共查询到20条相似文献,搜索用时 46 毫秒
1.
轻量级神经网络的出现显著降低了目标检测算法在移动端部署的难度,当前已有许多运算量小、精度较高的卷积神经网络在多个公共数据集上取得了不错的效果。然而,在基于图像的电力巡检领域,图像中目标检测与识别的效率对于及时排除电力故障具有重要意义,尤其是针对基于无人机的巡检,实现在线实时的故障检测更有意义。为了实现绝缘子和间隔棒这些重要巡检目标的移动端实时检测,本文提出了一种基于YOLOV5n的针对电力设备检测与分类的轻量级网络模型,算法在YOLOV5n的基础上优化网络,通过减少一系列的卷积层并舍去一部分的捷径分支,提高网络的并行程度并降低网络的深度。最终设计出模型更轻量、精确度更高的YOLOV5n-1、YOLOV5n-2,基于自建的电力巡检数据集进行测试,实验结果表明,提出的算法比YOLOV5n减少了27%的GFLOPs,检测时间降低了24%,降低了硬件要求,更适合在移动端部署。 相似文献
2.
为了提高Tiny YOLOV3目标检测算法在行人检测任务中的准确率,对该算法进行了研究改进。首先对Tiny YOLOV3的特征提取网络进行深化,增强网络特征提取能力;然后在预测网络的两个检测尺度分别加入通道域注意力机制,对特征图的不同通道赋予不同的权重,引导网络更多关注行人的可视区域;最后,改进激活函数和损失函数并采用K-means聚类算法重新选择初始候选框。实验结果表明,改进后Tiny YOLOV3算法的准确率在VOC2007行人子集上达到77%,较Tiny YOLOV3提高8.5%,在INRIA数据集上达到92.7%,提高2.5%,运行速度分别达到每秒92.6帧和31.2帧。本文方法提高了行人的检测精度,保持了较快的检测速度,满足实时性运行需求。 相似文献
3.
4.
道路伤损检测是道路养护过程中的重要基础性环节,传统道路伤损检测方法存在检测成本高且效率低的缺陷。为准确快速检测道路伤损状况,提出了一种基于YOLOv5改进的道路伤损检测模型YOLO-C-α。通过引入注意力机制CBAM模块,提高检测模型的特征提取及特征融合能力,改善模型对道路损伤小目标的漏检问题;引入α-IoU损失函数替换原始网络CIOU损失函数,降低预测框的回归损失,提升预测框的定位精度。基于RDD2020道路伤损检测数据集展开对比实验,结果显示:YOLO-C-α模型平均准确度达到60.3%,相比于原始模型平均精度提升1.4%,其F1值为60.2,相比于原始模型提升1%,且对于不同天气状况下的路面损伤均有较高的检测性能,实验环境每张图片的检测速度为6.3 ms,模型大小40.6 Mb。结果表明:本文基于YOLOv5m改进的算法抗干扰能力较强,能更准确地检测出多种天气状况下道路伤损目标,可为道路伤损实时检测及智慧化道路养护提供参考。 相似文献
5.
6.
为了提高交通目标检测的精度和效率,提出一种改进YOLOv5s的交通场景多目标检测方法,在YOLOv5s的主干网络中引入高效的层聚合网络结构来提高模型学习目标特征的能力,引入了通道注意力和空间注意力结合的卷积注意力模块(BAM)机制,进一步提高网络模型的特征提取能力,通过采用α-IoU作为边界框回归损失函数,提高了边界框回归精度。实验结果表明,改进的目标检测模型相较于YOLOv5s原模型在检测精度上提升了2.4%,模型参数量和模型大小分别降低了20.9%和19.1%。实现了在不同时间段准确且高效的检测交通场景的多种目标,保证了实时检测的应用需求。 相似文献
7.
针对目前钢板缺陷检测精度和速度的不足,提出了一种改进的YOLOv3检测算法.首先使用小波-中值滤波处理缺陷图像,清除图像里的噪声使图像更平滑.然后在原有网络中的密集连接网络(Darknet-53)上增加一个尺度输出增强算法对小目标缺陷的识别能力.最后为了增强算法模型的准确性对算法原有的损失函数进行优化,得到改进版的YO... 相似文献
8.
