: 10.3969/j.issn.0372-2112.2019.02.010

• 学术论文 • 上一篇    下一篇

基于随机投影深度函数的停车场车辆提取方法

李玉1, 王亚琼1, 赵雪梅2, 赵泉华1   

  1. 1. 辽宁工程技术大学测绘与地理科学学院遥感科学与应用研究所, 辽宁阜新 123000;
    2. 中国科学院遥感与数字地球研究所, 北京 100094
  • 收稿日期:2017-08-07 修回日期:2018-06-22 出版日期:2019-02-25
    • 通讯作者:
    • 王亚琼
    • 作者简介:
    • 李玉 男,1963年生于吉林长春,博士,教授,主要研究方向为遥感影像处理;赵雪梅 女,1989年出生于辽宁阜新,博士后,主要研究方向为遥感图像分割及基于深度学习的Landsat图像分类.
    • 基金资助:
    • 国家自然科学基金 (No.41301479,No.41271435); 辽宁省自然科学基金 (No.2015020090)

Extraction of Vehicles in Parking Lot Based on Random Projection Depth Function

LI Yu1, WANG Ya-qiong1, ZHAO Xue-mei2, ZHAO Quan-hua1   

  1. 1. Institute for Remote Sensing Science and Application. School of Geomatics, Liaoning Technical University, Fuxin, Liaoning 123000, China;
    2. Institute of Remote Sensing and Digital Earth of the Chinese Academy of Sciences, Beijing 100094, China
  • Received:2017-08-07 Revised:2018-06-22 Online:2019-02-25 Published:2019-02-25

摘要: 为精确提取露天停车场内颜色混杂的车辆,提出一种基于随机投影深度函数的车辆提取方法.随机投影深度函数可有效区分RGB彩色空间中数据集的数据中心与离群值,充分利用各车辆颜色特征的复杂性及其与停车场背景颜色特征的差异性,凸显具有离群值颜色特征的车辆.首先,利用随机投影深度函数对彩色遥感影像中各像素颜色特征进行排序得到深度场影像;然后,对深度场影像做形态学闭运算并选取合适的随机投影深度值作为阈值,二值化闭运算后的深度场影像,实现车辆初始提取;最后,结合决策树分析与形态学运算实现车辆精确提取.实验结果表明,随机投影深度函数可有效处理彩色遥感影像中各种颜色车辆所表现的"同物异谱"现象,在深度场影像中凸显不同颜色的车辆,有效提高车辆提取效率;辅助以简单的后处理可实现遥感影像中不同场景停车场车辆提取.

关键词: 遥感影像, 随机投影深度函数, 停车场车辆提取, 形态学操作

Abstract: An algorithm of extraction of vehicles based on random projection depth function is proposed for accurately extracting vehicles with different colors in outdoor parking lots.The random projection depth function can effectively distinguish the center and outlier of the data set in RGB color space,and in this way,the vehicles whose color characteristics act as outlier are highlighted.First,the random projection depth function is used to sort the color characteristic of each pixel to obtain random projection depth value,forming the depth field image; Then,morphological closed operation is carried out for the depth field image,and an appropriate random projection depth value is selected as the threshold to binarize the image; Finally,the vehicles are accurately extracted from the parking lot by decision tree algorithm and morphological operations.The experimental results show that the random projection depth function can effectively deal with the "same body with different spectrum" phenomenon of various color vehicles in remote sensing images.The vehicles of different colors are highlighted in the depth field image,which can effectively improve the efficiency of vehicle extraction,and extraction of vehicles from parking lot of remote sensing images can be realized accurately by combining random projection depth function and simple post-processing.

Key words: remote sensing image, random projection depth function, extraction of vehicles in parking lot, morphological operations

中图分类号: