In the rapidly developing e-commerce and innovative logistics sectors, the accuracy and efficiency of package dimension measurement technology are crucial for accurate shipping cost calculation. This study proposes a novel, low-cost 3D dimension measurement system based on deep learning and a dual-camera stereo vision compensation model to improve the reliability of measurement software in computer vision computing. This system uses standard camera equipment and integrates the YOLO algorithm to establish image-based object size estimation. The design concept mimics the human eye's binocular vision mechanism. It utilizes the parallax between the center points of two cameras to estimate distance, thereby optimizing computational performance and recognition speed. A mathematical parallax model is applied to improve depth accuracy. To mitigate computational errors and enhance reliability, a two stage linear regression method combined with a weighted error compensation model is used to correct for systematic deviations between depth estimates and ground truth measurements. Experimental results show that this framework achieves an average error of less than 1 cm and an overall accuracy of 89% for 3D dimension estimation of various logistics packages under uncontrolled conditions, meeting the accuracy requirements for practical logistics billing. These results demonstrate that leveraging flexible software-based stereo vision computing can significantly improve the feasibility and reliability of automated measurement solutions in real-world logistics applications.

Deep learning is developing rapidly, and the emergence of many network architectures has brought significant breakthroughs to training recognition models. Due to the maturity of edge computing technology, we can perform regional image training through distributed nodes, which significantly improves the training model's accuracy while performing transfer learning to achieve better performance. In image processing technology, high-precision recognition of non-luminous images can currently be achieved by modeling, if we replace the visual recognition target with a glowing digital panel, the recognition rate cannot be the same as the static text recognition rate. This article uses Keras to build a convolutional neural networks deep learning model to identify glowing light-emitting diodes (LED) digits, incremental learning to complete transfer learning on edge computing nodes, and an integrated IoT architecture to achieve better recognition results. In the experiment, the verification results obtained from the distributed training nodes were successfully combined to model and retrain the nodes. The proposed distributed learning method can increase the accuracy from 70% to 89%. At the same time, the misclassified images can be retrained by integrating the transfer learning model with the distributed learning results, and the accuracy reaches more than 92%.

The demand for automated dimensional measurement technology is growing daily with global supply chain management challenges. Accurate dimensional measurement is critical to optimizing packaging design, reducing shipping costs, improving warehouse efficiency, and enhancing the inventory management process. Traditional manual measurement methods are time-consuming and error-prone due to measurement and data collection errors. There are hidden concerns about the accuracy of the data, and they cannot meet the fast-paced, high-efficiency demands of modern logistics. To address this problem, this study introduces an automatic three-dimensional object size measurement system based on dual-view imaging technology. The experimental environment uses top-angle and side-angle lenses, combined with the image processing technology of the OpenCV framework, to accurately measure the object's outermost three-dimensional dimensions. The image processing algorithm used in the research can automatically identify the outline of an object and calculate its length, width, and height. The development of the system includes precise camera calibration, simultaneous image acquisition, and image processing technology to ensure the accuracy and consistency of measurement results. The research demonstrates the experimental verification of accurately using images to calculate the contour feature information of objects. It conducts measurements based on various samples, proving that it can measure the outer dimensions of objects of various materials and shapes and has strong practicability and broad application prospects. The contribution of the research not only provides new solutions for dimensional measurement in the logistics field but also the experimental empirical model will expand the development of logistics warehouse automation systems with innovative and forward-looking data calculation and application.