针对传统钢板表面缺陷检测方法效果差、缺陷定位不准确等问题,提出一种基于改进RetinaNet-GHM的深度学习检测算法。首先,引入路径聚合特征金字塔网络融合浅层和深层语义信息,提升网络对小目标的检测效果;然后,使用GHMC和GHMR损失函数对缺陷进行分类和定位;最后,引入高斯形式的软化非极大值抑制算法,提高检测精度。实验结果表明,改进的RetinaNet-GHM算法的平均精度均值为76.7%,裂纹、夹杂、斑块、麻点、压入氧化铁皮以及划痕六类缺陷的平均精度分别为45.2%、88.2%、94.2%、86.1%、65.1%和87.4%。通过与其他经典算法相比,改进的RetinaNet-GHM算法具有较好的检测效果 相似文献
9.
针对目前安全帽佩戴检测模型在复杂环境下对于密集目标和遮挡目标存在漏检的问题,基于YOLOv3框架,提出了一种新的安全帽佩戴检测算法YOLOv3-H。YOLOv3-H算法在主干网络提取的3个特征层后加入多尺度自注意力机制,提升模型捕捉有用信息的能力;同时对K-means聚类算法进行优化,获得适合安全帽佩戴检测的最优先验框,从而提升网络模型的检测精度;最后采用GIOULoss作为目标定位损失函数,使得网络可以沿着预测框与真实框重叠度高的方向进行优化,从而加快模型的收敛速度。在Safety Helmet Wearing detect公开数据集上进行实验,结果表明,YOLOv3-H算法相比原有YOLOv3,平均检测精度提升了7.05%。 相似文献
10.
11.
Clarence W. DE SILVA 《国外电子测量技术》2021,8(1):14-28
The enhanced definition of Mechatronics involves the four underlying characteristics of integrated, unified, unique, and systematic approaches. In this realm, Mechatronics is not limited to electro-mechanical systems, in the mul-ti-physics sense, but involves other physical domains such as fluid and thermal. This paper summarizes the mecha-tronic approach to modeling. Linear graphs facilitate the development of state-space models of mechatronic systems, through this approach. The use of linear graphs in mechatronic modeling is outlined and an illustrative example of sound system modeling is given. Both time-domain and frequency-domain approaches are presented for the use of linear graphs. A mechatronic model of a multi-physics system may be simplified by converting all the physical do-mains into an equivalent single-domain system that is entirely in the output domain of the system. This approach of converting (transforming) physical domains is presented. An illustrative example of a pressure-controlled hydraulic actuator system that operates a mechanical load is given. 相似文献
12.
13.
This paper deals with instrumenting a mechatronic system, through the incorporation of suitable sensors, actuators, and other required hardware. Sensors (e.g., semiconductor strain gauges, tachometers, RTD temperature sensors, cameras, piezoelectric accelerometers) are needed to measure (sense) unknown signals and parameters of a system and its environment. The information acquired in this manner is useful in operating or controlling the system, and also in process monitoring; experimental modeling (i.e., model identification); product testing and qualification; product quality assessment; fault prediction, detection and diagnosis; warning generation; surveillance, and so on. Actuators (e.g., stepper motors, solenoids, dc motors, hydraulic rams, pumps, heaters/coolers) are needed to “drive” a plant. Control actuators (e.g., control valves) perform control actions, and in particular they drive control devices. Micro-electromechanical systems (MEMS) use microminiature sensors and actuators. MEMS sensors commonly use piezoelectric, capacitive, electromagnetic and piezoresistive principles. MEMS devices provide the benefits of small size and light weight (negligible loading errors), high speed (high bandwidth), and convenient mass-production (low cost). The process of instrumentation involves the identification of proper sensors, actuators, controllers, and signal modification/interface hardware, and software with respect to their functions, operation, parameters, ratings, and interaction with each other, so as to achieve the performance requirements of the overall system, and interfacing/integration/tuning of the selected devices into the system, for a given application. This paper presents the key steps of instrumenting a mechatronic system, in a somewhat general and systematic manner. Examples are described to illustrate several key procedures of instrumentation. 相似文献
14.
