Shenyang Institute of Automation,Chinese Academy Of Sciences
Not a member yet
23582 research outputs found
Sort by
Infrared and Visible Image Fusion with Improved Residual Dense Generative Adversarial Network
No full text目前基于深度学习的红外与可见光图像融合算法,通常无法感知源图像显著性区域,导致融合结果没有突出红外与可见光图像各自的典型特征,无法达到理想的融合效果.针对上述问题,设计一种适用于红外与可见光图像融合任务的改进残差密集生成对抗网络结构.首先,使用改进残差密集模块作为基础网络分别构建生成器与判别器,并引入基于注意力机制的挤压激励网络来捕获通道维度下的显著特征,充分保留红外图像的热辐射信息和可见光图像的纹理细节信息;其次,使用相对平均判别器,分别衡量融合图像与红外图像、可见光图像之间的相对差异,并根据差异指导生成器保留缺少的源图像信息;最后,通过在TNO等多个图像融合数据集上的实验结果证明,所提方法能够生成目标清晰、细节丰富的融合图像,相比基于残差网络的融合方法，边缘强度和平均梯度分别提升了64.56%和64.94%.</p
Design and Motion Performance Analysis of Turbulent AUV Measuring Platform
No full textThe use of a multi-functional autonomous underwater vehicle (AUV) as a platform for mak-ing turbulence measurements in the ocean is developed. The layout optimization of the turbulence package and platform motion performance are limitation problems in turbulent AUV design. In this study, the computational fluid dynamics (CFD) method has been used to determine the optimized layout position and distance of the shear probe integrated into an AUV. When placed 0.8 D ahead of the AUV nose along the axis, the shear probe is not influenced by flow distortion and can contact the water body first. To analyze the motion of the turbulence AUV, the dynamic model of turbulence AUV for planar flight is obtained. Then, the mathematical equations of speed and angle of attack under steady-state motion have also been obtained. By calculating the hydrodynamic coefficients of the turbulence AUV and given system parameters, the simulation analysis has been conducted. The simulation results demonstrated that the speed of turbulent AUV is 0.5–1 m/s, and the maximum angle of attack is less than 6.5◦, which meets the observation requirements of the shear probe. In addition, turbulence AUV conducted a series of sea-trials in the northern South China Sea to illustrate the validity of the design and measurement. Two continuous profiles (1000 m) with a horizontal distance of 10 km were completed, and numerous high-quality spatiotemporal turbulence data were obtained. These profiles demonstrate the superior flight performance of turbulence AUV. Analysis shows that the measured data are of high quality, with the shear spectra being in very good agreement with the Nasmyth spectrum. Dissipation rates are consistent with background shear. When shear velocity is weak, the measurement of dissipation rate is 10−10 W Kg−1. All indications are that the turbulence AUV is suitable for long-term, contiguous ocean microstructure measurements, which will provide data needed to understand the temporal and spatial variability of the turbulent processes in the oceans.</p
Large full-thickness wounded skin regeneration using 3D-printed elastic scaffold with minimal functional unit of skin
No full textTraditional tissue engineering skin are composed of living cells and natural or synthetic scaffold. Besize the time delay and the risk of contamination involved with cell culture, the lack of autologous cell source and the persistence of allogeneic cells in heterologous grafts have limited its application. This study shows a novel tissue engineering functional skin by carrying minimal functional unit of skin (MFUS) in 3D-printed polylactide-co-caprolactone (PLCL) scaffold and collagen gel (PLCL + Col + MFUS). MFUS is full-layer micro skin harvested from rat autologous tail skin. 3D-printed PLCL elastic scaffold has the similar mechanical properties with rat skin which provides a suitable environment for MFUS growing and enhances the skin wound healing. Four large full-thickness skin defects with 30 mm diameter of each wound are created in rat dorsal skin, and treated either with tissue engineering functional skin (PLCL + Col + MFUS), or with 3D-printed PLCL scaffold and collagen gel (PLCL + Col), or with micro skin islands only (Micro skin), or without treatment (Normal healing). The wound treated with PLCL + Col + MFUS heales much faster than the other three groups as evidenced by the fibroblasts migration from fascia to the gap between the MFUS dermis layer, and functional skin with hair follicles and sebaceous gland has been regenerated. The PLCL + Col treated wound heals faster than normal healing wound, but no skin appendages formed in PLCL + Col-treated wound. The wound treated with micro skin islands heals slower than the wounds treated either with tissue engineering skin (PLCL + Col + MFUS) or with PLCL + Col gel. Our results provide a new strategy to use autologous MFUS instead seed cells as the bio-resource of engineering skin for large full-thickness skin wound healing.</p
Unsupervised learning of depth estimation from imperfect rectified stereo laparoscopic images
No full textBackground: Learning-based methods have achieved remarkable performances on depth estimation. However, the premise of most self-learning and unsupervised learning methods is built on rigorous, geometrically-aligned stereo rectification. The performances of these methods degrade when the rectification is not accurate. Therefore, we explore an approach for unsupervised depth estimation from stereo images that can handle imperfect camera parameters. Methods: We propose an unsupervised deep convolutional network that takes rectified stereo image pairs as input and outputs corresponding dense disparity maps. First, a new vertical correction module is designed for predicting a correction map to compensate for the imperfect geometry alignment. Second, the left and right images, which are reconstructed based on the input image pair and corresponding disparities as well as the vertical correction maps, are regarded as the outputs of the generative term of the generative adversarial network (GAN). Then, the discriminator term of the GAN is used to distinguish the reconstructed images from the original inputs to force the generator to output increasingly realistic images. In addition, a residual mask is introduced to exclude pixels that conflict with the appearance of the original image in the loss calculation. Results: The proposed model is validated on the publicly available Stereo Correspondence and Reconstruction of Endoscopic Data (SCARED) dataset and the average MAE is 3.054 mm. Conclusion: Our model can effectively handle imperfect rectified stereo images for depth estimation
