基于改进YOLOv5s的CNN-Swin Transformer森林野生动物图像目标检测算法

【目的】为提高野生动物在复杂森林环境中的检测精度,促进森林野生动物保护技术发展,提出一种基于YOLOv5s网络模型、针对陷阱相机所摄取森林野生动物图像的改进检测算法。【方法】以包含湖南壶瓶山国家级自然保护区几种典型森林野生动物在内的数据集为研究对象,首先,对真实标注框图像进行裁剪、归一化和缩放处理,随机将2~4张裁剪图像拼贴组成新的数据集元素,以丰富和增强数据集图像信息;其次,使用一种基于通道注意力思想的加权通道拼接方法,在通道拼接时引入权重改变通道数量,通过反向传播训练方法不断更新权重以增加重要特征信息的通道层数;接着,引入Swin Transformer模块与CNN网络相结合,为卷积神经网络特征提取加入自注意力机制,融合2种网络特征提取层的优势,提高特征提取的感受野;最后,选择更优的α-DIoU损失函数替代GIoU损失函数,针对边界框重叠面积和中心点距离造成的损失,引入新的几何因素惩罚项。【结果】在相同试验条件和数据集下,相比原YOLOv5s网络模型,改进算法极大提高检测的平均准确率和平均回归率,均值平均精度由74.1%提升至88.4%,获得14.3%的精度提升,同时也超过YOLOv3、YOLOXs、RetinaNet、Faster R-CNN等其他流行目标检测算法。【结论】针对陷阱相机所摄取森林野生动物图像背景与目标对比度低、遮挡重叠严重,致使检测误检率、漏检率高等问题,在检测算法中提出一系列改进措施,为我国森林野生动物的保护和数据获取提供一种新的可行性方案和思路。

基于Transformer改进YOLOv5的交通标志检测算法

交通标志检测作为自动驾驶的组成部分直接影响着行车安全。针对现有算法对图像中尺寸小、被遮挡的标志存在漏检、误检的问题,文中提出了基于改进YOLOv5的交通标志检测算法。首先对原模型注意力缺失的问题经过对比后构建了BiFormer-y,使模型可以更好获取长期依赖;接着针对层数较深造成的具有丢失特征的缺陷,利用残差结构重新设计检测层,从而更好地保留特征;最后对耦合头的空间错位问题引入解耦头并进行优化。CCTSDB2021的实验表明,精确率、召回率、mAP分别为97.0、95.9、97.9与先进工作相比具有明显优势。

基于Transformer与信息融合的绝缘子缺陷检测方法

针对现有绝缘子航拍图像背景复杂,闪络、破损缺陷检测困难的问题,本文提出一种全局与局部信息融合(GLIF)-YOLOv8s绝缘子缺陷检测算法。该算法采用EfficientFormerV2作为主干网络,以提高模型对全局信息的提取能力;基于全局与局部信息设计特征增强模块,通过信息融合减少深层网络信息的丢失。在绝缘子缺陷数据集上进行消融实验与对比实验,结果表明:本文算法在绝缘子缺陷数据集上的平均精度均值为91.6%,其对闪络和破损缺陷的检测平均精度分别达到82.3%和92.9%;与其他主流算法相比,本文算法的检测框置信度更高。

Over Voltage Fault due to Disconnection of Consumer’s Transformer Neutral Wire

Practically,the load currents in three phases are asymmetric in the power system.It means that the impedances are different in all three phases.If the consumer’s transformer neutral cut off and/or was disconnected from the neutral of power supply source,then there will be some trouble and failure occurred.The current in the neutral wire drops down to zero when the neutral wire is cut off and the phase currents of all three-phase equal to each other since there was no return wire.The currents are equal but the voltages at the phase consumers are different.Especially for residential single-phase consumers,the voltage at the consumers of the phase varies differently for three phase systems when the neutral wire was disconnected at consumer side and even the voltage at the consumers one or two of those three phases becomes over nominal voltage or reaches nearly line voltage.In this case,the electronic appliances in that phase will be fed by high voltage than the rated value and they can be broken down.In the power system of UB(Ulaanbaatar)city,there are some occasional such kind of failures every year.Obviously,many electronic appliances were broken down due to high voltage and the electricity utility companies respond for service charge of damaged parts.

