【计算机科学】【2021.05】基于深度学习的航空和卫星图像植物物种检测

本文为俄罗斯西伯利亚联邦大学（作者：Anastasiia Safonova）的博士论文，共80页。

毫无疑问，保护植物物种对于保护自然、气候和人类福祉至关重要。传统上，保护受威胁植物物种的任务通常是通过自然资源管理人员的直接实地监督来完成的。由于传感器技术的进步以及无人机等飞行器的发展，高分辨率遥感数据正成为大面积监测受威胁植物物种的重要资源。这些RS数据通常是多光谱图像，具有三个或更多波段，分辨率高达3cm/像素，提供所考虑的植物物种的正交或准正交视图。这些信息可能不如自然图像提供的信息完整；然而，监测位于难以进入的大面积区域的树种可能就足够了。分析遥感图像的任务通常是使用需要高水平的人工干预的经典算法。

在过去的十年中，深度学习（DL）模型和深度卷积神经网络（CNN）在从自然图像中提取空间模式方面取得了令人瞩目的成果。事实上，CNN构成了最先进的计算机视觉任务，包括图像分类、目标检测和分割。然而，在高分辨率的正交和准正交图像中，深度CNN的潜力还没有得到充分的挖掘，特别是在植物物种保护方面。

这篇论文是第一篇探讨深度CNN、资料预处理和高解析度RS资料在解决植物物种保护问题上的潜力研究。特别地，本文在三个不同的自然科学问题上对深度CNN模型进行了研究和分析：

1.利用DL检测无人机图像中被小蠹虫侵害的冷杉。

2.基于MaskR-CNN的无人机多分辨率图像分割橄榄树生物量估计。

3.利用YOLO结构检测无人机图像中的云杉（Picea Abies）是否感染小蠹虫。

本论文的主要目标是建立一个稳健而精确的动态线性模型，以利用无人机影像来监测不同植物种类。实现的具体目标是：●为所考虑的三个问题中的每一个建立三个高质量的数据集。●设计适当的预处理方法，减少特征和注释中的噪声和不确定性。●本论文前两章的研究结果已发表在JCR的Q1和Q2期刊上。第三章的研究成果已提交给IEEE地球科学与遥感学报。

本文共分为五章。第一章介绍了本文研究的问题、背景和目的。第二章介绍了建立的数据集、检测模型和预处理技术，以解决无人机图像中被小蠹虫侵害的冷杉检测问题。第三章介绍了无人机多分辨率图像分割中橄榄树生物量的数据集、分割模型和预处理技术。第四章给出了利用YOLO架构在无人机图像中检测云杉感染小蠹虫的数据集、模型和预处理技术。最后，第五章是结论与展望。

It is unquestionable that the conservation of plant species is essential for the preservation of nature, climate,and human wellbeing. Classically, the task of conserving threatened plantspecies was generally done through direct field supervision by naturalresources managers. Thanks to the advances in sensor technologies and also inaircrafts such as unmanned aerial vehicles (UAV), high resolution remotesensing (RS) data are becoming an important resource for monitoring threatenedplant species in large areas. Such RS data are usually multi-spectral images,with three or more bands, of up to 3 cm/pixel resolution, providing anorthogonal or quasi-orthogonal view of the considered plant species. Thisinformation may not be as complete as the information provided by naturalimages; however, it might be sufficient to monitor tree species located in verylarge areas with difficult access. The task of analyzing RS images is usuallyperformed using classical algorithms that require a high level of human intervention.In the last ten years, Deep Learning (DL) models in general and deepConvolutional Neural Networks (CNNs) in particular have shown impressive results in extracting spatial patterns from naturalimages. Indeed, CNNs constitute the state-of-the-art inall computer visiontasks, in image classification, object detection, and segmentation.Nevertheless, the potential of deep CNNs have not been fully explored in high resolution orthogonal and quasi-orthogonal images, especially in plant speciesconservation. This thesis presents one of the first studies in exploring the potential of deep CNNs, data preprocessing and high resolution RS data, in addressing plant species conservation problems. In particular, this thesis presents the results and analysis of deep CNNmodels in three different problems from natural sciences:1.The detection of Firtrees (Abies Sibirica) damaged by the bark beetle in UAV images using DL. 2.Theestimation of olive tree biovolume from UAV multi-resolution image segmentationusing MaskR-CNN. 3.The detection of Spruce trees (Picea Abies) infected by bark beetle in UAV images using YOLOs architectures. The main objectiveof this thesis is to develop robust and accurate DL models for the monitoringof different plant species using UAV images. The particular objectives toachieve the main objective are:●To build three high-quality datasets for eachone of the three considered problems.●To design the appropriate pre-processing methods that reduce noise and uncertainty in the features and annotations.●To developrobust and accurate CNN-based models for each case study.The results of thefirst two chapters of this thesis have been published in two journals ranked asQ1 and Q2 in JCR. The results of the third chapter have been submitted to “IEEETransactions on Geoscience and Remote Sensing”.

This thesis is organized into five chapters. The first chapter introducesthe considered problems, background, and objectives of the thesis. Chapter twopresents the built dataset, detection model and pre-processing technique to address the detection of Fir trees (Abies Sibirica) damaged by the bark beetle in UAV images. Chapter three presents the built dataset,segmentation model and pre-processing techniques to estimate olive treebiovolume from UAV multi-resolution image segmentation. Chapter four gives thebuilt dataset, models, and pre-processing technique to address the detection ofSpruce trees (Picea Abies) infected by bark beetle in UAV images using YOLOs architectures. Finally, Chapter five provides conclusions and futurework.

下载地址：

https://url92.ctfile.com/f/1850492-517632653-a7ceb2

（访问密码：3660）

更多精彩文章请关注公众号：