In this study, a displacement sensing system based on a variable spacing grating was proposed and experimentally demonstrated. The variable spacing grating was fabricated by electron beam lithography and reactive ion etching technique, and the initial line density was 950 L/mm. Double collimators were selected for transmit incident light and receive diffracted light respectively, and the incident and diffracted light angles were maintained at 46 degrees. A one-dimensional displacement platform was controlled to move the light spot on the grating surface, during 40 mm displacement, the central wavelength of the diffraction spectra was changed from 944.01 nm to 1553.03 nm gradually. Then, the light spot was moved back to its initial position, and the central wavelength of the diffraction spectra were adjusted from 1553.03 nm to 944.37 nm. The displacement sensitivity and linearity were 15.310 nm/mm and 0.986, respectively. The system has a large displacement measurement range and exhibits good repeatability.

In this paper, the sensor was proposed by combination of grapefruit photonic crystal fiber (GPCF) and femtosecond laser fabricated fiber Bragg grating (FBG) based on Sagnac interferometer. The GPCF was sandwiched between two single-mode fibers (SMFS) to form a SMF-GPCF-SMF structure, which was based on intermodal interference. The FBG with a central wavelength of 1535.32 nm was inscribed through polyimide coating and cladding by the femtosecond laser point-by-point inscribing method. The spectrum drift data of sensing structure were collected and the dual-parameters matrix of temperature-strain was constructed to realize the simultaneous measurement of temperature and strain. The experimental results showed that the linear range of the strain measurement was from 0 με to 2,000 με at different temperature, the wavelength of FBG and GPCF was red drift and blue drift, respectively, the average strain sensitivity at different temperature was 1.25 pm/ με and −2.05 pm/ με , all the linearity R 2 are higher than 0.999, and has the high repeatability. The linear range of temperature measurement was from 20 °C to 450 °C, the wavelength of FBG and GPCF was red drift and blue drift, respectively, the temperature sensitivity of FBG and GPCF was 15.17 pm °C −1 and −8.87 pm °C −1 , the linearity was 0.99983 and 0.99989, respectively.

提出用于卫星姿态控制的光纤陀螺故障自动定位方法,以光纤陀螺光源控制电路产生的光功率为故障信号提取目标,根据不同状态下光纤陀螺的光功率对故障信号进行分类,并利用定位函数对数据进行分析,确定光纤陀螺故障类型和故障位置,以此实现光纤陀螺故障自动定位.实验证明,本方法定位误差小于传统方法,更适用于光纤陀螺故障自动定位.

A wavelength-tunable double-ring cavity erbium-doped random fiber laser (RFL) based on a Mach-Zehnder interferometer (MZI) was proposed in this study. The MZI structure, which was based on two 2 x 2 optical fiber couplers, exhibits a comb-filtering effect. Rayleigh scattering (RS) was generated by a 10-km-long single-mode fiber (SMF) to provide randomly distributed feedback. A double-ring cavity structure based on forward and backward RS was built using two 2 x 2 fiber couplers. The threshold of the laser was 104 mW, and the slope efficiency of the total power output to the pump power was approximately 1.2%. When the pump power was 158 mW, a laser beam with a wavelength of 1555.3 nm was formed. By adjusting the polarization controller (PC), a single-wavelength switched RFL at 1547.9 nm and 1559.6 nm was obtained. The maximum peak power difference was less than 0.67 dB, the signal-to-noise ratio (SNR) was greater than 17.27 dB, and the peak power fluctuation was less than 0.92 dB during a scan time of 30 min at 25 degrees C. Regarding dual-wavelength switchable laser emission, two groups of different wavelengths with an SNR greater than 16.88 dB, power fluctuation less than 0.81 dB, and wavelength fluctuation less than 0.26 nm could be achieved under the same monitoring conditions. By further adjusting the PC, three- and four-wavelength RFLs were achieved, and the SNRs exceeded 16.31 dB and 14.64 dB, respectively. The proposed RFL has promising application prospects for distributed sensing systems and speckle imaging.