Study on Over-Voltage of 220 kV Transformer＇s Neutrals and Protection Strategy

In order to limit short-circuit current and satisfy the need of relay setting, only part of the 220 kV power transformer is grounded, so on the neutrals of ungrounded transformers will appear the over-voltage. Currently the value of over-voltage on transformer neutral point and the corresponding protection strategy is based on theoretical formula. This article uses the PSCAD/EMTDC software to calculate the over-voltage on the neutral point of 4 ungrounded power transformers in Chongqing 220 kV power grid. The result shows that the power frequency transient over-voltage on the neutrals may reach 178 kV. If the single-phase grounding fault occurs in ungrounded power system, the power frequency over-voltage on the neutrals will be more serious and rise to 138.6 kV. Non-full phase operation may cause serious ferro-resonance over-voltage on the neutrals of no-load transformer, which may last for seconds and may rise to 723.7 kV, causing serious threat to the transformer neutrals and line-side equipment. The article also studies the gap parameter which should be taken on the neutrals of 220 kV transformers at the end.

基于S-YOLO V5和Vision Transformer的视频内容描述算法

视频内容描述的自动生成是结合计算机视觉和自然语言处理等相关技术提出的一种新型交叉学习任务。针对当前视频内容生成描述模型可读性不佳的问题,本研究提出一种基于S-YOLO V5和Vison Transformer(ViT)的视频内容描述算法。首先,基于神经网络模型KATNA提取关键帧,以最少帧数进行模型训练;其次,利用S-YOLO V5模型提取视频帧中的语义信息,并结合预训练ResNet101模型和预训练C3D模型提取视频静态视觉特征和动态视觉特征,并对两种模态特征进行融合;然后,基于ViT结构的强大长距离编码能力,构建模型编码器对融合特征进行长距离依赖编码;最后,将编码器的输出作为LSTM解码器的输入,依次输出预测词,生成最终的自然语言描述。通过在MSR-VTT数据集上进行测试,本研究模型的BLEU-4、METEOR、ROUGEL和CIDEr分别为42.9、28.8、62.4和51.4;在MSVD数据集上进行测试,本研究模型的BLEU-4、METEOR、ROUGEL和CIDEr分别为56.8、37.6、74.5以及98.5。与当前主流模型相比,本研究模型在多项评价指标上表现优异。

融合Transformer和改进PANet的YOLOv5s交通标志检测

针对交通标志检测速度慢和目标大小与类别极度不平衡等问题,提出一种融合Transformer和改进PANet网络的YOLOv5s交通标志检测算法。首先在不增加模型复杂度的前提下,将主干网络末端与Transformer融合以提高网络特征提取能力;其次由于所采用交通标志数据集的目标尺度太小,导致网络32倍大尺度检测层检测效果不佳,故不采用相关网络层,同时采用K-means算法得出适合的预测候选框;然后改进损失函数以解决正负样本极度不平衡问题。最后将所提出的改进算法在Jetson AGX Xavier平台上部署验证。实验结果表明,所提算法检测性能更佳,其准确率和召回率在原网络的基础上分别提高了2.2%和0.7%,模型参数量和计算复杂度分别减少了25.8%和10.1%。在Xavier上的检测速度达到76FPS,满足实时交通标志检测的要求且易于在实际场景部署。

Transforming growth factor -β neutralizing antibodies inhibit subretinal fibrosis in a mouse model

