Shenyang Institute of Automation,Chinese Academy Of Sciences
Not a member yet
23582 research outputs found
Sort by
Ocean Circulation in the Challenger Deep Derived From Super-Deep Underwater Glider Observation
No full textOcean circulation in the Challenger Deep is rarely investigated due to sparse observations. Based on hydrographic data collected by two super-deep Chinese "Sea-Wing" gliders during September-October 2018, we analyzed water properties and ocean circulation structures in this trench. Results showed that the horizontal distribution of water properties is approximately a three layer structure, changing from a northeast-southwest structure to a north-south structure then to an east-west structure as the depth increases from 3,000 to 7,000 m. This structure is a joint result of the water uplift, advection, and diffusion in the trench. The westward geostrophic flow, with 1.29 Sv (1 Sv = 10(6) m(3) s(-1)) volume transport, dominates the deep layer of the Challenger Deep, gradually weakening with depth. The unexpectedly large volume transport, twice that of previous studies, might be due to temporal variations of LCDW (Lower Circumpolar Deep Water) intrusion from the Southern Ocean and different reference levels
Special issue on interpretation of deep learning: prediction, representation, quantification and visualization
No full textA multifunctional robotic system toward moveable sensing and energy harvesting
No full textThe operations of the triboelectric nanogenerator (TENG) highly rely on the availability of the mechanical motion source depending on the location, time, weather, etc. Thus, it is highly demanded to achieve moveable mechanical sensing and energy harvesting, but it is also challenging considering the system complexity, size, and weight. To this end, an electrostatic robotic system capable of locomotion, energy harvesting, and vibration sensing is proposed in this paper. The main body of the robot is an integration of an electrostatic actuator and a TENG, both of which share the same structure and materials and utilize the mechanical and electrical characteristics of electrostatic effects, respectively. The prototype is lightweight (2.46 g) and compact (3.7 cm in height and 9.1 cm in length), consisting of two conductive films as the main body in a zipper-like form and two flat conductive films as feet. Here we demonstrate the multifunctionality of this prototype by driving the robot crawling on the ground at a speed of 2.2 mm/s at maximum with a mini camera for monitoring, anchoring by electrostatic adhesive feet at the aim location where vibration is strong, sensing the vibration frequency accurately while having an average relative error of 8.7% in measuring the amplitude, and harvesting the energy by TENG. Such a multifunctional robotic system may enable broad potential applications in structural health monitoring, environmental surveillance, rescue, risky intervention, etc
Research on Stereo Vision Technology Based on Improved Region Growing Method
No full textAbstract
Image segmentation is the cornerstone of image analysis and image processing, its main difficulty is the ill-posedness of image segmentation. The region growing method is the most commonly used method in image segmentation. Its advantages are fast calculation speed, lower algorithm difficulty, and easy understanding. This article uses the area growing algorithm and TOF combined with binocular fusion technology. Discussed how to select the seed points in the region growing method. The algorithm proposed in this paper has high accuracy and high matching quality in the boundary area of the object and the area with large difference in depth. It has both matching time and reliability. It has good results and overcomes the limitations of the active and passive distance methods in the vision system
Path planning of mobile robot based on improved DDQN
No full textAbstract
Aiming at the problem of overestimation and sparse rewards of deep Q network algorithm in mobile robot path planning in reinforcement learning, an improved algorithm HERDDQN is proposed. Through the deep convolutional neural network model, the original RGB image is used as input, and it is trained through an end-to-end method. The improved deep reinforcement learning algorithm and the deep Q network algorithm are simulated in the same two-dimensional environment. The experimental results show that the HERDDQN algorithm solves the problem of overestimation and sparse reward better than the DQN algorithm in terms of success rate and reward convergence speed, Which shows that the improved algorithm finds a better strategy than the DQN algorithm
Structural Dependency Self-attention Based Hierarchical Event Model for Chinese Financial Event Extraction
No full textDocument-level event extraction (DEE) now draws a huge amount of researchers’ attention. Not only the researches on sentence-level event extraction have obtained a great progress, but researchers realize that an event is usually described by multiple sentences in a document especially for fields such as finance, medicine, and judicature. Several document-level event extraction models are proposed to solve this task and obtain improvements on DEE task in recent years. However, we noticed that these models fail to exploit the entity dependency information of trigger and arguments, which ignore the dependency information between arguments, and between the trigger and arguments especially for financial domain. For DEE task, a model needs to extract the event-related entities, i.e., trigger and arguments, and predicts its corresponding roles. Thus, the entity dependency information between trigger and argument, and between arguments are essential. In this work, we define 8 types of structural dependencies and propose a document-level Chinese financial event extraction model called SSA-HEE, which explicitly explores the structure dependency information of candidate entities and improves the model’s ability to identify the relevance of entities. The experimental results show the effectiveness of the proposed model