Recent studies on deep-learning-based small defection segmentation approaches are trained in specific settings and tend to be limited by fixed context. Throughout the training, the network inevitably learns the representation of the background of the training data before figuring out the defection. They underperform in the inference stage once the context changed and can only be solved by training in every new settings. This eventually leads to the limitation in practical robotic applications where contexts keep varying. To cope with this, instead of training a network context by context and hoping it to generalize, why not stop misleading it with any limited context and start training it with pure simulation? In this paper, we propose the network SSDS that learns a way of distinguishing small defections between two images regardless of the context, so that the network can be trained once for all. A small defection detection layer utilizing the pose sensitivity of phase correlation between images is introduced and is followed by an outlier masking layer. The network is trained on randomly generated simulated data with simple shapes and is generalized across the real world. Finally, SSDS is validated on real-world collected data and demonstrates the ability that even when trained in cheap simulation, SSDS can still find small defections in the real world showing the effectiveness and its potential for practical applications. Code is available here

In-flight objects capture is extremely challenging. The robot is required to complete trajectory prediction, interception position calculation and motion planning within tens of milliseconds. As in-flight uneven objects are affected by various kinds of forces, which leads to the time-varying acceleration, motion prediction for them is difficult. In order to compensate the system’s non-linearity, we propose using a recurrent neural network model, which we call the Neural Acceleration Estimator (NAE), to estimate the varying acceleration by observing a small fragment of previous deflected trajectory without any prior information. Moreover, end-to-end training with Differantiable Filter (NAE-DF) gives a supervision for measurement uncertainty and further improves the prediction accuracy. Experimental results show that motion prediction with NAE and NAE-DF is superior to other methods and has a good generalization performance on unseen objects. We test our methods on a robot, performing velocity control in real world and respectively achieve 83.3% and 86.7% success rate on a ploy urethane banana and a gourd. We also release an object in-flight dataset containing 1,500 trajectorys for uneven objects, which can be found on the project website:https://sites.google.com/view/neural-motion-prediction.

In the long-term deployment of mobile robots, changing appearance brings challenges for localization. When a robot travels to the same place or restarts from an existing map, global localization is needed, where place recognition provides coarse position information. For visual sensors, changing appearances such as the transition from day to night and seasonal variation can reduce the performance of a visual place recognition system. To address this problem, we propose to learn domain-unrelated features across extreme changing appearance, where a domain denotes a specific appearance condition, such as a season or a kind of weather. We use an adversarial network with two discriminators to disentangle domain-related features and domain-unrelated features from images, and the domain-unrelated features are used as descriptors in place recognition. Provided images from different domains, our network is trained in a self-supervised manner which does not require correspondences between these domains. Besides, our feature extractors are shared among all domains, making it possible to contain more appearance without increasing model complexity. Qualitative and quantitative results on two toy cases are presented to show that our network can disentangle domain-related and domain-unrelated features from given data. Experiments on three public datasets and one proposed dataset for visual place recognition are conducted to illustrate the performance of our method compared with several typical algorithms. Besides, an ablation study is designed to validate the effectiveness of the introduced discriminators in our network. Additionally, we use a four-domain dataset to verify that the network can extend to multiple domains with one model while achieving similar performance.

The fused use of the long-focus camera and telemetry LiDAR is necessary for advanced long-distance detection requirements in autonomous driving vehicles and unmanned subways, and the extrinsic parameter calibration is a precondition for multi-sensor fusion use. However, the inaccuracy of long-focus camera detection and low accuracy of data association between LiDAR-camera make the calibration a challenging problem. In this paper, we propose a novel calibration method that treat the detection poses as variables in the optimization framework, which significantly reduces the impact of detection errors on calibration accuracy. Furthermore, we construct 3 error terms to obtain better estimation of the variables, and the designed error terms only require plane-to-plane data association, avoiding incorrect data association that leads to failure in extrinsic parameter estimation. Then we made a full comparison with other methods through real-world experiments, which showed that our method achieves the best calibration result.