AIM:To determine the involvement of the transforming growth factor(TGF)-β with the development of experimental subretinal fibrosis in a mouse model.· METHODS:Subretinal fibrosis was induced by subretinal injection of macrophage-rich peritoneal exudate cells(PECs) and the local expression of TGF-β isoforms was assessed by quantitative real-time reverse transcription-polymerase chain reaction(RT-PCR) and enzyme-linked immunosorbent assay(ELISA) at various time points.In addition,we investigated the effect of TFG-β-neutralizing antibodies(TGF-β NAb) on subretinal fibrosis development.· RESULTS:TGF-β1 and TGF-β2 mRNA level was significantly elevated at day 2 after subretinal fibrosis induction and increased further to 5 and 6.5-fold respectively at day 5,reaching the peak.TGF-β3 mRNA was not detected in the present study.The result of ELSIA showed that active TGF-β1 and TGF-β2 levels were upregulated to 10-fold approximately,while total TGF-β1 and TGF-β2 levels were even upregulated more than 10-fold and more than 20-fold respectively in subretinal fibrosis mice in comparison with na?觙ve mice at day 5.TGF-β NAb resulted in a reduced subretinal fibrosis areas by 65% compared to animals from control group at day 7.· CONCLUSION:Our results indicate that TGF-β signaling may contribute to the pathogenesis of subretinal fibrogenesis and TGF-β inhibition may provide an effective,novel treatment of advanced and late-stage neovascular age-related macular degeneration.·

Application of fuzzy analytic hierarchy process and neural network in power transformer risk assessment

In operation,risk arising from power transformer faults is of much uncertainty and complicacy.To timely and objectively control the risks,a transformer risk assessment method based on fuzzy analytic hierarchy process(FAHP) and artificial neural network(ANN) from the perspective of accuracy and quickness is proposed.An analytic hierarchy process model for the transformer risk assessment is built by analysis of the risk factors affecting the transformer risk level and the weight relation of each risk factor in transformer risk calculation is analyzed by application of fuzzy consistency judgment matrix;with utilization of adaptive ability and nonlinear mapping ability of the ANN,the risk factors with large weights are used as input of neutral network,and thus intelligent quantitative assessment of transformer risk is realized.The simulation result shows that the proposed method increases the speed and accuracy of the risk assessment and can provide feasible decision basis for the transformer risk management and maintenance decisions.

Application of sparse S transform network with knowledge distillation in seismic attenuation delineation

Time-frequency analysis is a successfully used tool for analyzing the local features of seismic data.However,it suffers from several inevitable limitations,such as the restricted time-frequency resolution,the difficulty in selecting parameters,and the low computational efficiency.Inspired by deep learning,we suggest a deep learning-based workflow for seismic time-frequency analysis.The sparse S transform network(SSTNet)is first built to map the relationship between synthetic traces and sparse S transform spectra,which can be easily pre-trained by using synthetic traces and training labels.Next,we introduce knowledge distillation(KD)based transfer learning to re-train SSTNet by using a field data set without training labels,which is named the sparse S transform network with knowledge distillation(KD-SSTNet).In this way,we can effectively calculate the sparse time-frequency spectra of field data and avoid the use of field training labels.To test the availability of the suggested KD-SSTNet,we apply it to field data to estimate seismic attenuation for reservoir characterization and make detailed comparisons with the traditional time-frequency analysis methods.

Large Power Transformer Fault Diagnosis and Prognostic Based on DBNC and D-S Evidence Theory

Power transformer is a core equipment of power system, which undertakes the important functions of power transmission and transformation, and its safe and stable operation has great significance to the normal operation of the whole power system. Due to the complex structure of the transformer, the use of single information for condition-based maintenance (CBM) has certain limitations, with the help of advanced sensor monitoring and information fusion technology, multi-source information is applied to the prognostic and health management (PHM) of power transformer, which is an important way to realize the CBM of power transformer. This paper presents a method which combine deep belief network classifier (DBNC) and D-S evidence theory, and it is applied to the PHM of the large power transformer. The experimental results show that the proposed method has a high correct rate of fault diagnosis for the power transformer with a large number of multi-source data.

