基于人工智能的地形目标判读  

Terrain Target Interpretation Based on Artificial Intelligence

作　　者：汪莉莉[1] 吴帅 WANG Lili;WU Shuai(Naval University of Engineering,Wuhan 430033;No.96823 Troops of PLA,Kunming 650219)

机构地区：[1]海军工程大学,武汉430033 [2]96823部队,昆明650219

出　　处：《舰船电子工程》2020年第11期148-152,共5页Ship Electronic Engineering

摘　　要：地形目标判读是地形分析的前提和基础。论文提出了一种基于Mask R-CNN模型的地形目标判读方法。选择模拟大学场坪和居民地的高精度遥感图像作为研究对象,对提出的方法进行了初步验证。实验结果表明,该方法能较好地分类、定位和分割出地形目标,便于地形量化分析。该方法适用于公共场坪、居民地、工厂、水源、交通设施等多种地形目标,适用于可见光、红外、多光谱、激光雷达、SAR等多种图像,而且适用于机场、港口、要塞等多种重点目标判读,对于减轻人员工作量,推动地形和重点目标判读的自动化、智能化发展具有重要意义。Terrain target interpretation is the premise and foundation of terrain analysis.This paper presents a method of ter⁃rain target interpretation based on Mask R-CNN model.The high-precision remote sensing images of simulated university fields and residential areas are selected as the research objects,and the proposed method is preliminarily verified.The experimental results show that this method can classify,locate and segment terrain targets well,which is convenient for terrain quantization analysis.The method is suitable for many kinds of terrain targets such as public area,residential area,factory,water source,traffic facili⁃ties,and for many kinds of images such as visible light,infrared,multi-spectrum,lidar and SAR,and applies to the airport,port,fort and other key target interpretation.This method is of great significance in reducing the workload of personnel and promoting the development of automatic and intelligent interpretation of terrain and key targets.

关 键 词：人工智能 Mask R-CNN 地形目标 判读 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]