This article presents a high-precision single-camera inertial measurement unit (IMU) extrinsic calibration method by tightly fusing the visual information from other cameras. Specifically, multiple additional cameras are added to the monocular camera-IMU system for assisting calibration as we theoretically prove that more cameras used in calibration can lead to smaller lower bound on the covariance of the estimated extrinsic parameters, which then results in better calibration accuracy. Moreover, we provide two degenerative motion conditions in the resulting multicamera visual-inertial system, which impair the calibration accuracy and should be avoided in real application whenever possible. More importantly, we present the requirement of minimum motion for a reliable extrinsic calibration to provide the practical guideline. Finally, the full validation on both simulation and real-world data is demonstrated. By evaluating the Cramér-Rao lower bound on the covariance, the proposed camera-IMU calibration method is shown to be statistically efficient for accurate calibration with errors less than 0.01 m in translation and 0.5° in rotation, which is consistent with the theoretical analysis in this article.

The technology of detecting and identifying multitarget is of significance for robotic visual navigation within the unknown, unstructured, complex scenes, whose result could be considered as important references of 3D map building and path planning for mobile robot. Traditional algorithms of target detection can be applied to 2D images without depth information generally, which is disturbed by few factors such as illumination, view and scale easily. Therefore, an approach on visual detecting multi-target of unstructured and complex scenes based on RGBD images is proposed in this article to solve the above problem, which is composed of extracting descriptor of rotation and scale invariance feature, local encoding of targets, random ferns classifier training, Hough map generation, Hough voting theoretical model and local maximum search. Experimental results have shown that the proposed approach reduce the calculation of extracting and matching local feature, improve the accuracy of object recognition and detection in unknown complex environments, be capable of well robust against few disturbing factors i.e. rotation, scale, illumination, occlusion and non-rigid body deformation.

Map based visual inertial localization is a crucial step to reduce the drift in state estimation of mobile robots. The underlying problem for localization is to estimate the pose from a set of 3D-2D feature correspondences, of which the main challenge is the presence of outliers, especially in changing environment. In this paper, we propose a robust solution based on efficient global optimization of the consensus maximization problem, which is insensitive to high percentage of outliers. We first introduce translation invariant measurements (TIMs) for both points and lines to decouple the consensus maximization problem into rotation and translation subproblems, allowing for a two-stage solver with reduced search space. Then we show that (i) the rotation can be estimated by minimizing TIMs using only 1-dimensional branch-and-bound (BnB), (ii) the translation can be estimated by running 1-dimensional search for each of the three axes with prioritized progressive voting. Compared with the popular randomized solver, our solver achieves deterministic global convergence without requiring an initial value. Furthermore, ours is exponentially faster compared with existing BnB based methods. Finally, our experiments on both simulation and real-world datasets demonstrate that the proposed method gives accurate pose estimation even in the presence of 90% outliers (only 2 inliers).

Human-robot interaction (HRI) is considered as one of the key techniques of space intelligence robots. Few typical features of complicated human actions need to be captured and understood accurately by intelligent robot to ensure free communication and interaction between both above in real-time, especially for identifying and tracking hand state. There are four approaches for gesture recognition, including algorithm based on Kinect V2 SDK, model-based particle swarm optimization algorithm (PSO), deep learning algorithm and gesture module based on Baidu AI platform. These have been selected to compare in the form of principle, calculating time, robustness, range, environmental adaptability and correct rate respectively. The comparative results have generalized that both the second and the third algorithm have better performance than other algorithms in the above aspects such as calculating efficiency, robustness, detecting range, external disturbance and correct ratio. Particularly, the second algorithm is not only suitable for close range, but also suitable for multi-view cases. However, the third algorithm can have better performance, but depends on precise network model and weights by introducing lots of positive and negative gesture samples.

Face recognition which is of few advantages such as natural and non-contact to realize fluent interaction and cooperation between human and robot, has been one of important and common issues in the fields of computer vision and biometrics identification. However, the achievement of face recognition also meet few issues such as disturbances or variations in facial expression, pose, shade and environmental illumination to solve. For this reason, an autonomous face identification system based on deep learning is proposed in this article, which should be divided into 4 stages. Firstly, RGB-D images including one or more faces are captured by Kinect v2. Secondly, an algorithm of multi-view faces detection has been proposed by introducing candidate regions after filters of local binary Haar-like feature into Multi-layer perceptron (MLP) in order to obtain every candidate face area. Thirdly, typical face feature points such as left eye, right eye, nose tip, left corner of the mouth and the right corner of the mouth are located and aligned by Stacked Auto-Encoder (SAE) accurately. Finally, VIPLFaceNet has been applied to identify the similarity and difference between the image to be determined and any template in the face image database. Experimental results have shown that the proposed system not only can detect multi-faces belonging to different persons, but also could achieve well identification results with the correctness of no less than 70% regardless of few disturbance of pose, expression and illumination.

