低秩稀疏分解方法因其好的检测性能在红外小目标检测领域受到广泛关注。然而,现有低秩稀疏分解方法在复杂场景中仍然面临检测性能不高、检测速度较慢等问题。虽然现有的低秩塔克分解方法在复杂场景下取得了令人满意的检测性能,但其需依...低秩稀疏分解方法因其好的检测性能在红外小目标检测领域受到广泛关注。然而,现有低秩稀疏分解方法在复杂场景中仍然面临检测性能不高、检测速度较慢等问题。虽然现有的低秩塔克分解方法在复杂场景下取得了令人满意的检测性能,但其需依赖经验预先定义秩:若秩估计过大或过小,会导致漏检或虚警。而且,不同场景中秩的大小不一样,限制了实际应用。为了解决这一问题,本文采用非凸秩接近范数约束低秩塔克分解的潜在因子,无需手动设置秩,从而显著提升了算法在不同场景中的鲁棒性。进一步地,设计了基于对称高斯-赛德尔的交替方向乘子法(symmetric GaussSeidel based alternating direction method of multipliers algorithm,sGSADMM)来求解所提模型。与现有基于交替方向乘子法相比,sGSADMM算法通过利用更多结构信息,实现了更高的求解精度。大量实验表明,所提方法在检测性能和背景抑制等方面均优于现有的先进算法。展开更多
特殊环境下道路目标的三维感知对汽车的全天时、全气候自动驾驶具有重要意义,红外双目视觉模仿人眼实现微光/无光等特殊环境下目标的立体感知,目标检测与匹配是双目视觉立体感知的关键技术。针对当前分步实现目标检测与目标匹配的过程...特殊环境下道路目标的三维感知对汽车的全天时、全气候自动驾驶具有重要意义,红外双目视觉模仿人眼实现微光/无光等特殊环境下目标的立体感知,目标检测与匹配是双目视觉立体感知的关键技术。针对当前分步实现目标检测与目标匹配的过程冗杂问题,提出了一个可以同步检测与匹配红外目标的深度学习网络——SODMNet(Synchronous Object Detection and Matching Network)。SODMNet融合了目标检测网络和目标匹配模块,以目标检测网络为主要架构,取其分类与回归分支深层特征为目标匹配模块的输入,与特征图相对位置编码拼接后通过卷积网络输出左右图像特征描述子,根据特征描述子之间的欧氏距离得到目标匹配结果,实现双目视觉目标检测与匹配。与此同时,采集并制作了一个包含人、车辆等标注目标的夜间红外双目数据集。实验结果表明,SODMNet在该红外双目数据集上的目标检测精度mAP(Mean Average Precision)提升84.9%以上,同时目标匹配精度AP(Average Precision)达到0.5777。结果证明,SODMNet能够高精度地同步实现红外双目目标检测与匹配。展开更多
The concept of emissivity has been with the scientific and engineering world since Planck formulated his blackbody radiation law more than a century ago.Nevertheless,emissivity is an elusive concept even for ex⁃perts....The concept of emissivity has been with the scientific and engineering world since Planck formulated his blackbody radiation law more than a century ago.Nevertheless,emissivity is an elusive concept even for ex⁃perts.It is a vague and fuzzy concept for the wider community of engineers.The importance of remote sensing of temperature by measuring IR radiation has been recognized in a wide range of industrial,medical,and environ⁃mental uses.One of the major sources of errors in IR radiometry is the emissivity of the surface being measured.In real experiments,emissivity may be influenced by many factors:surface texture,spectral properties,oxida⁃tion,and aging of surfaces.While commercial blackbodies are prevalent,the much-needed grey bodies with a known emissivity,are unavailable.This study describes how to achieve a calibrated and stable emissivity with a blackbody,a perforated screen,and a reliable and linear novel IR thermal sensor,18 dubbed TMOS.The Digital TMOS is now a low-cost commercial product,it requires low power,and it has a small form factor.The method⁃ology is based on two-color measurements,with two different optical filters,with selected wavelengths conform⁃ing to the grey body definition of the use case under study.With a photochemically etched perforated screen,the effective emissivity of the screen is simply the hole density area of the surface area that emits according to the blackbody temperature radiation.The concept is illustrated with ray tracing simulations,which demonstrate the approach.Measured results are reported.展开更多
