摘要
Abstract
To address the challenge of identifying target areas in unmanned aerial vehicle(UAV)remote sensing images with geographical information,where the targets are large,densely distributed,and set against a complex background,this study proposes a method for accurate multi-target recognition.The image is divided into multiple sub-regions using spectral clustering technology.Features such as brightness,contrast,and information entropy are extracted,and the optimal target region is determined using the entropy weighting method.Similarity is then assessed,and an adaptive threshold iterative method is applied to accurately recognize the target regions.After determining the target regions,layered bag-of-features(BoF)and scale-invariant feature transform(SIFT)features,spectral clustering(SC)features,and Hu Moments invariants are captured within the regions.A radial basis function(RBF)support vector machine(SVM)model is trained using cross-validation,and the recognition probability of each feature is computed.By combining a multi-feature decision-level weighted fusion strategy,this method enables the recognition of geographical targets,such as houses and land,in UAV remote sensing images.The case study results show that the proposed method can accurately recognize multiple targets in UAV remote sensing images,with an accuracy rate exceeding 95%,demonstrating its practical application value.关键词
地理信息/无人机遥感图像/多目标/精准识别Key words
geographical information/unmanned aerial vehicle(UAV)remote sensing image/multi-target/accurate recognition分类
信息技术与安全科学
