This paper mainly focuses on the issues about generic multi-scale object perception for detection or recognition. A novel computational model in visually-feature space is presented for scene & object representatio...This paper mainly focuses on the issues about generic multi-scale object perception for detection or recognition. A novel computational model in visually-feature space is presented for scene & object representation to purse the underlying textural manifold statistically in nonparametric manner. The associative method approximately makes perceptual hierarchy in human-vision biologically coherency in specific quad-tree-pyramid structure, and the appropriate scale-value of different objects can automatically be selected by evaluating from well-defined scale function without any priori knowledge. The sufficient experiments truly demonstrate the effectiveness of scale determination in textural manifold with object localization rapidly.展开更多
In this paper, we presented a method of using the l as er scanning triangulation for the non-contact 3D surface profile measurement of large-scale object. The characteristic of large-scale object non-contact mea surem...In this paper, we presented a method of using the l as er scanning triangulation for the non-contact 3D surface profile measurement of large-scale object. The characteristic of large-scale object non-contact mea surement is analyzed and the measuring method is proposed. Main factors influenc ing measurement precision such as image distortion and accurate designation of s peckle center are analyzed and methods of solving these problems are proposed. W e designed a combined filter by which the pulse noise and the Gaussian noise of speckle image can be eliminated efficiently. Using the characteristic of intensi ty distribution of laser speckle image we proposed a new approximating method th at could locate the center of laser speckle image at sub-pixel. The auxiliary v ariables are set to linearize the relationship between the image displacement an d the distance, the accurate values of laser triangulation system parameters cou ld be calibrated accurately and the measuring precision is increased remarkabl y. Using the above techniques we designed a measuring system based on laser sc anning triangulation. The results of the experiment show that these methods can raise the measuring precision of large-scale 3D surface profile effectively.展开更多
A novel multi level image segmentation methodology is been proposed with the aim of extracting the salient object,keeping in view,only a small part of the visual scene undergoes attention and reaches the level of awar...A novel multi level image segmentation methodology is been proposed with the aim of extracting the salient object,keeping in view,only a small part of the visual scene undergoes attention and reaches the level of awareness while rest of details are futile.Taking advantage of multilevel gray scale quantization,image prominent object is separated from background,keeping in view the fact;salient object is having high contrast as compared to the background.The inutile fragments were removed using morphological operations of opening and closing and making the image smoothened with Gaussian filter.The optimum threshold is selected for the binary conversion and final extrication of the salient object from the image.The experimental data indicates that hybrid approach leads to improved segmentation with the apparent assertion of prime object extraction.展开更多
【目的】矿产资源是人类生存和经济发展的重要物质基础,开展矿山监测、建立矿山监测模型对矿产资源的高效开发和矿区环境保护具有重要意义。针对露天矿区背景复杂、目标尺度多样且小目标聚集的特点,本研究旨在构建兼顾监测精度与效率的...【目的】矿产资源是人类生存和经济发展的重要物质基础,开展矿山监测、建立矿山监测模型对矿产资源的高效开发和矿区环境保护具有重要意义。针对露天矿区背景复杂、目标尺度多样且小目标聚集的特点,本研究旨在构建兼顾监测精度与效率的轻量化模型,以提升矿区目标地物监测的准确性和效率。【方法】现有遥感数据集存在的样本单一、地域局限等问题,因此本文基于0.9 m天地图与1.8 m谷歌影像构建了不同气候背景、大范围和多种地物的六大露天煤矿基地OMTSFD(Open-pit Mine Typical Surface Features Dataset)数据集,提出改进的YOLO11-DAE算法进行模型训练与验证。首先,在骨干网络和特征金字塔中引入C3K2-DBB模块以增强多尺度特征捕获能力;其次,采用ADown模块替换网络下采样卷积,增强了模块对不同特征的表征能力,减少了低对比度场景的细节丢失;最后,采用E_Detect高效检测头降低模型复杂度和参数量,实现模型轻量化。【结果】实验表明,YOLO11-DAE的每秒帧数(Frames Per Second,FPS)为528.100,模型推理速度较快,精确率(Precision,P)、召回率(Recall,R)、综合评价指标(F1-Score,F1)、平均精度均值(Mean Average Precision,mAP)分别达到0.932、0.894、0.913和0.950,显著优于YOLOv5n、YOLOv8n和YOLOv10n算法,相较于YOLOv11n各项指标分别提高7.600%、10.000%、8.800%、8.000%。【结论】YOLO11-DAE算法能够满足矿区实时监测,并适用于多尺度、多背景等复杂场景的目标识别,实现了高精度、低漏检率的监测目标,达到了模型可应用性与实时性的平衡。展开更多
为了解决施工场景下安全帽佩戴检测时,由于人员密集、遮挡和复杂背景等原因造成的小目标漏检和错检的问题,提出一种基于YOLOv8n的双重注意力机制的跨层多尺度安全帽佩戴检测算法。首先,设计微小目标检测头,以提高模型对小目标的检测能力...为了解决施工场景下安全帽佩戴检测时,由于人员密集、遮挡和复杂背景等原因造成的小目标漏检和错检的问题,提出一种基于YOLOv8n的双重注意力机制的跨层多尺度安全帽佩戴检测算法。首先,设计微小目标检测头,以提高模型对小目标的检测能力;其次,在特征提取网络中嵌入双重注意力机制,从而更加关注复杂场景下目标信息的特征捕获;然后,将特征融合网络替换成重参数化泛化特征金字塔网络(RepGFPN)改进后的跨层多尺度特征融合结构S-GFPN(Selective layer Generalized Feature Pyramid Network),以实现小目标特征层信息和其他特征层的多尺度融合,并建立长期的依赖关系,从而抑制背景信息的干扰;最后,采用MPDIOU(Intersection Over Union with Minimum Point Distance)损失函数来解决尺度变化不敏感的问题。在公开数据集GDUT-HWD上的实验结果表明,改进后的模型比YOLOv8n的mAP@0.5提升了3.4个百分点,对蓝色、黄色、白色和红色安全帽的检测精度分别提升了2.0、1.1、4.6和9.1个百分点,在密集、遮挡、小目标、反光和黑暗这5类复杂场景下的可视化检测效果也优于YOLOv8n,为实际施工场景中安全帽佩戴检测提供了一种有效方法。展开更多
文摘This paper mainly focuses on the issues about generic multi-scale object perception for detection or recognition. A novel computational model in visually-feature space is presented for scene & object representation to purse the underlying textural manifold statistically in nonparametric manner. The associative method approximately makes perceptual hierarchy in human-vision biologically coherency in specific quad-tree-pyramid structure, and the appropriate scale-value of different objects can automatically be selected by evaluating from well-defined scale function without any priori knowledge. The sufficient experiments truly demonstrate the effectiveness of scale determination in textural manifold with object localization rapidly.
