Single image deraining aims to recover the rain-free image from rainy image. Most existing deraining methods based on deep learning do not utilize the global information of rainy image effectively, which makes them lose much detailed and structural information after processing. Focusing on this issue, this paper proposes a single image deraining algorithm based on Swin Transformer. The network mainly includes a shallow features extraction module and a deep features extraction network. The former exploits the context information aggregation module to adapt to the distribution diversity of rain streaks and extracts the shallow features of rainy image. The latter uses Swin Transformer to capture the global information and long-distance dependencies between different pixels, in combination with residual convolution and dense connection to strengthen features learning. Finally, the derained image is obtained through a global residual convolution. In addition, this paper proposes a novel comprehensive loss function that constrains the similarity of image edges and regions synchronously to further improve the quality of derained image. Extensive experimental results show that, compared with the two state-of-the-art methods, MSPFN and MPRNet, the average peak signal-to-noise ratio of derained images of our method increases by 0.19 dB and 2.17 dB, and the average structural similarity increases by 3.433% and 1.412%. At the same time, the model parameters of the proposed network decreases by 84.59% and 34.53%, and the forward propagation time reduces by 21.25% and 26.67%.
Aimed at the problems of low resolution, low contrast,and dark color of images recovered by traditional dehazing processing, an improved images dehazing method is proposed and applied to the unmanned aerial vehicle (UAV) aerial building image processing. First, to solve the problem that the value of global atmospheric light is easily affected by the scene objects, a method of atmospheric light with minimum variance of color attenuation prior projection is proposed. The difference image of brightness and saturation is constructed to solve the region where the minimum variance occurres, and the estimation of global atmospheric light is determined. Then, the regional atmospheric light is fused with the global atmospheric light by using the depth information of the image scene, and a new atmospheric light image is obtained. Finally, the haze line based on the non-local information prior theory in view of the transmittance is optimized. Moreover, this paper proposes a method based on the theory of haze line and guide relative to the total variation regularization algorithm. The transmission rate is fixed through calculating transmittance reliability function. A large amount of useless texture information existing in the image is eliminated, which enhances the precision of transmission rate estimation. It effectively improves the image quality of thick haze and abrupt depth-of-field in UAV aerial shooting scene. The experimental results show that, compared with other algorithms, the average gradient, contrast, haze aware density evaluator, and blur coefficient of the recovered images are improved by 12.2%, 7.0%, 11.9%, and 12.5%, respectively. The operation time of the proposed algorithm is shorter than that of some other algorithms, and the processed aerial images are clearer, which are more consistent with the visual perception of human eyes.
Precise measurement of the thermal insulation performance of thermal barrier coatings (TBCs) under the thermal gradient environment is important for the design and development of TBCs. A phosphorescent sensor TBC which contains an Eu doped yttria-stabilized zirconia (YSZ:Eu) surface layer, a YSZ intermediate layer, and a YSZ:Dy bottom layer, is designed and prepared by electron beam physical vapor deposition (EB-PVD). Based on the thermal quenching characteristics of phosphorescence signal, the surface temperature of the YSZ coating and the interface temperature of the bond-coat/YSZ layer are measured online in a temperature gradient environment, and the real thermal insulation effect of the EB-PVD YSZ thermal barrier coating is evaluated. The results show that the EB-PVD YSZ coating with a thickness of 113 μm can achieve an average temperature decrease of 66.5 ℃. The average thermal conductivity of the coating is (0.87±0.15) W/(m·K) in the temperature range between 400 and 700 ℃, which is slightly lower than the value (0.95±0.02) W/(m·K) obtained by using the traditional laser flash method. The above results validate the reliability of online phosphorescence temperature measurement technique, and provide an effective method to monitor the thermal insulation effect of TBCs in real time.
Aimed at the joint vibration problem caused by the load change of humanoid flexible joint, a torque compensation control method based on state observer is proposed. By controlling the motor to output a torque increment equivalent to the disturbance torque, the joint torque can quickly balance the load torque and shorten the oscillation process of the elastic element passively adapting to the load change. A state observer for estimating load disturbance torque and motor speed is designed, whose convergence is proved by the Lyapunov function. The control structure of the drive system based on the proportional integral-intergral proportional (PI-IP) speed regulator is established, and the observer output feedforward link is added to the speed regulator to improve the anti-interference ability of the system. The simulation results show that compared with proportional integral differential (PID) control and joint force feedback proportional differential (PD) control, the proposed method can restore the motor speed to stability within 0.6 s after load change and realize joint vibration suppression within 1 s. Besides, the joint speed adjustment time is shortened by about 1.8 s and 0.9 s respectively, which effectively improves the dynamic adjustment ability of the system. Finally, the effectiveness of the proposed method was verified by experiments on an integrated flexible joint testing platform.
A novel robot skill learning method using dynamic movement primitive (DMP) and adaptive control is proposed. The existing DMP method learns actions from a single teaching trajectory, and its Gaussian basis function distribution mode is fixed, which is not suitable for multiple movement trajectories with different characteristics. Therefore, the Gaussian mixture model (GMM) and Gaussian mixture regression are introduced into DMP to enable the robot to learn skills from multi-teaching trajectory. Moreover, radial basis function neural network (RBFNN) is introduced into DMP to establish the RBF-DMP method, which is able to learn the central position and weight of Gaussian basis through gradient descent and improves the accuracy of skill modeling. Furthermore, an adaptive neural network controller is designed to control the learned actions of the manipulator in redemonstration. Finally, experiments on Franka Emika Panda manipulator prove the effectiveness of the proposed method.