文摘低秩稀疏分解方法因其好的检测性能在红外小目标检测领域受到广泛关注。然而,现有低秩稀疏分解方法在复杂场景中仍然面临检测性能不高、检测速度较慢等问题。虽然现有的低秩塔克分解方法在复杂场景下取得了令人满意的检测性能,但其需依赖经验预先定义秩:若秩估计过大或过小,会导致漏检或虚警。而且,不同场景中秩的大小不一样,限制了实际应用。为了解决这一问题,本文采用非凸秩接近范数约束低秩塔克分解的潜在因子,无需手动设置秩,从而显著提升了算法在不同场景中的鲁棒性。进一步地,设计了基于对称高斯-赛德尔的交替方向乘子法(symmetric GaussSeidel based alternating direction method of multipliers algorithm,sGSADMM)来求解所提模型。与现有基于交替方向乘子法相比,sGSADMM算法通过利用更多结构信息,实现了更高的求解精度。大量实验表明,所提方法在检测性能和背景抑制等方面均优于现有的先进算法。
文摘特殊环境下道路目标的三维感知对汽车的全天时、全气候自动驾驶具有重要意义,红外双目视觉模仿人眼实现微光/无光等特殊环境下目标的立体感知,目标检测与匹配是双目视觉立体感知的关键技术。针对当前分步实现目标检测与目标匹配的过程冗杂问题,提出了一个可以同步检测与匹配红外目标的深度学习网络——SODMNet(Synchronous Object Detection and Matching Network)。SODMNet融合了目标检测网络和目标匹配模块,以目标检测网络为主要架构,取其分类与回归分支深层特征为目标匹配模块的输入,与特征图相对位置编码拼接后通过卷积网络输出左右图像特征描述子,根据特征描述子之间的欧氏距离得到目标匹配结果,实现双目视觉目标检测与匹配。与此同时,采集并制作了一个包含人、车辆等标注目标的夜间红外双目数据集。实验结果表明,SODMNet在该红外双目数据集上的目标检测精度mAP(Mean Average Precision)提升84.9%以上,同时目标匹配精度AP(Average Precision)达到0.5777。结果证明,SODMNet能够高精度地同步实现红外双目目标检测与匹配。
文摘The concept of emissivity has been with the scientific and engineering world since Planck formulated his blackbody radiation law more than a century ago.Nevertheless,emissivity is an elusive concept even for ex⁃perts.It is a vague and fuzzy concept for the wider community of engineers.The importance of remote sensing of temperature by measuring IR radiation has been recognized in a wide range of industrial,medical,and environ⁃mental uses.One of the major sources of errors in IR radiometry is the emissivity of the surface being measured.In real experiments,emissivity may be influenced by many factors:surface texture,spectral properties,oxida⁃tion,and aging of surfaces.While commercial blackbodies are prevalent,the much-needed grey bodies with a known emissivity,are unavailable.This study describes how to achieve a calibrated and stable emissivity with a blackbody,a perforated screen,and a reliable and linear novel IR thermal sensor,18 dubbed TMOS.The Digital TMOS is now a low-cost commercial product,it requires low power,and it has a small form factor.The method⁃ology is based on two-color measurements,with two different optical filters,with selected wavelengths conform⁃ing to the grey body definition of the use case under study.With a photochemically etched perforated screen,the effective emissivity of the screen is simply the hole density area of the surface area that emits according to the blackbody temperature radiation.The concept is illustrated with ray tracing simulations,which demonstrate the approach.Measured results are reported.
