Noise interference critically impairs the stability and data accuracy of sensing systems.However,current suppression strategies fail to concurrently mitigate intrinsic system noise and extrinsic environmental noise.Th...Noise interference critically impairs the stability and data accuracy of sensing systems.However,current suppression strategies fail to concurrently mitigate intrinsic system noise and extrinsic environmental noise.This study introduces a composite denoising approach to address this challenge.This method is based on the ameliorated ellipse fitting algorithm(AEFA)and adaptive successive variational mode decomposition(ASVMD).This algorithm employs AEFA to eliminate system noise tightly coupled with direct-current and alternating-current components in the interference signal,thereby obtaining a phase signal containing only environmental noise.The ASVMD technique adaptively extracts environmental noise components predominantly present in the phase signal.To achieve optimal decomposition results automatically,the permutation entropy criterion is employed to refine decomposition parameters.The correlation coefficient is utilized to differentiate effective components from noise components in the decomposition results.Experimental results indicate that the combined AEFA and ASVMD algorithm effectively suppresses both system and environmental noises.When applied to 50 Hz vibration signal processing,the proposed approach achieves a noise reduction of 17.81 dB and a phase resolution of 35.14μrad/√Hz.Given the excellent performance of the noise suppression,the proposed approach holds great application potential in high-performance interferometric sensing systems.展开更多
针对暗环境动态特征轮廓模糊、盲区遮挡情况,高效准确地检测跟踪动态目标特征,对灾害救援、搜寻跟踪具有实际意义。为实现暗环境下模糊轮廓特征的有效检测跟踪,提出一种时空关联机制的红外目标实时检测深度学习网络(Spatial Local Dynam...针对暗环境动态特征轮廓模糊、盲区遮挡情况,高效准确地检测跟踪动态目标特征,对灾害救援、搜寻跟踪具有实际意义。为实现暗环境下模糊轮廓特征的有效检测跟踪,提出一种时空关联机制的红外目标实时检测深度学习网络(Spatial Local Dynamic You Only Look Once Version 8,SLD-YOLOv8),设计非局部自适应Non-local模块和空间通道卷积关联模块,对原YOLOv8网络的瓶颈层Bottleneck CSP进行优化。为有效提取深层空间多尺度表征信息,增加用于小目标检测的160×160检测层和动态检测头,较好地提升暗环境中目标跟踪的边界回归性能,并实时有效地推理出目标特征的相对深度位置信息。实验结果表明,改进后的红外目标检测算法对暗环境下的动态特征检测具有较好的鲁棒性和准确性,其平均精度评估指标mAP_0.5和mAP_0.5:0.95比原模型提高了5.6%和4.5%,证明了新算法对暗环境目标跟踪的有效性。展开更多
文摘Noise interference critically impairs the stability and data accuracy of sensing systems.However,current suppression strategies fail to concurrently mitigate intrinsic system noise and extrinsic environmental noise.This study introduces a composite denoising approach to address this challenge.This method is based on the ameliorated ellipse fitting algorithm(AEFA)and adaptive successive variational mode decomposition(ASVMD).This algorithm employs AEFA to eliminate system noise tightly coupled with direct-current and alternating-current components in the interference signal,thereby obtaining a phase signal containing only environmental noise.The ASVMD technique adaptively extracts environmental noise components predominantly present in the phase signal.To achieve optimal decomposition results automatically,the permutation entropy criterion is employed to refine decomposition parameters.The correlation coefficient is utilized to differentiate effective components from noise components in the decomposition results.Experimental results indicate that the combined AEFA and ASVMD algorithm effectively suppresses both system and environmental noises.When applied to 50 Hz vibration signal processing,the proposed approach achieves a noise reduction of 17.81 dB and a phase resolution of 35.14μrad/√Hz.Given the excellent performance of the noise suppression,the proposed approach holds great application potential in high-performance interferometric sensing systems.
文摘针对暗环境动态特征轮廓模糊、盲区遮挡情况,高效准确地检测跟踪动态目标特征,对灾害救援、搜寻跟踪具有实际意义。为实现暗环境下模糊轮廓特征的有效检测跟踪,提出一种时空关联机制的红外目标实时检测深度学习网络(Spatial Local Dynamic You Only Look Once Version 8,SLD-YOLOv8),设计非局部自适应Non-local模块和空间通道卷积关联模块,对原YOLOv8网络的瓶颈层Bottleneck CSP进行优化。为有效提取深层空间多尺度表征信息,增加用于小目标检测的160×160检测层和动态检测头,较好地提升暗环境中目标跟踪的边界回归性能,并实时有效地推理出目标特征的相对深度位置信息。实验结果表明,改进后的红外目标检测算法对暗环境下的动态特征检测具有较好的鲁棒性和准确性,其平均精度评估指标mAP_0.5和mAP_0.5:0.95比原模型提高了5.6%和4.5%,证明了新算法对暗环境目标跟踪的有效性。