文摘In this paper, we presented a method of using the l as er scanning triangulation for the non-contact 3D surface profile measurement of large-scale object. The characteristic of large-scale object non-contact mea surement is analyzed and the measuring method is proposed. Main factors influenc ing measurement precision such as image distortion and accurate designation of s peckle center are analyzed and methods of solving these problems are proposed. W e designed a combined filter by which the pulse noise and the Gaussian noise of speckle image can be eliminated efficiently. Using the characteristic of intensi ty distribution of laser speckle image we proposed a new approximating method th at could locate the center of laser speckle image at sub-pixel. The auxiliary v ariables are set to linearize the relationship between the image displacement an d the distance, the accurate values of laser triangulation system parameters cou ld be calibrated accurately and the measuring precision is increased remarkabl y. Using the above techniques we designed a measuring system based on laser sc anning triangulation. The results of the experiment show that these methods can raise the measuring precision of large-scale 3D surface profile effectively.
文摘A novel multi level image segmentation methodology is been proposed with the aim of extracting the salient object,keeping in view,only a small part of the visual scene undergoes attention and reaches the level of awareness while rest of details are futile.Taking advantage of multilevel gray scale quantization,image prominent object is separated from background,keeping in view the fact;salient object is having high contrast as compared to the background.The inutile fragments were removed using morphological operations of opening and closing and making the image smoothened with Gaussian filter.The optimum threshold is selected for the binary conversion and final extrication of the salient object from the image.The experimental data indicates that hybrid approach leads to improved segmentation with the apparent assertion of prime object extraction.
文摘【目的】矿产资源是人类生存和经济发展的重要物质基础,开展矿山监测、建立矿山监测模型对矿产资源的高效开发和矿区环境保护具有重要意义。针对露天矿区背景复杂、目标尺度多样且小目标聚集的特点,本研究旨在构建兼顾监测精度与效率的轻量化模型,以提升矿区目标地物监测的准确性和效率。【方法】现有遥感数据集存在的样本单一、地域局限等问题,因此本文基于0.9 m天地图与1.8 m谷歌影像构建了不同气候背景、大范围和多种地物的六大露天煤矿基地OMTSFD(Open-pit Mine Typical Surface Features Dataset)数据集,提出改进的YOLO11-DAE算法进行模型训练与验证。首先,在骨干网络和特征金字塔中引入C3K2-DBB模块以增强多尺度特征捕获能力;其次,采用ADown模块替换网络下采样卷积,增强了模块对不同特征的表征能力,减少了低对比度场景的细节丢失;最后,采用E_Detect高效检测头降低模型复杂度和参数量,实现模型轻量化。【结果】实验表明,YOLO11-DAE的每秒帧数(Frames Per Second,FPS)为528.100,模型推理速度较快,精确率(Precision,P)、召回率(Recall,R)、综合评价指标(F1-Score,F1)、平均精度均值(Mean Average Precision,mAP)分别达到0.932、0.894、0.913和0.950,显著优于YOLOv5n、YOLOv8n和YOLOv10n算法,相较于YOLOv11n各项指标分别提高7.600%、10.000%、8.800%、8.000%。【结论】YOLO11-DAE算法能够满足矿区实时监测,并适用于多尺度、多背景等复杂场景的目标识别,实现了高精度、低漏检率的监测目标,达到了模型可应用性与实时性的平衡。
文摘为了解决施工场景下安全帽佩戴检测时,由于人员密集、遮挡和复杂背景等原因造成的小目标漏检和错检的问题,提出一种基于YOLOv8n的双重注意力机制的跨层多尺度安全帽佩戴检测算法。首先,设计微小目标检测头,以提高模型对小目标的检测能力;其次,在特征提取网络中嵌入双重注意力机制,从而更加关注复杂场景下目标信息的特征捕获;然后,将特征融合网络替换成重参数化泛化特征金字塔网络(RepGFPN)改进后的跨层多尺度特征融合结构S-GFPN(Selective layer Generalized Feature Pyramid Network),以实现小目标特征层信息和其他特征层的多尺度融合,并建立长期的依赖关系,从而抑制背景信息的干扰;最后,采用MPDIOU(Intersection Over Union with Minimum Point Distance)损失函数来解决尺度变化不敏感的问题。在公开数据集GDUT-HWD上的实验结果表明,改进后的模型比YOLOv8n的mAP@0.5提升了3.4个百分点,对蓝色、黄色、白色和红色安全帽的检测精度分别提升了2.0、1.1、4.6和9.1个百分点,在密集、遮挡、小目标、反光和黑暗这5类复杂场景下的可视化检测效果也优于YOLOv8n,为实际施工场景中安全帽佩戴检测提供了一种有效方法。