融合YOLOv5s与Swin Transformer的森林火灾检测

针对传统的森林火灾检测方法检测精度不佳、可靠性低等不足,提出一种基于YOLOv5s融合Swin Transformer的火灾检测方法.为实现森林火灾的实时性检测,提出了YOLOv5s-SwinT的改进识别方法.基于Transformer模型的应用,解决了卷积神经网络的运算局部性以及全局特征提取等不足.融合Swin Transformer与YOLOv5s卷积神经网络模型,并将其应用于森林火灾检测的机器视觉任务中.引入α-IoU损失函数替换GIOU损失函数,并在骨干网络中融入CA注意力机制轻量模块,提升了整体网络的特征提取能力以及获取高质量和高精准度的定位图像区域,进行边界框生成及预测,改善小目标检测漏检及检测精度差等问题.实验结果表明,融合YOLOv5s-Swin T的改进识别方法在森林火灾检测任务中,可实现mAP值达到74.2%,相比YOLOv5s提高了4.5%,并设计了GUI界面直接部署到PC端实现实时火灾检测需求,为森林火灾检测视觉任务提供了一种有效的检测方法.

A New Approach with Three Dimension Figure and ANSI/IEEE C57.104 Standard Rule Diagnoses Transformer’s Insulating Oil

The dissolved gas analysis (DGA) is an effective method for detecting incipient faults in immersed oil power transformers. In this paper, we investigate the DGA methods and employ the ANSI/IEEE C57.104 standards (guidelines for the interpretation of gases generated in oil-immersed transformers) and IEC Basic Gas Ratio method to design a heuristic power transformer fault diagnosis tool in practice. The proposed tool is implemented by a MATLAB program and it can provide users a transformer diagnosis result. The user keys in the data of H2, CH4, C2H2, C2H4, and C2H6 gases dissolved from the immersed oil transformer’s insulating oil measured by ASTM D3612. The analyzed results will be represented in texts and figures. The real measured data of the transformer oil were taken from Taiwan Power Company substations to verify the validation and accuracy of the developed diagnosis tool.

Difficulties in Parameters Identification of Traction Transformer Multi-port Equivalent Scheme with Hidden Construction Defect

While modeling a power supply system for an electric railway traction, knowing equivalent circuits of locomotives supplied this way is an essential issue. In alternating current traction, it is important to diagnose inter alia processes taking place in transformers installed on electric vehicles. This article presents specific phenomena occurring during the work of mono-phase, multi-winding, multisystem （systems AC： 50 Hz, 16.7 Hz） laboratory traction transformer. It also shows difficulties encountered during the process of identifying multi-port equivalent scheme＇s elements of the described device, in which a construction defect occurs.

A Contribution to the Analysis of Frequency Response and Impedance Measurements to Evaluation and Diagnosis of Power Transformers

This paper has an objective to show a developed quantitative criterion,based in two mathematical variables that explicit the deviation degree of a normal situation,applying simultaneously data from terminal impedances and frequency response.Based in more than 100-measured equipment,of different applications(step-up transformer,transmission transformer,etc.,),for a period of 10 years,the work presents some examples of practical application of this methodology in Brazilian Electrical System.

Study on Power Transformers Fault Diagnosis Based on Wavelet Neural Network and D-S Evidence Theory

>Transformer faults are quite complicated phenomena and can occur due to a variety of reasons.There have been several methods for transformer fault synthetic diagnosis,but each of them has its own limitations in real fault diagnosis applications.In order to overcome those shortcomings in the existing methods,a new transformer fault diagnosis method based on a wavelet neural network optimized by adaptive genetic algorithm(AGA)and an improved D-S evidence theory fusion technique is proposed in this paper.The proposed method combines the oil chromatogram data and the off-line electrical test data of transformers to carry out fault diagnosis.Based on the fusion mechanism of D-S evidence theory,the comprehensive reliability of evidence is constructed by considering the evidence importance,the outputs of the neural network and the expert experience.The new method increases the objectivity of the basic probability assignment(BPA)and reduces the basic probability assigned for uncertain and unimportant information.The case study results of using the proposed method show that it has a good performance of fault diagnosis for transformers.

