Abstract
With the continuous development of energy infrastructure, a large number of distributed new energy, distributed energy storage, and various power electronic equipment are connected to distribution communities. The access to a large amount of equipment not only increases the workload of power operators, but also leads to a complex field operation environment, which will threaten the security of field operators. Traditional monitoring strategy for distribution community operation site relies on manual operation, which is tedious, labor-intensive, and error-prone. To solve this problem, this paper proposes a security monitoring strategy for distribution community operation sites based on an intelligent image processing method. Firstly, a power image enhancement method based on multi-filtering algorithm is proposed. The filter operators are used to smooth and denoise the image, enhance the edge and detail information, and expand the security monitoring data set of operation sites by merging the filtered images. Secondly, an object detection method for personnel and security protection tools based on the improved Faster R-CNN algorithm is proposed, and a pyramid structure is constructed to improve the detection and localization accuracy of small objects such as safety helmets and safety belts. To simplify the object detection task and the complexity of the labeling system, it is disassembled into four single tasks: operators, non-operators, safety helmets, and safety belts. Finally, the detection frame labels are fused by calculating the overlap of the object detection boxes to describe the security of the operator. Experiments show that the method proposed in this paper can effectively locate the object and accurately describe the security of the operator.
Introduction
The modernization of power distribution networks has introduced a diverse array of technologies, including distributed renewable energy sources, battery storage systems, and electric vehicle (EV) charging infrastructure. While these advancements are critical for the energy transition, they significantly complicate the operational environment at distribution sites. Power operators now face increased workloads and more hazardous conditions, necessitating rigorous adherence to safety protocols.
Historically, site safety has been monitored manually. This approach is not only labor-intensive but also susceptible to human oversight, particularly in cluttered or high-activity environments. While computer vision and deep learning offer potential for automation, their application in this domain is hindered by two primary factors: the scarcity of specialized, high-quality labeled datasets for power operations, and the inherent difficulty of detecting small safety items—such as helmet straps or safety belt buckles—using standard object detection architectures.
Methodology
This paper presents an integrated security monitoring framework designed to overcome these hurdles through a three-stage process: data augmentation via multi-filtering, an improved detection architecture, and a task-disassembly approach for safety assessment.
Multi-Filter Image Enhancement and Data Augmentation
To address the data scarcity problem, the researchers developed a "multi-filter" augmentation strategy. Instead of relying solely on manual labeling, which is time-consuming, the method applies four distinct filtering algorithms to a base set of 600 original images: Gaussian, Laplace, Laplacian of Gaussian (LoG), and Unsharp Masking.
These filters serve to denoise images or sharpen edge details without altering the spatial coordinates of the objects within the frame. This allows for "label reuse," where the original bounding box annotations remain valid for the filtered versions, effectively quadrupling the training data available for the deep learning model.
Improved Faster R-CNN with Feature Pyramids
The core of the detection system is an improved version of the Faster R-CNN model. Standard detection models often struggle with small objects because high-level feature maps lose fine-grained spatial information. To mitigate this, the authors incorporated a Feature Pyramid Network (FPN) structure. This architecture fuses high-level semantic information with low-level detail, significantly enhancing the model's ability to localize and identify small safety equipment like helmets and belts in complex backgrounds.
Task Disassembly and Label Fusion
To reduce the complexity of the labeling system and the learning burden on the model, the detection task was disassembled into four simpler sub-tasks: identifying operators, non-operators, safety helmets, and safety belts. Rather than training a single complex classifier for "safe operator" vs. "unsafe operator," the model learns to identify these components independently. A final "Label Fusion" step uses bounding box overlap (Intersection over Union) to determine if a detected operator is correctly wearing the required safety gear, resulting in a comprehensive security report.
Results and Discussion
The proposed framework was evaluated against a test set of 200 images. The results demonstrate that the improved Faster R-CNN significantly outperforms industry-standard models like SSD and YOLO.
- Detection Accuracy: The model achieved an Average Precision (AP) of 0.9051 for safety helmets and 0.8674 for safety belts. In comparison, the YOLO model achieved only 0.76 and 0.80 respectively.
- Impact of Augmentation: The multi-filter approach proved highly effective. Specifically, Gaussian filtering was found to be the most beneficial for safety belt detection, increasing the AP by 6.3% compared to training on unfiltered data.
- System Performance: The label fusion method successfully combined individual detections to reach a final safety identification accuracy of 96%, up from a 91% baseline.
Conclusion
The intelligent image processing strategy proposed in this paper provides a robust solution for the automated security monitoring of power distribution sites. By leveraging multi-filter data augmentation and an FPN-enhanced Faster R-CNN, the system achieves high accuracy in detecting small safety equipment and assessing operator compliance. Future research will focus on improving real-time performance and testing the system's resilience in extreme weather and low-light conditions.
Summary updated