As an interdisciplinary course, Machine Vision combines AI and digital image processing methods. This paper develops a comprehensive machine vision experiment on forest wildfire detection that organically integrates digital image processing, machine learning and deep learning technologies. Although the research on wildfire detection has made great progress, many experiments are not suitable for students to operate. Also, the detection with high accuracy is still a big challenge. In this paper, we divide the task of forest wildfire detection into two modules, which are wildfire image classification and wildfire region detection. We propose a novel wildfire image classification algorithm based on Reduce-VGGnet, and a wildfire region detection algorithm based on the optimized CNN with the combination of spatial and temporal features. The experimental results show that the proposed Reduce-VGGNet model can reach 91.20% in accuracy, and the optimized CNN model with the combination of spatial and temporal features can reach 97.35% in accuracy. Our framework is a novel way to combine research and teaching. It can achieve good detection performance and can be used as a comprehensive experiment for Machine Vision course, which can provide the support for talent cultivation in machine vision area.

With the continuous improvement of the national medical system, health monitoring combined with cloud computing and Internet of Things has become a concern. This study mainly discusses the design of the medical health monitoring IoT system based on cloud computing. From the user to the health service provider, there are three devices: sensor terminal, gateway terminal, and service platform. The sensor terminal is used to measure physiological indicators, such as blood pressure, electrocardiogram, blood oxygen saturation, heart rate, and other physiological indicators; the gateway terminal is used to link the sensor terminal to receive physiological indicators and forward them to the business platform; the gateway is also used to receive health information and other instructions issued by the server. In the community service mode, users can be divided into groups according to the community and region, and the corresponding service doctors and agent customer service personnel (nurses) can be assigned. Users can collect personal physiological indicators at home or outside through the medical terminal. These indicators and information are transmitted to the background health platform system through the mobile GSM-TD communication network. Users can also view their own historical health records and opinions of health consultants through the web/WAP website. Through the integration ability of the health cloud platform, relying on the interconnection with HIS, LIS, and other information systems of professional medical institutions, we jointly operate special value-added services, such as appointment registration, maternal and child healthcare, and medical communication (doctor-patient interaction), so that users can enjoy the remote service and guidance of professional medical institutions by subscribing to health value-added services. The CPU utilization rate is 40%, the internal utilization rate is 7.44 G, the memory utilization rate is 11.8%, and the network bandwidth is 591.87 M. During the whole test process, the indicators are stable, and there are no restart, crash, and other phenomena, so the system performance meets the design requirements.

Non-intrusive load identification can infer the operating status of various electrical appliances in a user’s household based on the electrical parameters collected by the overall household electricity meter. It helps users utilize electrical energy more scientifically and rationally. Due to significant differences in the working principles of appliances and their energy consumption, the effectiveness of load identification has not yet met practical application requirements. This paper proposes a load identification method based on a dual-branch attention GRU fusion network structure, the network integrates GRU, Temporal Convolution Network (TCN), and attention mechanisms. It overcomes the limitation of GRU networks in extracting ultra-long-term temporal dependencies and compensates for TCN’s less effective exploration of medium and short-term temporal dependencies compared to GRU. As a result, it achieves better identification performance. Experiments on public datasets indicate that the proposed method achieves better load state identification performance compared to other methods.