红外热成像测温是及时发现电缆附件异常发热缺陷的重要方法,但面对海量巡检图像,传统的人工诊断方式费时费力,且过分依赖人工经验。已有研究中为了实现电气设备红外图像的智能诊断,大多提取特定特征量作为所搭建神经网络模型的输入,而...红外热成像测温是及时发现电缆附件异常发热缺陷的重要方法,但面对海量巡检图像,传统的人工诊断方式费时费力,且过分依赖人工经验。已有研究中为了实现电气设备红外图像的智能诊断,大多提取特定特征量作为所搭建神经网络模型的输入,而所提取的特征量也依赖于人工选择。为此,提出了一种基于Faster RCNN(faster regions with convolutional neural networks features)与Mean-Shift的电缆附件缺陷红外图像自动诊断方法。该方法首先基于Faster RCNN网络实现诊断对象的识别定位;之后利用Mean-Shift聚类算法提取过热区域;最后计算温度参数,并根据相应诊断标准得到诊断结果。利用实际巡检图像进行测试,结果表明:所提方法能够在不同拍摄角度、不同背景下准确定位诊断目标与过热区域,继而实现过热缺陷的自动诊断。研究对于实际工程中电缆附件的缺陷诊断具有一定的参考价值。展开更多
针对田间密植环境棉花精准打顶时,棉花顶芽因其小体积特性所带来识别困难问题,该研究提出一种改进型快速区域卷积神经网络(Faster Region Convolutional Neural Networks,Faster R-CNN)目标检测算法实现大田环境棉花顶芽识别。以Faster ...针对田间密植环境棉花精准打顶时,棉花顶芽因其小体积特性所带来识别困难问题,该研究提出一种改进型快速区域卷积神经网络(Faster Region Convolutional Neural Networks,Faster R-CNN)目标检测算法实现大田环境棉花顶芽识别。以Faster R-CNN为基础框架,使用RegNetX-6.4GF作为主干网络,以提高图像特征获取性能。将特征金字塔网络(Feature Pyramid Network,FPN)和导向锚框定位(Guided Anchoring,GA)机制相融合,实现锚框(Anchor)动态自适应生成。通过融合动态区域卷积神经网络(Dynamic Region Convolutional Neural Networks,Dynamic R-CNN),实现训练阶段检测模型自适应候选区域(Proposal)分布的动态变化。最后在目标候选区域(Region of Interest,ROI)中引入目标候选区域提取器(Generic ROI Extractor,GROIE)提高图像特征融合能力。采集自然环境下7种不同棉花总计4819张图片,建立微软常见物体图像识别库2017(Microsoft Common Objects in Context 2017,MS COCO 2017)格式的棉花顶芽图片数据集进行试验。结果表明,该研究提出方法的平均准确率均值(Mean Average Precision,MAP)为98.1%,模型的处理帧速(Frames Per Second,FPS)为10.3帧/s。其MAP在交并比(Intersection Over Union,IOU)为0.5时较Faster R-CNN、RetinaNet、Cascade R-CNN和RepPoints网络分别提高7.3%、78.9%、10.1%和8.3%。该研究算法在田间对于棉花顶芽识别具有较高的鲁棒性和精确度,为棉花精准打顶作业奠定基础。展开更多
小目标是指图像中覆盖区域较小的一类目标.与常规目标相比,小目标信息量少,训练数据难以标记,这导致通用的目标检测方法对小目标的检测效果不好,而专门为小目标设计的检测方法往往复杂度过高或不具有通用性.在分析现有目标检测方法的基...小目标是指图像中覆盖区域较小的一类目标.与常规目标相比,小目标信息量少,训练数据难以标记,这导致通用的目标检测方法对小目标的检测效果不好,而专门为小目标设计的检测方法往往复杂度过高或不具有通用性.在分析现有目标检测方法的基础上,提出了一种面向小目标的多尺度快速区域卷积神经网络(faster-regions with convolutional neural network, Faster-RCNN)检测算法.根据卷积神经网络的特性,修改了Faster-RCNN的网络结构,使网络可以同时使用低层和高层的特征进行多尺度目标检测,提升了以低层特征为主要检测依据的小目标检测任务的精度.同时,针对训练数据难以标记的问题,使用从搜索引擎上获取的数据来训练模型.因为这些训练数据与任务测试数据分布不同,又利用下采样和上采样的方法对目标高分辨率的训练图像进行转化,使训练图像和测试图像的特征分布更类似.实验结果表明:所提出的方法在小目标检测任务上的平均精度均值(mean average precision, mAP)可以比原始的Faster-RCNN提高约5%.展开更多
随着交通愈加发达,道路愈加拥堵,如何实时准确地获取车辆基本信息以便交通部门及时管理特定路段和路口的车辆显得日益重要.对交通视频中车辆的检测和识别,不仅需要实时检测,还要保证其准确性.针对实际情况中车辆之间的遮挡、光照的变化...随着交通愈加发达,道路愈加拥堵,如何实时准确地获取车辆基本信息以便交通部门及时管理特定路段和路口的车辆显得日益重要.对交通视频中车辆的检测和识别,不仅需要实时检测,还要保证其准确性.针对实际情况中车辆之间的遮挡、光照的变化、阴影、道路旁树枝的晃动、背景中固定对象的移动等因素严重影响检测与识别的精度的问题,提出基于Faster-RCNN(Faster-Regions with CNN features)的车辆实时检测改进算法.首先采用k-means算法对KITTI数据集的目标框进行聚类,得到合适的长宽比,并增加一组尺度(64~2)以适应差异较大的车辆尺寸;然后改进区域提案网络,降低计算量,优化网络结构;最后在训练阶段采用多尺度策略,降低漏检率,提高精确率.实验结果表明:改进后的车辆检测算法的mAP(mean Average Precision)达到了82.20%,检测速率为每张照片耗时0.03875 s,基本能够满足车辆实时检测的需求.展开更多
