Xiao Yu

Insulator defect detection is of great significance to ensure the normal operation of power transmission and distribution networks. In response to the problems of low speed, low accuracy, and difficulty in deploying to embedded terminals in existing insulator defect detection, this paper proposes a lightweight insulator defect detection model based on an improved YOLOv5s, named LiteYOLO-ID. Firstly, to significantly reduce the model parameters while maintaining detection accuracy, we design a new lightweight convolution module called EGC (ECA-GhostNet-C2f). Secondly, based on the EGC module, we construct the EGC-CSPGhostNet backbone network, which optimizes the feature extraction process and achieves model compression. Additionally, we design a lightweight neck network, EGC-PANet, to further reduce the parameter count and achieve efficient feature fusion. Experimental results show that on the IDID-Plus dataset, compared to the original YOLOv5s model, not only does LiteYOLO-ID reduce the model parameters by 47.13%, but it also improves the mAP(0.5) by 1%. Furthermore, the generalization of the model is validated on the Pascal VOC dataset and the SFID dataset. Importantly, after TensorRT optimization, the inference speed of the LiteYOLO-ID algorithm on the Jetson TX2 NX reaches 20.2 FPS, meeting the real-time detection requirements of insulator defects. Our code, weight models, and datasets can be obtained at the following URL: https://github.com/LuYang-2023/Insulator-defect-detection.

In recent years, deep learning has been widely applied in the diagnosis of oil and gas pipeline failures due to its powerful feature representation capabilities. However, in practical applications, the diagnostic accuracy falls short of meeting actual requirements due to complex environmental interference, coupled with existing methods that focus solely on spatial feature relationships. To improve diagnostic performance, this paper eliminates the influence of complex backgrounds through optimal neural-immune domain in the image preprocessing stage. Subsequently, an Immune Depth Presentation Network capable of effectively capturing inter-feature correlations in the channel dimension is proposed, and it is integrated with a conventional convolutional neural network to construct the Immune Depth Presentation Convolutional Neural Network model. To validate the effectiveness and stability of the model, we compared the proposed algorithm in this paper with classic algorithms such as BPNN, KNN, VGG16, VGG19, Resnet34, and Resnet50. We conducted multiple experiments, and the results indicate that the proposed algorithm achieved the highest average testing accuracy, demonstrating stable performance.

Recently, with the development of intelligent search and rescue system, humans trapped in disaster detection plays an increasingly crucial role in our daily life, and a sea of related technologies have been proposed. However, most of these algorithms have a low accuracy rate in detecting persons in sophisticated backgrounds. In this paper, we come up with a human detection algorithm based on the improved I-Unet to solve the challenge of low people-recognition rate in complex disaster scenarios. This paper replaces traditional convolution with dilated convolution to extract more features from images. In the meanwhile, we add batch normalization after convolution to speed up the training process. To address the problem of class imbalance, this paper uses the focal loss function to prevent confusion between the human category and other confusing types. Experimental results show that this algorithm applied in human detection is effective in complex scenes, and more accurate than the Unet-based method.

Due to the advantages of safety and economy, it has become a trend to use unmanned aerial vehicles (UAVs) instead of humans to inspect high-voltage transmission lines. Considering the manual inspection process and the few-shot learning, a two-stage method for insulator anomaly detection is proposed. In the first stage, a positioning-restoration-cropping method is discussed for insulator string detection and processing. In the second stage, an insulator anomaly detection model called a multi-scale feature reweighting (MFR) network is built. With the help of few-shot object detection, the detection of five kinds of anomaly insulator caps, such as falling off, breakage and ablation is realized. The mean average precision (mAP) of the proposed method is 88.76%.

In criminal cases, the information collection and analysis of the scene plays an important role in solving cases. Handprint mark can be easily remaining and frequently occurring mark of criminals. It may occasionally be recognized as potential physical evidence during criminal investigations. The infrared thermal imager can fully collect the thermal radiation information of objects, so it is convenient to get the remaining heat source marks in the environment, and to obtain key information that is missing from other methods. Handprints tend to disappear over time. Like cell apoptosis is regulated by a rigorous mechanism of a series of genes, handprints also regulated by the mechanism of mark disappearance affected by surrounding pixels. In fact, we can easily get handprint images which hands left a period of time, but hope to restore the initial handprint image and obtain the characteristics of the criminal hand. Therefore, a method of infrared handprint image restoration based on apoptotic mechanism is proposed. Training depends on the existing infrared handprint images, and we can find a relational model between the former and the latter images. Finally, the rough shape of the original infrared handprint can be deduced.