Stiffness regulation analysis of multi⁃stage metamorphic mechanism
No full text刚度调控策略作为实现多级变胞机构的关节柔顺性,提高多级变胞机构的集成度,实现多级变胞机构紧凑化的重要方法,对其进行深入研究有着十分重要的意义.针对多级变胞机构刚度调控复杂的问题,本文通过对弹簧长度、最佳弹簧刚度以及最佳弹簧安装位置的确定,提出了针对多级变胞机构的刚度调控策略.实验结果表明,在一定范围内,最佳弹簧刚度随着安装位置l_3的增大而减小,当最佳弹簧刚度达到最小值后,最佳弹簧刚度随着安装位置l_3的增大而增大,两者之间为非线性关系.</p
Efficient weakly supervised LIBS feature selection method in quantitative analysis of iron ore slurry
No full textOn-stream analysis of the element content in ore slurry plays an important role in the control of the mineral flotation process. Therefore, our laboratory developed a LIBS-based slurry analyzer named LIBSlurry, which can monitor the iron content in slurries in real time. However, achieving high-precision quantitative analysis results of the slurries is challenging. In this paper, a weakly supervised feature selection method named spectral distance variable selection was proposed for the raw spectral data. This method utilizes the prior information that multiple spectra of the same slurry sample have the same reference concentration to assess the important weight of spectral features, and features selected by this prior can avoid over-fitting compared with a traditional wrapper method. The spectral data were collected on-stream of iron ore concentrate slurry samples during the mineral flotation process. The results show that the prediction accuracy is greatly improved compared with the full-spectrum input and other feature selection methods; the root mean square error of the prediction of iron content can be decreased to 0.75%, which helps to realize the successful application of the analyzer.</p
Carbon Black/PDMS Based Flexible Capacitive Tactile Sensor for Multi-Directional Force Sensing
No full textFlexible sensing tends to be widely exploited in the process of human–computer interactions of intelligent robots for its contact compliance and environmental adaptability. A novel flexible capacitive tactile sensor was proposed for multi-directional force sensing, which is based on carbon black/polydimethylsiloxane (PDMS) composite dielectric layer and upper and lower electrodes of carbon nanotubes/polydimethylsiloxane (CNTs/PDMS) composite layer. By changing the ratio of carbon black, the resolution of carbon black/PDMS composite layer increases at 4 wt%, and then decreases, which was explained according to the percolation theory of the conductive particles in the polymer matrix. Mathematical model of force and capacitance variance was established, which can be used to predict the value of the applied force. Then, the prototype with carbon black/PDMS composite dielectric layer was fabricated and characterized. SEM observation was conducted and a ratio was introduced in the composites material design. It was concluded that the resolution of carbon sensor can reach 0.1 N within 50 N in normal direction and 0.2 N in 0–10 N in tangential direction with good stability. Finally, the multi-directional force results were obtained. Compared with the individual directional force results, the output capacitance value of multi-directional force was lower, which indicated the amplitude decrease in capacity change in the normal and tangential direction. This might be caused by the deformation distribution in the normal and tangential direction under multi-directional force.</p
An optimal method based on HOG-SVM for fault detection
No full textIn this paper, an improved method based on HOG-SVM (histogram of oriented gradient characteristic and support vector machine) is proposed for fault diagnosis. First, by converting mechanical vibration signals to 3-D (three dimensional) images, this proposed method can extract the T-HOG (improved HOG) feature of 3-D images precisely. With the optimal method, all characteristic information of mechanical vibration signal, including fault characteristic signal and health characteristic signal, are converted into characteristic 3-D image. Then, fault information can be accurately recognized though R-SVM's (optimal SVM) classification. Furthermore, the new method which is tested on two kinds of field tests, including rail and gear box fault diagnosis, has achieved high detection accuracy of 97.3% and 96.7% respectively. Finally, compared with other ML and signal feature extraction methods, the proposed method shows superiority in fault diagnosis, which is significant for industry safety and reliability.</p
Dynamic mobile robot collaborative computing offloading based on improved deep reinforcement learning
No full text移动边缘计算是解决机器人大计算量任务需求的一种方法；传统算法基于智能算法或凸优化方法，迭代时间长。深度强化学习可以通过一次前向传递即可求解，但只可以针对固定数量机器人进行求解。通过对深度强化学习分析研究，在深度强化学习神经网络中输入层前进行输入规整，在输出层后添加卷积层，使得网络能够自适应满足动态移动机器人数量的卸载需求。最后通过仿真实验验证，与自适应遗传算法和强化学习进行对比，验证了本文算法的有效性及可行性。</p
A multiagent deep deterministic policy gradient-based distributed protection method for distribution network
No full textRelay protection system plays an important role in the safe and stable operation of distribution network (DN), and the traditional model-based relay protection algorithms are difficult to solve the impact of the increasing uncertainty caused by distributed generation (DG) access on the security of DN. To solve this issue, first, the relay protection characteristics of DN under DG access are analyzed; second, the DN relay protection problem is transformed into multiagent reinforcement learning (RL) problem; third, a DN distributed protection method based on multiagent deep deterministic policy gradient (MADDPG) is proposed. The advantage of this method is that there is no need to build a DN security model in advance; therefore, it can effectively overcome the impact of uncertainty caused by DG access on DN security . Extensive experiments show the effectiveness of the proposed algorithm.</p