Insulators need to be regularly inspected to ensure the normal operation of the power system. Currently, the detection of insulators in power distribution network poses greater challenges due to the wider variety of types and smaller differences between classes. Furthermore, most of distribution network insulators are small targets, background will generate significant interference. To enhance detail in feature maps, a detail feature enhancement module is proposed. In the proposed module, shallow features are integrated into deep features, and object regions of the feature maps are adjusted by attention mechanism. Additionally, to enhance the discriminant ability of classification features, an auxiliary classification module is introduced. Typical insulators are used as examples and category score of the detected insulators are adjusted, thereby improving the ability to distinguish similar insulators. Finally, the proposed auxiliary classification module is employed to construct an insulator detection algorithm based on Faster R-CNN. Notably, the experimental results conducted on a distribution network insulator dataset show that the proposed algorithm can achieve a mean average precision of 96.6%, which can accurately identify the majority of insulators.

Power line detection plays an important role in automatic power inspection systems based on the vision sensor and unmanned aerial vehicle (UAV). The positioning of the power line is essential to assist in planning and navigation along the power line during the UAV-based power inspection process. For the along-line inspection problem of UAV power inspection, we developed a power line extraction (PLE) algorithm based on parallel line detection and straight line stretching. The proposed algorithm first extracts the image edges by the Canny algorithm and then constructs criterion functions for PLE based on power line feature information, e.g., length, width, and direction. Finally, the information on power line location, direction, width, and so on will be combined to get the exact power line regions. In order to effectively evaluate the accuracy and efficiency of the PLE algorithm, we constructed a power line dataset for UAV inspection and proposed a PLE effect evaluation method. The experimental results demonstrate that the proposed method can effectively reduce the influence of ground environment and other interference factors. Compared with other methods, the proposed algorithm is significantly better than the comparison methods in detecting power lines for the constructed dataset. Moreover, it is still a higher detection accuracy than the comparison methods when the power line is far away from UAV.

In recent years, point clouds have been widely used in powerline inspection, smart cities, autonomous driving, and other fields. The deep learning-based point cloud processing methods have attracted more and more attention due to the developments of laser scanning technology and machine learning. However, the recent methods largely ignore global contextual relationships and do not make full use of the complementation between local feature and high-level geometric information. To the problem, we propose a novel deep neural network, namely, the Dense connection-based Kernel Point Network (DenseKPNet), which can greatly expand the receptive field of kernel point convolution to extract rich semantic context information and valuable geometric features from the local region effectively. Specifically, we first design a multiscale convolution kernel point module to extract initial geometric features from coarse to fine. Then, we design the dense connection module to efficiently learn more expressive local geometric features while capturing rich contextual information. In addition, we propose the kernel point convolution attention module (KPCAM), which can capture global interdependencies between points and strengthen the discriminativeness of effective features. We evaluate our method on public indoor and outdoor datasets. The qualitative and quantitative experimental results show the effectiveness of DenseKPNet. The mIoU of the proposed method on S3DIS and semantic3D datasets can reach 68.9% and 77.9%, respectively.

Key objects semantic segmentation and power lines extraction from the electric power corridor point cloud are critical steps in power line inspection. However, the massive amount of point cloud data and missing power line points pose a challenge to the object extraction. To complete extraction of power lines and other essential objects, a method called SS-IPLE is proposed, which is based on Pointwise Multilayer-Perceptron semantic segmentation network. The method consists of two main parts: electric power corridor semantic segmentation and individual power lines extraction. In the segmentation step, SCF-Net is employed as our primary segmentation network, and the network can process large-scale point clouds. To further improve the segmentation ability of SCF-Net in the corridor, a local coding module (LCM) is designed to construct the SCFL-Net. In the individual power lines extraction step, individual power lines are extracted by flexible grid filtering, effectively overcoming the point missing problem. The corridor point cloud semantic segmentation and individual power lines extraction experiments are conducted in different corridors collected in suburban areas. Promising results are obtained for both semantic segmentation and individual power lines extraction, with a mIou of point cloud semantic segmentation and a mean extraction rate of 96.70% and 96.56%, respectively.

Accurate and dependable weed detection technology is a prerequisite for weed control robots to do autonomous weeding. Due to the complexity of the farmland environment and the resemblance between crops and weeds, detecting weeds in the field under natural settings is a difficult task. Existing deep learning-based weed detection approaches often suffer from issues such as monotonous detection scene, lack of picture samples and location information for detected items, low detection accuracy, etc. as compared to conventional weed detection methods. To address these issues, WeedNet-R, a vision-based network for weed identification and localization in sugar beet fields, is proposed. WeedNet-R adds numerous context modules to RetinaNet’s neck in order to combine context information from many feature maps and so expand the effective receptive fields of the entire network. During model training, meantime, a learning rate adjustment method combining an untuned exponential warmup schedule and cosine annealing technique is implemented. As a result, the suggested method for weed detection is more accurate without requiring a considerable increase in model parameters. The WeedNet-R was trained and assessed using the OD-SugarBeets dataset, which is enhanced by manually adding the bounding box labels based on the publicly available agricultural dataset, i.e. SugarBeet2016. Compared to the original RetinaNet, the mAP of the proposed WeedNet-R increased in the weed detection job in sugar beet fields by 4.65% to 92.30%. WeedNet-R’s average precision for weed and sugar beet is 85.70% and 98.89%, respectively. WeedNet-R outperforms other sophisticated object detection algorithms in terms of detection accuracy while matching other single-stage detectors in terms of detection speed.

Unmanned surface vehicle (USV)-based floating-waste detection presents significant challenges. Due to the water surface’s high reflectivity, there are often light spots and reflections in images captured by USVs. Furthermore, floating waste often consists of numerous small objects that prove difficult to detect, posing a robustness challenge for object-detection networks. To address these issues, we introduce a new dataset collected by USV, FloatingWaste-I, which accounts for the effects of light in various weather conditions, including sunny, cloudy, rainy and nighttime scenarios. This dataset comprises two types of waste: bottles and cartons. We also propose the innovative floating-waste-detection network, YOLO-Float, which incorporates a low-level representation-enhancement module and an attentional-fusion module. The former boosts the network’s low-level representation capability while the latter fuses the highest- and lowest-resolution feature map to improve the model robustness. We evaluated our method by using both the public dataset FloW-img and our FloatingWaste-I dataset. The results confirm YOLO-Float’s effectiveness, with an AP of 44.2% on the FloW-img dataset, surpassing the existing YOLOR, YOLOX and YOLOv7 by 3.2%, 2.7% and 3.4%, respectively.