Visual localization has attracted considerable attention due to its low-cost and stable sensor, which is desired in many applications, such as autonomous driving, inspection robots and unmanned aerial vehicles. However, current visual localization methods still struggle with environmental changes across weathers and seasons, as there is significant appearance variation between the map and the query image. The crucial challenge in this situation is that the percentage of outliers, i.e. incorrect feature matches, is high. In this paper, we derive minimal closed form solutions for 3D-2D localization with the aid of inertial measurements, using only 2 point matches or 1 point match and 1 line match. These solutions are further utilized in the proposed 2-entity RANSAC, which is more robust to outliers as both line and point features can be used simultaneously and the number of matches required for pose calculation is reduced. Furthermore, we introduce three feature sampling strategies with different advantages, enabling an automatic selection mechanism. With the mechanism, our 2-entity RANSAC can be adaptive to the environments with different distribution of feature types in different segments. Finally, we evaluate the method on both synthetic and real-world datasets, validating its performance and effectiveness in inter-session scenarios.

Radar, optical and infrared sensors are the main loads in space-based detection system. In this paper, multisensor information fusion method is applied to space target recognition. SVM, RFM and CNN is used in the recognition method. In this paper, the feature level fusion method of multisensor image is compared with extracting the features without fusion. The advantage of fusion for improving the recognition success rate is highlighted. Finally, the D-S evidence theory method is used in decision level fusion.

Inverse Synthetic Aperture Radar (ISAR) imaging technology has been developed over half a century, and its theoretical research and imaging methods have gradually matured. and its theoretical research and imaging technology. Electromagnetic simulation software based on 3 d modeling software and build targets scattering point model, and using the model of two-dimensional SAR imaging simulation analysis, the use of traditional range - doppler algorithm based on cross-correlation method of envelope alignment processing and by the minimum variance normalized amplitude phase correction processing, thus completing the process of ISAR 2-d imaging. Then the effects of pulse repetition rate and fast time sampling rate on the 2-d imaging results of target ISAR are analyzed.

Data registration between 3D point cloud and CAD model of non-cooperative object has been considered as one of key technologies for estimating spatial position and orientation of target spacecraft. The registration result will directly affect the success or failure of on-orbit capture mission for space manipulator. Usually, 3D CAD model needs to discretize into point cloud of model which can be applied to match the corresponding 3D measuring point clouds. In this article, a coarse registration algorithm of curvature features based on distance constraint consistency is proposed to solve data registration between 3D point cloud and CAD model of non-cooperative object. According to the principle of invariant curvature of rigid transformation, a set of curvature feature points which satisfies the consistency of distance constraint can be selected to calculate rotation matrix and translation vector between both sets of point clouds. Experimental results have shown that the proposed registration algorithm can achieve higher registration accuracy to provide reliable initial values of transformation parameters for subsequent fine registration work.

空间机械臂需要在空间站外表面进行移动来实现表面巡检或抵达目标位置执行空间任务,需要通过对合作靶标的测量来获取空间机械臂与合作靶标间的位姿信息,因此设计出一种合理可靠的合作靶标尤为关键.提出一种合作靶标,针对该合作靶标提出一种识别及特征点定位算法,利用相机和合作靶标对算法进行验证.实验结果表明,该算法可在0.5～3.0 m的距离内准确识别出合作靶标,特征点的坐标定位精度可达到0.0052 mm,算法快速、稳定,抗干扰能力强.

Based on the mobile robot platform with multi-ultrasonic ranging sensors, a method is proposed to obtain local maps by ultrasonic ranging, and to obtain global maps through detection in the process of motion, so as to complete path planning and navigation. This method effectively fuses multi-sensor to detect obstacles around the mobile robot. The global map is constructed by local map, which overcomes the shortcoming of ultrasonic ranging. The method is easy to implement. The validity of this method is verified by a certain type of mobile robot platform.