2021, 10(3):1-11. DOI: 10.12146/j.issn.2095-3135.20201020001
Abstract:This paper presents an improved faster R-CNN algorithm based on the application of unmanned vending machine selling bottled drinks. Firstly, the residual network ResNet-50 is used as the feature extraction network to deepen the depth of target feature extraction and learning. Then, the number and style of anchor frame in regional proposal network (RPN) is improved according to the morphological characteristics of bottled beverage products. Finally, a multi-dimensional feature map fusion network is proposed to enhance the detection performance of small targets. The experimental results showed that, the loss value tends to converge after 10 000 iterations of model training. Average precision values of 10 categories of bottled beverage products are all larger than 90%. And the comprehensive detection recognition rate mean average precision value is 93.26%, which is improved 20% compared with the original model.
2021, 10(3):12-21. DOI: 10.12146/j.issn.2095-3135.20201030001
Abstract:In this paper, a triangular centroid distances (TCDs) descriptor that integrates corner information is proposed to solve the detection problem in the case of overlapping workpieces. The proposed descriptor detects the corner points and local contour direction of the target, which can be used to detect the suspected contour segments on the template contour. Then, the improved TCDs feature matrices are extracted from the target contour and the suspected contour segments of the templates. Finally, by calculating the distances between target contour matrix and each suspected contour feature matrix, the suspicious contour segment with the smallest distance can be determined as the corresponding workpiece. The experimental results showed that, not only recognition rate but also the efficiency of the proposed algorithm can be improved compared with traditional TCDs algorithm.
2021, 10(3):22-34. DOI: 10.12146/j.issn.2095-3135.20201215001
Abstract:With the prevalence of mobile devices and the gradually severe information overloading in big data era, how to accurately recommend apps to users based on their interests and their user behaviours becomes an important problem. Most existing recommender systems suffer from common weaknesses such as making recommendations that are too dull and lacking diversity, and they also rarely consider the user behaviours when making recommendations, resulting in sub-optimal recommendation performance. In this paper, we propose a multi-dimensional app recommender system that takes both the hierarchically clustered user features and user behaviours into account when making recommendations. After being processed by singular value decomposition (SVD) that performs denoising and dimensionality reduction, the user features are then clustered into several user groups by the hierarchical clustering. The user behaviour information is subsequently leveraged together with the user features and inputted into the Bayesian model to produce an app recommendation list in the descending order of the probability of being accurate. Hence, the accuracy of the app recommendation is improved. This paper finally implements the proposed recommender system under a distributed framework, which ensures the efficiency and the reliability of the system. The experiments demonstrate an average within-group deviation of 0.4 after the SVD is applied. The recall of the app recommender system after the hierarchical clustering raises to 73%, which shows a 5%～16% performance boost compared with collaborative-filtering-based and association-rule-based methods. All evaluation metrics of the Bayesian model also presents a performance gain of around 10%, which verifies the effectiveness of the proposed recommender system.
2021, 10(3):35-46. DOI: 10.12146/j.issn.2095-3135.20201231001
Abstract:The thermal management of proton exchange membrane fuel cells (PEMFCs) influences the safety, durability, and operating efficiency of hydrogen fuel cell vehicles. A thermal management control method is proposed for PEMFCs in this research to maintain the temperature at the inlet and outlet of the stack at the set values. A two-dimensional fuzzy controller is designed based on the temperature changes at the inlet and outlet of the stack in the thermal management system model of PEMFCs, where the membership function of the fuzzy controller is optimized by using the genetic algorithm, so that the control precision of the fuzzy controller is higher. A hydrogen fuel cell vehicle from the Autonomie software is used to validate the proposed thermal management method of the PEMFC on two standard vehicle driving conditions. Simulation results show that the fuzzy controller optimized by the genetic algorithm presents the higher control accuracy than the one without the optimization.