THz Super-Resolution Imaging Based on Complex Laplacian Prior Deconvolution Algorithm
No full textDue to the long wavelength of the Terahertz (THz) wave, the imaging quality is seriously deteriorated with diffraction. To solve this problem, a simple but very effective approach based on prior knowledge and wave nature was introduced in this paper. In this prior, the image gradients are represented by Laplacian to constrain the gradient of the high-resolution image and the enhanced image when performing single image super-resolution and sharpness enhancement. Moreover, the deconvolution algorithm is expended to a complex dimension. Low-Resolution (LR) THz image was simulated by convolution the High-Resolution (HR) image with real-measured Point-Spread Function (PSF) to ensure the applicability. The numerical experiments illustrate the efficiency and effectiveness of the proposed method in terms of Peak Signal-to-Noise Ratio (PSNR), Mean-Square Error (MSE) and Structural Similarity (SSIM). Super-Resolution (SR) results show that the proposed method has good performance in convergence and suppressing ringing or jaggy artifacts
AI powered electrochemical multi-component detection of insulin and glucose in serum
No full textMulti-component detection of insulin and glucose in serum is of great importance and urgently needed in clinical diagnosis and treatment due to its economy and practicability. However, insulin and glucose can hardly be determined by traditional electrochemical detection methods. Their mixed oxidation currents and rare involvement in the reaction process make it difficult to decouple them. In this study, AI algorithms are introduced to power the electrochemical method to conquer this problem. First, the current curves of insulin, glucose, and their mixed solution are obtained using cyclic voltammetry. Then, seven features of the cyclic voltammetry curve are extracted as characteristic values for detecting the concentrations of insulin and glucose. Finally, after training using machine learning algorithms, insulin and glucose concentrations are decoupled and regressed accurately. The entire detection process only takes three minutes. It can detect insulin at the pmol level and glucose at the mmol level, which meets the basic clinical requirements. The average relative error in predicting insulin concentrations is around 6.515%, and that in predicting glucose concentrations is around 4.36%. To verify the performance and effectiveness of the proposed method, it is used to determine the concentrations of insulin and glucose in fetal bovine serum and real clinical serum samples. The results are satisfactory, demonstrating that the method can meet basic clinical needs. This multi-component testing system delivers acceptable detect limit and accuracy and has the merits of low cost and high efficiency, holding great potential for use in clinical diagnosis.</p
ISTDet: An efficient end-to-end neural network for infrared small target detection
No full textInfrared small target detection has made many breakthroughs in early warning, guidance and battlefield intelligence. However, infrared small target occupies less pixels and lacks color and texture features, which makes infrared small target detection a challenging subject. To achieve the infrared small target detection, an efficient end-to-end network ISTDet is proposed in this paper. ISTDet mainly consists of two modules, including image filtering module and infrared small target detection module. The image filtering module is proposed to obtain the confidence map, aiming to enhance the response of infrared small targets and suppress the response of background. The infrared small target detection module takes the infrared image activated by the confidence map as input, aiming to speculate the category and position of the infrared small targets. Multi-task loss function is used to train the ISTDet in an end-to-end way. Finally, we do comparative experiments on five infrared small target sequences to demonstrate the detection performance of ISTDet. The results show ISTDet has better performance for infrared small target detection compared with other detectors
QoE-Driven Edge Caching in Vehicle Networks Based on Deep Reinforcement Learning
No full textThe Internet of vehicles (IoV) is a large information interaction network that collects information on vehicles, roads and pedestrians. One of the important uses of vehicle networks is to meet the entertainment needs of driving users through communication between vehicles and roadside units (RSUs). Due to the limited storage space of RSUs, determining the content cached in each RSU is a key challenge. With the development of 5G and video editing technology, short video systems have become increasingly popular. Current widely used cache update methods, such as partial file precaching and content popularity- and user interest-based determination, are inefficient for such systems. To solve this problem, this paper proposes a QoE-driven edge caching method for the IoV based on deep reinforcement learning. First, a class-based user interest model is established. Compared with the traditional file popularity- and user interest distribution-based cache update methods, the proposed method is more suitable for systems with a large number of small files. Second, a quality of experience (QoE)-driven RSU cache model is established based on the proposed class-based user interest model. Third, a deep reinforcement learning method is designed to address the QoE-driven RSU cache update issue effectively. The experimental results verify the effectiveness of the proposed algorithm.</p