针对废墟环境下红外图像人体检测任务中存在的图像分辨率低且人体特征不明显的问题,基于YOLO框架设计了一种包含重参数化(re-parameterization)和多尺度大核卷积(multi-scale large kernel convolution)的红外图像人体检测网络RML-YOLO(...针对废墟环境下红外图像人体检测任务中存在的图像分辨率低且人体特征不明显的问题,基于YOLO框架设计了一种包含重参数化(re-parameterization)和多尺度大核卷积(multi-scale large kernel convolution)的红外图像人体检测网络RML-YOLO(re-parameterization multi-scale large kernel convolution)。该网络通过空间和通道重构注意力模块,将注意值集中到对检测任务更重要的区域。通过Sobel算子强化边缘特征,提高对不同姿态人体的检测能力。RML-YOLO的有效性在自制数据集上得到验证。在只有1.8×10~6可学习参数的情况下,模型的AP50和AP50-75分别达到了91.2%和87.3%,与参数量相近的YOLOv8-n相比分别提高了4.4%和5.3%。结果表明,RML-YOLO显著提高了利用红外图像进行废墟环境下人体检测的精度。展开更多
Let α≥ 0 and 0 〈 ρ ≤ n/2, the boundedness of hypersingular parameterized Marcinkiewicz integrals μΩ,α^ρ with variable kernels on Sobolev spaces Lα^ρ and HardySobolev spaces Hα^ρ is established.
In this paper,the parameterized Marcinkiewicz integrals with variable kernels defined by μΩ^ρ（f）（x）=（∫0^∞│∫│1-y│≤t Ω（x,x-y）/│x-y│^n-p f（y）dy│^2dt/t1＋2p）^1/2 are investigated.It is proved that ...In this paper,the parameterized Marcinkiewicz integrals with variable kernels defined by μΩ^ρ（f）（x）=（∫0^∞│∫│1-y│≤t Ω（x,x-y）/│x-y│^n-p f（y）dy│^2dt/t1＋2p）^1/2 are investigated.It is proved that if Ω∈ L∞（R^n） × L^r（S^n-1）（r〉（n-n1p＇/n） is an odd function in the second variable y,then the operator μΩ^ρ is bounded from L^p（R^n） to L^p（R^n） for 1 〈 p ≤ max{（n＋1）/2,2}.It is also proved that,if Ω satisfies the L^1-Dini condition,then μΩ^ρ is of type（p,p） for 1 〈 p ≤ 2,of the weak type（1,1） and bounded from H1 to L1.展开更多
文摘针对废墟环境下红外图像人体检测任务中存在的图像分辨率低且人体特征不明显的问题,基于YOLO框架设计了一种包含重参数化(re-parameterization)和多尺度大核卷积(multi-scale large kernel convolution)的红外图像人体检测网络RML-YOLO(re-parameterization multi-scale large kernel convolution)。该网络通过空间和通道重构注意力模块,将注意值集中到对检测任务更重要的区域。通过Sobel算子强化边缘特征,提高对不同姿态人体的检测能力。RML-YOLO的有效性在自制数据集上得到验证。在只有1.8×10~6可学习参数的情况下,模型的AP50和AP50-75分别达到了91.2%和87.3%,与参数量相近的YOLOv8-n相比分别提高了4.4%和5.3%。结果表明,RML-YOLO显著提高了利用红外图像进行废墟环境下人体检测的精度。
基金Supported by the National Natural Science Foundation of China under Grant Nos.6043302060773111(国家自然科学基金)+1 种基金the Program for New Century Excellent Talents in University of China under Grant No.NCET-05-0683(新世纪优秀人才支持计划)the Program for Changjiang Scholars and Innovative Research Team in University of China under Grant No.IRT0661(长江学者和创新团队发展计划)
基金Supported by the National Natural Science Foundation of China(1057115610871173)
文摘Let α≥ 0 and 0 〈 ρ ≤ n/2, the boundedness of hypersingular parameterized Marcinkiewicz integrals μΩ,α^ρ with variable kernels on Sobolev spaces Lα^ρ and HardySobolev spaces Hα^ρ is established.
基金Supported by the National Natural Science Foundation of China (1057115610871173)
文摘In this paper,the parameterized Marcinkiewicz integrals with variable kernels defined by μΩ^ρ（f）（x）=（∫0^∞│∫│1-y│≤t Ω（x,x-y）/│x-y│^n-p f（y）dy│^2dt/t1＋2p）^1/2 are investigated.It is proved that if Ω∈ L∞（R^n） × L^r（S^n-1）（r〉（n-n1p＇/n） is an odd function in the second variable y,then the operator μΩ^ρ is bounded from L^p（R^n） to L^p（R^n） for 1 〈 p ≤ max{（n＋1）/2,2}.It is also proved that,if Ω satisfies the L^1-Dini condition,then μΩ^ρ is of type（p,p） for 1 〈 p ≤ 2,of the weak type（1,1） and bounded from H1 to L1.