It well known that vehicle detection is an important component of the field of object detection.However,the environment of vehicle detection is particularly sophisticated in practical processes.It is comparatively dif...It well known that vehicle detection is an important component of the field of object detection.However,the environment of vehicle detection is particularly sophisticated in practical processes.It is comparatively difficult to detect vehicles of various scales in traffic scene images,because the vehicles partially obscured by green belts,roadblocks or other vehicles,as well as influence of some low illumination weather.In this paper,we present a model based on Faster ReCNN with NAS optimization and feature enrichment to realize the effective detection of multi-scale vehicle targets in traffic scenes.First,we proposed a Retinex-based image adaptive correction algorithm(RIAC)to enhance the traffic images in the dataset to reduce the influence of shadow and illumination,and improve the image quality.Second,in order to improve the feature expression of the backbone network,we conducted Neural Architecture Search(NAS)on the backbone network used for feature extraction of Faster ReCNN to generate the optimal cross-layer connection to extract multi-layer features more effectively.Third,we used the object Feature Enrichment that combines the multi-layer feature information and the context information of the last layer after cross-layer connection to enrich the information of vehicle targets,and improve the robustness of the model for challenging targets such as small scale and severe occlusion.In the implementation of the model,K-means clustering algorithm was used to select the suitable anchor size for our dataset to improve the convergence speed of the model.Our model has been trained and tested on the UN-DETRAC dataset,and the obtained results indicate that our method has art-of-state detection performance.展开更多
文摘红外热成像测温是及时发现电缆附件异常发热缺陷的重要方法,但面对海量巡检图像,传统的人工诊断方式费时费力,且过分依赖人工经验。已有研究中为了实现电气设备红外图像的智能诊断,大多提取特定特征量作为所搭建神经网络模型的输入,而所提取的特征量也依赖于人工选择。为此,提出了一种基于Faster RCNN(faster regions with convolutional neural networks features)与Mean-Shift的电缆附件缺陷红外图像自动诊断方法。该方法首先基于Faster RCNN网络实现诊断对象的识别定位;之后利用Mean-Shift聚类算法提取过热区域;最后计算温度参数,并根据相应诊断标准得到诊断结果。利用实际巡检图像进行测试,结果表明:所提方法能够在不同拍摄角度、不同背景下准确定位诊断目标与过热区域,继而实现过热缺陷的自动诊断。研究对于实际工程中电缆附件的缺陷诊断具有一定的参考价值。
文摘针对田间密植环境棉花精准打顶时,棉花顶芽因其小体积特性所带来识别困难问题,该研究提出一种改进型快速区域卷积神经网络(Faster Region Convolutional Neural Networks,Faster R-CNN)目标检测算法实现大田环境棉花顶芽识别。以Faster R-CNN为基础框架,使用RegNetX-6.4GF作为主干网络,以提高图像特征获取性能。将特征金字塔网络(Feature Pyramid Network,FPN)和导向锚框定位(Guided Anchoring,GA)机制相融合,实现锚框(Anchor)动态自适应生成。通过融合动态区域卷积神经网络(Dynamic Region Convolutional Neural Networks,Dynamic R-CNN),实现训练阶段检测模型自适应候选区域(Proposal)分布的动态变化。最后在目标候选区域(Region of Interest,ROI)中引入目标候选区域提取器(Generic ROI Extractor,GROIE)提高图像特征融合能力。采集自然环境下7种不同棉花总计4819张图片,建立微软常见物体图像识别库2017(Microsoft Common Objects in Context 2017,MS COCO 2017)格式的棉花顶芽图片数据集进行试验。结果表明,该研究提出方法的平均准确率均值(Mean Average Precision,MAP)为98.1%,模型的处理帧速(Frames Per Second,FPS)为10.3帧/s。其MAP在交并比(Intersection Over Union,IOU)为0.5时较Faster R-CNN、RetinaNet、Cascade R-CNN和RepPoints网络分别提高7.3%、78.9%、10.1%和8.3%。该研究算法在田间对于棉花顶芽识别具有较高的鲁棒性和精确度,为棉花精准打顶作业奠定基础。