A Modeling Method for Transformer Windings Under VFTO Based on S-parameters

To study the Very Fast Transient Over-voltage (VFTO) distribution in transformer windings in gas insulated substation (GIS), a systematic methodology based on S-parameters is presented for establishing high-frequency model of transformer windings. Firstly, voltage transfer functions are derived from S-parameters which are calculated or measured from transformer windings. Secondly, voltage transfer functions are fitted with rational functions by the vector fitting method and then the rational transfer functions are order-reduced by optimal Pade-approximation algorithm. Lastly, the resultant voltage transfer functions are synthesized by network technology. Computational results are consistent with simulation results of Electromagnetic Transient Program (EMTP) and confirm the feasibility and validity of proposed methodology.

An Effective Detection of Inrush and Internal Faults in Power Transformers Using Bacterial Foraging Optimization Technique

Power transformers in transmission network are utilized for increasing or decreasing the voltage level. Power Transformers fail to connect directly to the consumers that result in the less load fluctuations. Power transformer operation under any abnormal condition decreases the lifetime of the transformer. Power Transformer protection from inrush and internal fault is critical issue in power system because the obstacle lies in the precise and swift distinction between them. Due to the limitation of heterogeneous resources, occurrence of fault poses severe problem. Providing an efficient mechanism to differentiate between faults (i.e. inrush and internal) is the key for efficient information flow. In this paper, the task of detecting inrush and internal fault in power transformers is formulated as an optimization problem which is solved by using Hyperbolic S-Transform Bacterial Foraging Optimization (HS-TBFO) technique. The Gaussian Frequency- based Hyperbolic S-Transform detects the faults at much earlier stage and therefore minimizes the computation cost by applying Cosine Hyperbolic S-Transform. Next, the Bacterial Foraging Optimization (BFO) technique has been proposed and has demonstrated the capability of identifying the maximum number of faults covered with minimum test cases and therefore improving the fault detection efficiency in a wise manner. The HS-TBFO technique is evaluated and validated in various simulation test cases to detect inrush and internal fault in a significant manner. This HS-TBFO technique is investigated based on three phase power transformer embedded in a power system fed from both ends. Results have confirmed that the HS-TBFO technique is capable of categorizing the inrush and internal faults by identifying maximum number of faults with minimum computation cost as compared to the state-of-the-art works.

Amplitude spectrum compensation and phase spectrum correction of seismic data based on the generalized S transform

We propose a method for the compensation and phase correction of the amplitude spectrum based on the generalized S transform. The compensation of the amplitude spectrum within a reliable frequency range of the seismic record is performed in the S domain to restore the amplitude spectrum of reflection. We use spectral simulation methods to fit the time-dependent amplitude spectrum and compensate for the amplitude attenuation owing to absorption. We use phase scanning to select the time-, space-, and frequencydependent phases correction based on the parsimony criterion and eliminate the residual phase effect of the wavelet in the S domain. The method does not directly calculate the Q value; thus, it can be applied to the case of variable Q. The comparison of the theory model and field data verify that the proposed method can recover the amplitude spectrum of the strata reflectivity, while eliminating the effect of the residual phase of the wavelet. Thus, the wavelet approaches the zero-phase wavelet and, the seismic resolution is improved.

Ground roll attenuation using a time-frequency dependent polarization filter based on the S transform

The ground roll and body wave usually show significant differences in arrival time, frequency content, and polarization characteristics, and conventional polarization filters that operate in either the time or frequency domain cannot consider all these elements. Therefore, we have developed a time-frequency dependent polarization filter based on the S transform to attenuate the ground roll in seismic records. Our approach adopts the complex coefficients of the S transform of the multi-component seismic data to estimate the local polarization attributes and utilizes the estimated attributes to construct the filter function. In this study, we select the S transform to design this polarization filter because its scalable window length can ensure the same number of cycles of a Fourier sinusoid, thereby rendering more precise estimation of local polarization attributes. The results of applying our approach in synthetic and real data examples demonstrate that the proposed polarization filter can effectively attenuate the ground roll and successfully preserve the body wave.