Coordinate classification is an efficient approach to 2-D human pose estimation (HPE), treating keypoint predictions as sub-pixel bins along horizontal and vertical axes, thereby avoiding the computationally intensive upsampling process required in traditional heatmap-based methods. In this article, we introduce the Position Encoding Coordinate Classification (PECC) system, which enhances coordinate classification by embedding position information directly into keypoint feature representations through a novel position encoding mechanism. We further design a tailored attention mechanism, Filtering Amplified Attention (FAA), optimized for coordinate classification. FAA provides finer relative positional information, improving the system’s ability to model relationships between keypoints and enhancing coordinate localization accuracy. Our method maintains the efficiency of coordinate classification by utilizing 1-D vectors, significantly reducing model parameters and computational cost. Additionally, the incorporation of positional encoding enhances the system’s ability to effectively model and exploit spatial information within a coordinate-classification-based pose estimation framework. Extensive experiments on mainstream datasets demonstrate that PECC achieves superior accuracy and robustness in 2-D HPE, advancing the state-of-the-art in this domain.

Rapid advancement of industrial internet of things (IoT) technology has changed the supply chain network to an open system to meet the high demand for individualized products and provide better customer experiences. However the open-system supply chain has forced many small and midsize enterprises (SMEs) to adopt vertical integration by being divided into smaller companies with a distinctive business for each SME but a central alliance to produce a range of products and gain competencies. Therefore, existing models do not guarantee the protection of data privacy of individual SMEs. Moreover, especially for the IoT environment, collecting data in a secure way and revealing valuable knowledge in an IoT network is difficult. How to share data in a secure framework is of paramount importance in the internet of behavior field. In this article, a privacy-preserving data-mining framework is proposed for joint-venture industrial collaborative activities by combining federated learning and a “pre-large concept” of data-mining techniques. The novelty of the proposed approach is that, while mining high-utility itemsets (HUIs) from multiple datasets, it does not require direct data sharing. In the proposed method, the federated-learning framework can learn from aggregated learning parameters without scanning all data from different sets. The pre-large concept in this approach reduces the amount of scanning into different datasets. Thus, the approach makes it possible to train federated learning more quickly while protecting the privacy of individual data owners. The approach has been tested on real industrial datasets in a collaborative environment. Extensive experimental results show that the approach achieves high accuracy compared with conventional data-mining techniques while preserving the privacy of datasets.

Aiming at the problem of real-time and low accuracy of automatic recognition of human abnormal behaviour in a public area surveillance video, a simple multi-scale human anomaly behaviour detection algorithm based on video was proposed. Firstly, the binary image sequence of human body in surveillance video is acquired by background modelling method based on visual background extraction (ViBe). Then, the simple multi-scale algorithm is constructed by combining the aspect ratio, motion trajectory and video continuous interframe motion acceleration of the minimum circumscribed rectangle of the binarised image. The human target behaviour is judged, and then the normal behaviour of the human body - standing, walking, jogging, and abnormal behaviour - shouting for help, falling, punching, wandering, and sudden running are identified. The experimental results show that the human body moving target recognition by ViBe combined with simple multi-scale algorithm for abnormal behaviour detection has good real-time performance and high accuracy.

With the development of science and technology, analysis of large data is used more and more widely, such as electronic commerce, biotechnology, retail sale business, finance, education are inseparable from our life. Big data can show its value with data mining, moreover, it plays an important role in the field of education. Most researches on the exploration of education data have implications in predicting students' learning effectiveness, which can predict students' mid-term grades, final grades and semester grades with the big data analysis. The exploration of education data has been used by many research to apply the existing predictive model, thus can predict students who failed the semester exam. When there are some students who have a tendency, teachers can early to care about their learning conditions, find out the reason why they are poor in the performance, which can improve the quality of student learning. This study extends JAVA automatic grading system that our laboratory developed, increases students' operation behaviors and forecasting analysis models on the system, cooperates with four algorithms like random forest, supporting vector machine, neural network, Naive Bayes to predict and analyze the student's drop point of semester grades. This study can predict semester grade with 77% accuracy. Furthermore, this study also makes behavior analysis based on the collected data to find out the behavior model for students with low grades. The result of the study analysis can help teachers to early tutor students and improve their quality of teaching. Moreover, it plays an important role in reference value.

MapReduce paradigm has become a popular platform for massive data processing and Big Data applications. Although MapReduce was initially designed for high throughput and batch processing, it has also been used for handling many other types of applications and workloads due to its scalable and reliable system architecture. One of the emerging requirements for enterprise data-process computing is completion time guar- antee. However, there are only a few research works have been done for MapReduce jobs with deadline constraint. Therefore, in this paper, we aim to prevent jobs from missing deadline while maximizing the resource utilization and data locality of a MapReduce cluster. Our approach is to introduce a two-phase job scheduling mechanism which combines a job admission controller policy and a priority-based scheduling algorithm. We use a series of simulations over diverted workload to evaluate our system. The results show that our approach can guarantee job completion time in a heavy-loaded system, and achieve comparable data locality to the delay schedule algorithm in a light-loaded system. Furthermore, our approach can maximize system throughput by preventing system resources from being wasted by the jobs missing their deadlines.

The equipment lock has been an important tool for the power company to protect the electricity metering equipment. However, the conventional equipment lock has two potential problems: vandalism and counterfeiting. To fulfill the control and track the potential illegal behavior, the human labor and paper are required to proceed with related operations, resulting in the consumption of a large amount of human resources and maintenance costs. This study focused on the design of RFID technology applied to the traditional equipment lock, which, through the mobile and electronic technology, strengthens the management/operating convenience of the lock and provides the solutions for anti-counterfeiting and spoilage detection so that the national energy can be properly protected and fairly distributed.