2021, 10(3):47-60. DOI: 10.12146/j.issn.2095-3135.20210118001
Abstract:In order to improve the fuel economy and fuel cell lifetime of fuel cell hybrid vehicles, this research proposes an energy management strategy based on deep reinforcement learning (DRL). The strategy first adds a lifetime factor to reward signal of DRL, the lifetime of fuel cell is extended by limiting the power fluctuation. Then, the fuel cell system works in a high efficiency range by limiting the action space of DRL, improving the efficiency of the entire vehicle. After offline training under UDDS, WLTC, and Japan1015, it is applied in real time under NEDC to verify the adaptability of the proposed strategy. The results show that the proposed strategy can converge quickly in offline training, which proves its stability. Compared with dynamic programming-based strategy, the fuel economy difference in training cycles is only 5.58%, 3.03% and 4.65%, which is close to the optimal, and the promotion is 4.46%, 7.26% and 5.35% compared with reinforcement learning-based strategy. Compared with the DRL-based strategy without a lifetime factor, the proposed strategy reduces the average power fluctuation by 10.27%, 47.95%, and 10.85% under training cycles, which is beneficial to improve the fuel cell lifetime. In the real-time application, the fuel economy of the proposed strategy is improved by 3.39% compared with the reinforcement learningbased strategy, which proves its adaptability to unknown cycles.
2021, 10(3):63-68. DOI: 10.12146/j.issn.2095-3135.20210227001
Abstract:With the development of science and technology, orthopedic biomaterials have evolved from traditional inert materials into targeted functional regulating materials. During their degradation, the materials can also regulate the local bone metabolism and promote bone regeneration and functional reconstruction. Therefore, the development of intelligent materials will be the hot spot and the new direction of orthopedic biomaterials research in the next decade, while multidisciplinary integration is an inevitable trend for the orthopedic clinics.
2021, 10(3):69-77. DOI: 10.12146/j.issn.2095-3135.20210310001
Abstract:China is in the leading position of the world in research of antibacterial medical metals, and their applications are expected to effectively reduce the incidences of bacterial infections related to medical devices or implants, which has great clinical values. This article briefly introduces the innovative researches and preliminary applications of Cu-bearing antibacterial medical metals in China, and analyzes the opportunities and challenges in the future.
2021, 10(3):78-92. DOI: 10.12146/j.issn.2095-3135.20210310002
Abstract:Precision medicine is an important development direction of modern biomedicine, and the emergence of micro/nanorobots has promoted the development of precision medicine. These miniaturized robots are manufactured through self-assembly, electron beam deposition, and 3D printing, and they can initiate chemical reactions or move under the action of external fields such as ultrasound, light, magnetic and microorganisms (cells). They are widely used in biomedicine, they can load drug particles, biological reagents and living cells to achieve precise cargo delivery; perform operation and treat diseases as a small sized surgical tool; detect metal ions and other substances in organisms for early diagnosis of diseases; conduct medical imaging through different methods such as photoacoustic, magnetic resonance. In the past ten years, the research of micro/nano robots has made some progress in these areas, vigorously promoting the development of modern medicine.
2021, 10(3):93-99. DOI: 10.12146/j.issn.2095-3135.20210331001
Abstract:3D printed bone tissue model is important to the pre-operation planning, precision intraoperative location and post-surgery assessment. However, to print the same scale bone tissue model of patients depends on a series of processing from the CT image scanning, 3D reconstruction, 3D printing to post processes of the printed model. In the process, the quality of CT images and 3D reconstruction directly affects the precision and quality of the 3D printed model, especially the selection of which CT image series is a key problem that needs to be answered. Through the comparison of 3D reconstruction results between the bone series and stand series of CT images shown that the standard series produce better quality than the bone series. The conclusion is that the standard series may produce reconstructed 3D model of bone tissue in a better way than the bone series. And the conclusion provides a scientific proof for the selection of CT image series during the 3D reconstruction.