In IoT communication, to facilitate the management of the status of all devices and ensure availability under limited network performance, the MQTT communication protocol is used to construct a data exchange mechanism. However, MQTT's transmission mechanism needs built-in encryption, effectively displaying the transmission content in clear text to everyone, raising concerns about information security. At the same time, MQTT's subscription mechanism helps to manage many devices quickly, but this mechanism lacks a set of management protocols to limit the devices that can be subscribed. In the case where any node can subscribe, this study proposes a framework that uses the device's IP and port combined with a whitelist mechanism to distinguish legitimate and illegal subscriptions. After the device requests a subscription, it will be recorded in the Broker, and the message will be broadcast within the topic. However, if the node information is not stored in the whitelist, the message cannot be read because the information transmission process uses asymmetric public key encryption. This situation is called an illegal subscription unless the whitelist is updated and the node is allowed to receive the topic key content, thereby becoming a legitimate subscribing device. In this way, if there are more devices under this topic, you can effectively limit which devices can communicate with these different devices, eliminating concerns about information security. Proposing a new framework to improve the MQTT subscription mechanism not only maintains the convenience of the original protocol but also significantly enhances the protection of the information transmission process. In addition, through experimental content, it is proved that the low latency of the framework can ensure the real-time delivery of messages.

Through big data analytics and deep learning, users can uncover unseen hidden information from personal data in the cloud and derive health improvements from it - including genomic, microbial, medical history, comprehensive blood analysis chemistry, proteins and metabolites, and daily data from exercise devices and scales. The analysis and integration of this data has the potential to improve health and prevent disease. In the research, a set of tools to collect, compute and analyze data through the middle layer, combined with alliance learning and deep learning technologies to conduct sampling comparison training on medical data, while combining edge computing nodes and data hidden features of alliance learning for data construction and model verification. The outcomes can provide early warning of disease agents through data analysis, early symptom detection, and long-term monitoring before the disease becomes widespread. Daily behavior and health can be improved by supporting the P4 medical model, and by integrating these capabilities, many chronic diseases can be prevented from further devastating patients' health. In the experiments, we propose a framework for designing proxies through system construction, data quantification, description and analysis of patient-specific chronic disease risk, and data learning to validate the model construction.

Computer-based visual recognition technology combined with deep learning enables accurate image matching through interactive, multi-level comparison. Currently, popular search methods such as RNN, Faster RNN, etc. are used in computer vision to compute image information by performing hierarchical comparison of image objects. In today’s machine learning is used to construct training curves to predict corresponding results, but the accuracy rate will cause some data distortion due to overformation. Therefore, to solve the problem of interference is now physical, the abandonment method is developed to extract certain neurons and reduce the number of feature values. In this study, a visual image recognition framework is designed that uses advanced computer technology to mark and compare images, and to mark and eliminate blurred images. The experimental method successfully improves the prediction of accuracy after a judgment error by comparing the results of training with the results of deep learning verification. The accuracy of matrix formation and result prediction can reach 0.9812 without eliminating the image produced by the light-emitting elements. We remove noisy images based on the same sampling information. After re-training and re-predicting, the accuracy can reach 0.9847.

In the big data era, the distributed file system is getting more and more significant due to the characteristics of its scale-out capability, high availability, and high performance. Different distributed file systems may have different design goals. For example, some of them are designed to have good performance for small file operations, such as GlusterFS, while some of them are designed for large file operations, such as Hadoop distributed file system. With the divergence of big data applications, a distributed file system may provide good performance for some applications but fails for some other applications, that is, there has no universal distributed file system that can produce good performance for all applications. In this paper, we propose a hybrid file system framework, HybridFS, which can deliver satisfactory performance for all applications. HybridFS is composed of multiple distributed file systems with the integration of advantages of these distributed file systems. In HybridFS, on top of multiple distributed file systems, we have designed a metadata management server to perform three functions: file placement, partial metadata store, and dynamic file migration. The file placement is performed based on a decision tree. The partial metadata store is performed for files whose size is less than a few hundred Bytes to increase throughput. The dynamic file migration is performed to balance the storage usage of distributed file systems without throttling performance. We have implemented HybridFS in java on eight nodes and choose Ceph, HDFS, and GlusterFS as designated distributed file systems. The experimental results show that, in the best case, HybridFS can have up to 30% performance improvement of read/write operations over a single distributed file system. In addition, if the difference of storage usage among multiple distributed file systems is less than 40%, the performance of HybridFS is guaranteed, that is, no performance degradation.