Synthetic aperture radar (SAR) target recognition faces the challenge that there are very little labeled data. Although few-shot learning methods are developed to extract more information from a small amount of labeled data to avoid overfitting problems, recent few-shot or limited-data SAR target recognition algorithms overlook the unique SAR imaging mechanism. Domain knowledge-powered two-stream deep network (DKTS-N) is proposed in this study, which incorporates SAR domain knowledge related to the azimuth angle, the amplitude, and the phase data of vehicles, making it a pioneering work in few-shot SAR vehicle recognition. The two-stream deep network, extracting the features of the entire image and image patches, is proposed for more effective use of the SAR domain knowledge. To measure the structural information distance between the global and local features of vehicles, the deep Earth mover's distance is improved to cope with the features from a two-stream deep network. Considering the sensitivity of the azimuth angle in SAR vehicle recognition, the nearest neighbor classifier replaces the structured fully connected layer for K -shot classification. All experiments are conducted under the configuration that the SARSIM and the Moving and Stationary Target Acquisition and Recognition (MSTAR) dataset work as a source and target task, respectively. Our proposed DKTS-N achieved 49.26% and 96.15% under ten-way one-shot and ten-way 25-shot, whose labeled samples are randomly selected from the training set. In standard operating condition (SOC) as well as three extended operating conditions (EOCs), DKTS-N demonstrated overwhelming advantages in accuracy and time consumption compared with other few-shot learning methods in K -shot recognition tasks.

Most deep-learning based target detection methods have high computational complexity and memory consumption, and they are difficult to be deployed on edge devices with limited computing resources and memory. To tackle this problem, this paper proposes to learn a lightweight detector named Light-YOLOv4, and it is obtained from YOLOv4 through model compression. To this end, firstly, we perform sparsity training by applying L1 regularization to the channel scaling factors, so the less important channels and layers can be found. Secondly, channel pruning and layer pruning are enforced on the network to prune the less important parts, which could significantly reduce network's width and depth. Thirdly, the pruned model is retrained with knowledge distillation method to improve the detection accuracy. Fourthly, the model is quantized from FP32 to FP16, and it could further accelerate the model with almost no loss of detection accuracy. Finally, to evaluate Light-YOLOv4's performance on edge devices, Light-YOLOv4 is deployed on NVIDIA Jetson TX2. Experiments on SAR ship detection dataset (SSDD) demonstrate that the model size, parameter size and FLOPs of Light-YOLOv4 have been reduce by 98.63%, 98.66% and 91.30% compared with YOLOv4, and the detection speed has been increased to 4.2 times. While the detection accuracy of Light-YOLOv4 is only slightly reduced, for example, the mAP has only reduced by 0.013. Besides, experiments on Gaofen Airplane dataset also prove the feasibility of Light-YOLOv4. Moreover, compared with other state-of-the-art methods, such as SSD and FPN, Light-YOLOv4 is more suitable for edge devices.

No abstract

： The customer requirement modeling is a highly important part of a product development process for understanding the customers and market needs. But with the increasing of complexity of complex mechatronic products, it is necessary to involve multidisciplinary design teams, thus, the traditional customer requirements modeling for a single discipline team becomes difficult to be applied in a multidisciplinary team and project since team members with various disciplinary backgrounds may have different interpretations of the customers' requirements. This paper provides a new synthesized multidisciplinary customer requirements modeling method for obtaining and describing the common understanding of customer requirements and more importantly transferring them into a detailed and accurate product design specifications (PDS) to interact with different team members effectively. A case study of designing a high speed train verifies the rationality and feasibility of the proposed multidisciplinary requirement modeling method for complex mechatronic product development.Key words: complex mechatronic product, multidisciplinary customer requirements modeling, disciplinary specialty language dictionary, high-speed train 1.IntroductionWith the mass customization of product design and development, the individuality and diversity of customer requirements for a complex mechatronic product is ever increasing, the manufacturer enterprises are facing tremendous pressures and challenges to deal with the diverse and rapid changing customer and market requirements. The traditional "product-centric" design method for complex mechatronic products has been unable to adapt to the growing market competition, therefore, the enterprises should shift the design focus to the "customer-centric" design methods for complex mechatronic products. Customer requirement modeling thus becomes a highly significant part of a product development process and it also has been a research topic for years and used in the field of system and software development [1].The ultimate purpose of traditional customer requirements modeling is that realizes the mapping of CRs in the customer domain to PDS (a formalized specification of customers' requirements and a list of the product performance, environment, quality, reliability, security, life cycle and other elements with considering performance and cost constraints, design inputs, constraints and goals and so on [30]) in the designer domain to improve the development efficiency and reduce the development cost. Through many years of research and application, the process of traditional customer requirements modeling, encompassing requirement elicitation, requirement analysis and requirement verification, has been formed into a standardization process from a system and software engineering point of view [2]. The requirement elicitation is to extract Biographical notesMA Xiaojie, born in 1986, is currently a PhD candidate at Institute of Advanced Design and Manufacturing,

Ball-end milling cutter with tooth offset center is widely used in machining industry, because it has higher machining efficiency and better stability compared with the ball-end milling cutter without tooth offset center. In addition, the tooth offset center has lower wear rate of the tool tip so the life of the milling cutter is improved. However, up to present, there is no mature and effective theory for the design and manufacture of this kind of milling cutters. This article presents a new mathematical model for S-shaped edge curve of the ball end taking the tooth offset center into account, which can construct accurate S-shaped edge curve for the ball-end cutting tools with tooth offset center as well as without tooth offset center. This model overcomes the complex computation and bad adaptability of the traditional modeling method. At the same time, a five-axis grinding algorithm for rake face of the ball end is also presented in this article. Finally, based on the application programming interface of CATIA™, a three-dimensional computer-aided design and computer-aided manufacturing system is developed. The accuracy and effectiveness of the grinding algorithm are verified by simulation in VERICUT™ and machining experiment in tool grinding machine.