文摘小目标是指图像中覆盖区域较小的一类目标.与常规目标相比,小目标信息量少,训练数据难以标记,这导致通用的目标检测方法对小目标的检测效果不好,而专门为小目标设计的检测方法往往复杂度过高或不具有通用性.在分析现有目标检测方法的基础上,提出了一种面向小目标的多尺度快速区域卷积神经网络(faster-regions with convolutional neural network, Faster-RCNN)检测算法.根据卷积神经网络的特性,修改了Faster-RCNN的网络结构,使网络可以同时使用低层和高层的特征进行多尺度目标检测,提升了以低层特征为主要检测依据的小目标检测任务的精度.同时,针对训练数据难以标记的问题,使用从搜索引擎上获取的数据来训练模型.因为这些训练数据与任务测试数据分布不同,又利用下采样和上采样的方法对目标高分辨率的训练图像进行转化,使训练图像和测试图像的特征分布更类似.实验结果表明:所提出的方法在小目标检测任务上的平均精度均值(mean average precision, mAP)可以比原始的Faster-RCNN提高约5%.
文摘随着交通愈加发达,道路愈加拥堵,如何实时准确地获取车辆基本信息以便交通部门及时管理特定路段和路口的车辆显得日益重要.对交通视频中车辆的检测和识别,不仅需要实时检测,还要保证其准确性.针对实际情况中车辆之间的遮挡、光照的变化、阴影、道路旁树枝的晃动、背景中固定对象的移动等因素严重影响检测与识别的精度的问题,提出基于Faster-RCNN(Faster-Regions with CNN features)的车辆实时检测改进算法.首先采用k-means算法对KITTI数据集的目标框进行聚类,得到合适的长宽比,并增加一组尺度(64~2)以适应差异较大的车辆尺寸;然后改进区域提案网络,降低计算量,优化网络结构;最后在训练阶段采用多尺度策略,降低漏检率,提高精确率.实验结果表明:改进后的车辆检测算法的mAP(mean Average Precision)达到了82.20%,检测速率为每张照片耗时0.03875 s,基本能够满足车辆实时检测的需求.
基金This research was funded by the National Natural Science Foundation of China(grant number:61671470)the Key Research and Development Program of China(grant number:2016YFC0802900).
文摘It well known that vehicle detection is an important component of the field of object detection.However,the environment of vehicle detection is particularly sophisticated in practical processes.It is comparatively difficult to detect vehicles of various scales in traffic scene images,because the vehicles partially obscured by green belts,roadblocks or other vehicles,as well as influence of some low illumination weather.In this paper,we present a model based on Faster ReCNN with NAS optimization and feature enrichment to realize the effective detection of multi-scale vehicle targets in traffic scenes.First,we proposed a Retinex-based image adaptive correction algorithm(RIAC)to enhance the traffic images in the dataset to reduce the influence of shadow and illumination,and improve the image quality.Second,in order to improve the feature expression of the backbone network,we conducted Neural Architecture Search(NAS)on the backbone network used for feature extraction of Faster ReCNN to generate the optimal cross-layer connection to extract multi-layer features more effectively.Third,we used the object Feature Enrichment that combines the multi-layer feature information and the context information of the last layer after cross-layer connection to enrich the information of vehicle targets,and improve the robustness of the model for challenging targets such as small scale and severe occlusion.In the implementation of the model,K-means clustering algorithm was used to select the suitable anchor size for our dataset to improve the convergence speed of the model.Our model has been trained and tested on the UN-DETRAC dataset,and the obtained results indicate that our method has art-of-state detection performance.
