The workers of many fields must follow strict safety rules, and not complying with them may put these workers into severe danger. Oil and Gas workers are subject to hazards and accidents in their workplace frequently. The use of personal protective equipment (PPE) is of summary importance to professionals who work with heavy-duty machinery or in unsafe environments, reducing the risk of serious injuries or even death. The workers are supposed to wear PPE, some of which are gloves, hardhat, and Steel-Toed Boots, during their activities. However, PPE is commonly misused or unused, incurring the need for recurrent inspection. There is no guarantee that the safety equipment is being used correctly, safely and continuously. These threats to the safety of the workers are increased significantly when they are working offshore due to either the harsh conditions they might be working on or inherent dangers that O&G workplaces can offer to the workers such as machinery and dangerous areas with risk of collision and accidents. Detecting the lack of PPE can prevent injuries during work. For this purpose, we propose a surveillance system solution to automatically analyze video footage and detect oil and gas (O&G) workers who are not using adequate protective equipment. This project developed a multi-step detection system using Deep Learning techniques in a pipeline for monitoring workers through camera images. Being able to detect violations to the established rules is an important step towards reducing the impact of incidents and accidents. Using computer vision, deep neural networks, and video footage, we created a web solution for analyzing the imagery in real-time and issuing alerts when a violation happens. For this specific domain, we accomplished the best results by using YOLOv3 as a person detector in conjunction with the Xception network for classification. We achieved 98% precision for the classification step and 78% precision for the joint solution (detection and classification steps) while running in real-time in an NVIDIA Titan X GPU.

The most common methods to detect non-technical losses involve Deep Learning-based classifiers and samples of consumption remotely collected several times a day through Smart Meters (SMs) and Advanced Metering Infrastructure (AMI). This approach requires a huge amount of data, and training is computationally expensive. However, most energy meters in emerging countries such as Brazil are technologically limited. These devices can measure only the accumulated energy consumption monthly. This work focuses on detecting energy theft in scenarios without AMI and SM. We propose a strategy called HyMOTree intended for the hyperparameter tuning of tree-based algorithms using different multiobjective optimization strategies. Our main contributions are associating different multiobjective optimization strategies to improve the classifier performance and analyzing the model’s performance given different probability cutoff operations. HyMOTree combines NSGA-II and GDE-3 with Decision Tree, Random Forest, and XGboost. A dataset provided by a Brazilian power distribution company CPFL ENERGIA™ was used, and the SMOTE technique was applied to balance the data. The results show that HyMOTree performed better than the random search method, and then, the combination between Random Forest and NSGA-II achieved 0.95 and 0.93 for Precision and F1-Score, respectively. Field studies showed that inspections guided by HyMOTree achieved an accuracy of 76%.