Periodic denial-of-service (DoS) attack prevents the information exchange of measurement channel and control channel for switched positive systems through the network. Moreover, the coupling of system switching behavior and hybrid behavior due to DoS attack and positive constraint also increase the difficulty in secure control research. The secure control problem of networked switched positive systems under periodic DoS attack is studied and a modeling method based on switching behavior is proposed in this paper. Under variable periodic switching, the multiple linear copositive Lyapunov function is constructed to obtain the positive and asymptotically stable condition for the closed-loop system by limiting the relationship between the switching period and the DoS attack period, and a secure state feedback controller based on the positive system theory is designed, based on which, the constant periodic switching case is also discussed by using the positive system theory. Finally, the validity of the results is verified by a simulation example.
A decoupling and synchronization control strategy based on a model-compensated extended state observer is proposed in control of a dual-drive gantry positioning stage with asymmetric flexure-linked structures. A gantry dynamic model considering the features of the flexure-linked structures and the impact of load acceleration is built using Lagrangian equations. Translational and rotational control loops are designed according to this model, while extended state observers with compensation of known coupling items in the model are deployed in each control loop to estimate and attenuate the total disturbances composed of unmodeled dynamics, unknown couplings, etc. The experimental results show that the proposed control strategy, which is simple to realize and requires few parameters to be measured, can effectively decouple the dynamics of dual axes and attenuates disturbances, greatly improving the dynamic response and synchronization accuracy of the gantry stage.
Optical detection of partial discharge (PD) is an important way to reflect the insulation status of equipment. C4F7N/CO2 gas mixture is one of the most potential substitutes for SF6 at present, but there is a lack of research on its optical PD characteristics and diagnostic methods. In this paper, a PD multispectral array detection platform that can collect 7 characteristic bands is constructed, and 4 kinds of PD defects are produced. The similarities and differences of the PD multispectral characteristics in phase distribution, energy distribution, and feature stacking map under the conditions of 5 different ratios of C4F7N/CO2 gas mixture and pure SF6 gas are analyzed. Finally, a novel method of PD diagnosis based on multispectral features (MF) and k-nearest neighbors (KNN) is proposed. The experimental results show that the fault recognition accuracy in pure SF6 can reach 96.2%. The recognition rate of C4F7N/CO2 gas mixture is above 88%, and the highest accuracy rate is 91.1%. This method has a guiding significance for the PD diagnosis of environmentally friendly gas-insulated equipment, and provides a new route for traditional PD detection and diagnosis.
Local trajectory planning is one of the key technologies of the autonomous valet parking system. In this scenario, there exist problems such as long planning time, discontinuous curvature, and insufficient safety in local trajectory planning methods for intelligent vehicles. Aimed at these problems, this paper proposes a trajectory planning method for intelligent vehicles in parking scenarios. This method improves the real-time performance and security of the initial path search by improving the analytic expansions of the hybrid A* algorithm and introducing the risk function. Further, according to the initial path, the quadratic programming method is used to realize path smoothing and speed planning. Finally, the trajectory generation is completed. Simulation experiments show that the method can improve the real-time, smoothness, and safety of intelligent vehicle trajectory planning. In addition, in actual parking environment, the feasibility of the method is verified in real-world vehicle experiments.
Due to its huge amount of calculation and memory occupation, the traditional deep neural network is difficult to be deployed to embedded platform. Therefore, lightweight models have been developing rapidly. Among them, the lightweight architecture MobileNet proposed by Google has been widely used. To improve the performance, the model of MobileNet has developed from MobileNetV1 to MobileNetV3. However, the model has become more complex and its scale continues to expand, which is difficult to give full play to the advantages of lightweight model. To reduce the difficulty of deploying MobileNetV3 on embedded platform while maintaining its performance, a structured pruning method integrating the characteristics of MobileNetV3 is proposed to prune the lightweight model MobileNetV3-Large to obtain a more compact lightweight model. First, the model is trained by sparse regularization to obtain a sparse network model. Then, the product of the sparse value of convolution layer and scale factor of batch normalization layer is used to identify the redundant filter, which is structurally pruned, and experiment is conducted on CIFAR-10 and CIFAR-100 datasets. The results show that the proposed compression method can effectively compress the model parameters, and the compressed model can still ensure a good performance. While the accuracy remains unchanged, the number of parameters on CIFAR-10 in the model is reduced by 44.5% and calculation amount is reduced by 40%.
The thermal cycling system is the key component of the real-time fluorescent polymerase chain reaction (PCR) instrument, which determines the nucleic acid detection efficiency and result accuracy. Aiming at the problems that the thermal cycle of the traditional PCR detection system takes a long time and the temperature control is complicated, a real-time fluorescent PCR thermal cycle system with partition temperature control is designed, including the hardware circuit and mechanical structure of the thermal cycle system. The rapid thermal cycling is achieved by switching the test solution between different constant temperature zones, and the temperature is controlled using an incremental proportional integral derivative (PID) algorithm, with a control precision of ±0.1 ℃. Using Fluent software to establish heat transfer model, the thermal delay phenomenon of the test solution was analyzed to predict the temperature variation pattern of the test solution. A prototype is built for testing and verification, and the heating and cooling rates of the test liquid are 3.8 ℃/s and 4.4 ℃/s. It is verified that the proposed PCR thermal cycling system can effectively improve the detection efficiency.
