摘要
近年来无人机航拍技术逐步应用于野生动物保护,在很大程度上提高了考察效率。由于航拍图像与地面拍摄图像的特征差异较大,加之野生动物生存环境背景复杂,目前没有通用的方法可直接应用于野生动物航拍图像的检测与统计。本文回顾了智能检测和统计技术近年来的发展,针对无人机航拍野生动物图像的大场景、小目标、多尺度、复杂背景等特点,介绍了无人机航拍动物群数据集的选取与建立方法,以及基于深度学习的检测与统计方法,并进行了深层次地分析,归纳了各类方法的优势和可应用场景,总结了各方法的特点和适用范围,同时针对存在的问题给出了改进方向。
Recently,UAV aerial photography technology has been gradually applied to wildlife protection,which has greatly improved the investigation efficiency.Due to the great difference in characteristics between aerial images and ground images,and the complex background of wildlife living environment,there is no general method that can be directly applied to the detection and statistics of UAV aerial wildlife photography.In this paper,firstly,the development of intelligent detection and statistics technology in recent years is reviewed.Then,according to the characteristics of large scene,small target,multi scale and complex background of UAV aerial wildlife photography,the selection and establishment methods of UAV aerial wildlife dataset is introduced,and the detection and statistics methods based on deep learning as well.Finally,the advantages and applicable scenes of these methods are summarized,and the improvement direction is given.
作者
祝宁华
郑江滨
张阳
ZHU Ninghua;ZHENG Jiangbin;ZHANG Yang(Engineering Practice Training Center,Northwestern Polytechnical University,Xi’an 710072,China;School of Software,Northwestern Polytechnical University,Xi’an 710072,China;Xi’an Modern Control Technology Research Institute,Xi’an 710065,China)
出处
《航空工程进展》
CSCD
2023年第1期13-26,共14页
Advances in Aeronautical Science and Engineering
基金
陕西省重点研发计划(2021ZDLGY09-08)。
关键词
无人机航拍
生态保护
深度学习
迁移学习
目标检测
数量统计
UAV aerial photography
ecological protection
deep learning
transfer learning
object detection
statistics