Clarence W. DE SILVA 《国外电子测量技术》2019,6(4):72-84
An engineering system may consist of several different types of components, belonging to such physical “domains” as mechanical, electrical, fluid, and thermal. It is termed a multi-domain (or multi-physics) system. The present paper concerns the use of linear graphs (LGs) to generate a minimal model for a multi-physics system. A state-space model has to be a minimal realization. Specifically, the number of state variables in the model should be the minimum number that can completely represent the dynamic state of the system. This choice is not straightforward. Initially, state variables are assigned to all the energy-storage elements of the system. However, some of the energy storage elements may not be independent, and then some of the chosen state variables will be redundant. An approach is presented in the paper, with illustrative examples in the mixed fluid-mechanical domains, to illustrate a way to recognize dependent energy storage elements and thereby obtain a minimal state-space model. System analysis in the frequency domain is known to be more convenient than in the time domain, mainly because the relevant operations are algebraic rather than differential. For achieving this objective, the state space model has to be converted into a transfer function. The direct way is to first convert the state-space model into the input-output differential equation, and then substitute the time derivative by the Laplace variable. This approach is shown in the paper. The same result can be obtained through the transfer function linear graph (TF LG) of the system. In a multi-physics system, first the physical domains have to be converted into an equivalent single domain (preferably, the output domain of the system), when using the method of TFLG. This procedure is illustrated as well, in the present paper. 相似文献
15.
本文分析了压电射流角速度传感器的噪声特点,提出了利用自适应数字滤波抑制其噪声,归纳了自适应算法中的快速式滤波(FTF)最小二乘法的算法步骤,并且作用8098单片微机实现对该传感器的实时FTF滤波,给出了滤波效果实例。 相似文献
16.
称重传感器神经网络补偿器设计 总被引:6,自引:0,他引:6
吴忠强 《电子测量与仪器学报》2004,18(1):42-46
为满足快速称重的要求,设计出一种新型的称重传感器神经网络补偿器.仿真表明有效的提高了称重传感器的动态响应性能,有推广应用价值. 相似文献
17.
以区块链为代表的去中心化技术因其突出的可靠性和安全性已被广泛应用于分布式能源领域。区块链分片技术可以克服传统区块链系统低吞吐量、低可扩展性的性能缺陷,然而现有区块链分片系统大多采用随机分片模式,没有考虑系统中节点的能力差异以及分片的安全隐患。为了解决这类问题,文中提出了一种在分布式能源交易场景下区块链分片的匹配算法,该方法充分考虑了分布式能源交易中节点对节点间电气距离和分片内交易数的偏好,以降低能源传输损耗和交易成本,同时考虑节点信誉值,以反映其过去在交付承诺能源方面的表现,保证分片的安全性,并将问题抽象为具有外部性的多对一匹配模型进行求解,最后通过仿真实验验证了所提出的区块链分片方法优于传统的分片方法,具有良好的应用价值。 相似文献
18.
设计了一种基于Zigbee的无线传感器网络节点定位的矿下机车监控系统.在基于模糊思想的RSSI定位算法,采用了Sugeno 测度模型的匹配节点定位算法,并结合煤矿监测系统的实际应用,使此系统更加便于矿井机车的生产管理,提高安全性. 相似文献
19.
20.
崔海朋 《可编程控制器与工厂自动化(PLC FA)》2008,(10):114-116
通过对双光纤光栅并联匹配解调原理的研究,提出了基于MSP430F1612单片机的光纤光栅传感器解调系统。利用曲线拟合的方法对压电陶瓷驱动电压与波长偏移量的数据进行拟合计算,实现光纤光栅解调的目的。本文给出了该方案软硬件设计,并对匹配解调法的发展前景进行了论述。 相似文献