摘要
Abstract
Compared with single-band object detection technology,multispectral object detection technology greatly improves the accuracy of object detection and the robustness in dealing with complex environments by capturing the reflection or radiation information of objects in multiple spectral bands of different wavelengths.Therefore,it has extensive applications in fields such as remote sensing,agricultural detection,environmental protection,industrial production,and national defense security.However,the field of multispectral object detection still faces severe challenges at present:the lack of diverse high-quality datasets and efficient object detection algorithms seriously restricts further development and application of this technology.In view of this,this paper comprehensively explains the production method of multispectral object detection datasets and the important progress of multispectral object detection algorithms.First,the article systematically analyzes the construction process of multispectral datasets,including data acquisition,preprocessing,and data annotation,aiming to provide technical support for the subsequent construction of high-quality multispectral object detection datasets.Second,this paper comprehensively analyzes the historical context of the development of object detection algorithms.These algorithms cover object detection algorithms based on traditional feature extraction technologies,deep learning methods,and their improved versions.In addition,this paper summarizes the key improvements made by algorithm developers in terms of feature fusion,model architecture,and sub-networks to improve the performance of multispectral object detection based on deep learning-based object detection algorithms.Finally,this paper discusses future development direction of multispectral object detection technology,hoping to indicate potential research hotspots and application fields for researchers,and promote the wider application of multispectral object detection technology in actual scenarios and enhance its social value.关键词
多光谱成像技术/成像光谱仪/目标检测算法/多光谱目标检测算法/数据集/特征融合Key words
multispectral imaging technology/imaging spectrometer/object detection algorithm/multispectral object detection algorithm/dataset/feature fusion分类
信息技术与安全科学
