Abstract:In order to accurately identify the directional intention of lower limb functional disorder patients, a novel walking directional intention identification strategy is proposed considering individual differences and safety. Firstly, the structure of rehabilitation robot and the relationship between the pressure of the forearm and the directional intention are introduced. To ensure the patient to walk safely in any directions, a distancetype fuzzy reasoning algorithm with sparse antecedents is adopted under the prerequisite for reasoning about safe gait through rotation angle of knees. Then, to reduce the identification errors caused by individual differences and nonstable fuzzy rules, this paper proposes a rule evolutionary strategy, which can update the fuzzy inference rules. Finally, the algorithm is applied to multiobjective fuzzy reasoning experiments and pressure control experiments of walking rehabilitation robot, which prove that the algorithm can accurately identify arbitrary directional intention of patients and increase walk safety. This strategy of directional intention identification can be used in the rehabilitation of lower limb dysfunction in daily living and rehabilitation training.

Abstract:As a kind of tall moving obstacle, wind turbines have an inevitable impact on weather radar performance. The analysis of electromagnetic scattering characteristics of wind turbine is significant for the identification of wind farm disturbing and the evaluation of influence on weather radar. In this paper,considering the effect of the dielectric constant for cylindrical blade on the radar cross section (RCS), an analytical formula of the RCS calculation for cylindrical blade is proposed based on numerical theory, and the results are validated by simulation. Meanwhile, a real blade is modeled by using software UG in consideration of airfoil section. Therefore, the electromagnetic scattering characteristics including RCS and timefrequency domain analysis are established. Simulation results demonstrate the validity of the modelling method in comparison with the measured data. Finally, an analytical formula of real blade′s RCS is adapted with a certain range.

Abstract:In order to improve the detection precision of fly ash carbon content using microwave method and fill the blank of fly ash electromagnetic parameter measurement in broadband, the coaxial transmission line method and coaxial probe method are used to study the frequency response characteristics of the S parameter and complex dielectric constant of fly ash with different carbon contents in the frequency range of 3~18 GHz in experiment, respectively in this paper. The experiment results show that in the frequency range of 3~18 GHz and carbon content of greater than 7%, there is a good linear relationship between S21 and carbon content, the fitness degree is greater than 0.96, and the measurement sensitivity is positively correlated with frequency. In the frequency range of 10~18 GHz, the carbon content and dielectric constant ε′ show a good linear relationship, the fitting degree is greater than 0.98. The relative uncertainty is below 4%. In addition, the reason for the poor regularity and repeatability of the S11 and loss factor measurement results are analyzed from the test uncertainty and measurement methods.

Abstract:Traditional CNN models are suitable for the feature extraction of gray image sequences or the color image separate channels, which ignores the interdependency among the channels and destroys the color features of real world objects, thereby affects the accuracy rate of human body action recognition. In order to solve this problem, a human body action recognition method is proposed based on quaternion spatialtemporal convolutional neural network (QSTCNN). Firstly, codebook algorithm is adopted to process all the images in the sample set and extract the key regions of human body motion in the images. Then，the quaternion matrix expression of the color images is taken as the input of the QSTCNN. The spatial convolutional layer of CNN is expended as a quaternion spatial convolutional layer. The values of the red, green, and blue channels of the color images are considered simultaneously as a whole in a spatial convolutional layer to conduct the extraction of the action spatial features, and avoid the loss of spatial relationships. The dynamical information of adjacent frames is extracted in a temporal convolutional layer. Finally, experiment was conducted, in which QSTCNN, gray single channel CNN (GrayCNN) and RGB three channel CNN (3 ChannelCNN) were compared. The experiment result demonstrates that the QSTCNN boosts the performance of action recognition, the proposed method is superior to other popular methods and achieves the recognition rates of 85.34% and 80.2% in the Weizmann and UCF sports datasets, respectively.

Abstract:The iris images collected by iris recognition system used for smart mobile devices are seriously interfered under visible light. The interference reduces recognition accuracy and decreases the robustness. An iris recognition method based on analogous convolutional neural network and local feature extraction is proposed. Firstly, a haze removal algorithm using dark channel is utilized to enhance the iris texture and reduce the light interference. Then, the analogous convolutional neural network is used to reduce the image dimension, and the texture information is obtained with a binary image. The feature vector of the iris image is built by local feature extraction of the regions of the lower dimension image. Finally, Euclidean distance is utilized in matching process. To validate the performance of the proposed method, the 30 people’s 240 iris images (four indoor images for each person and four outdoor images for each person) in MICHEI iris gallery are tested, and a comparison is conducted with iris recognition methods of Gabor transform and Principal Component Analysis (PCA). The results show that the recognition accuracy under the condition of both indoor and outdoor images can reach 98.33% and the approach has better robustness under the indoor and outdoor light interference. The performance is superior to the Gabor transform and PCA methods. These demonstrate that the proposed method can satisfy the requirements of iris recognition on mobile device.

Abstract:The unmanned underwater vehicle (UUV) path tracking control provides the important technical foundation for various civilian and military applications of UUV. Aiming at the underactuation, nonholonomic constraint and model nonlinearity in the path tracking control of UUV, an UUV vertical path tracking controller is designed based on the nonlinear continuous model predictive control algorithm. The vertical motion model of UUV is established, and the prediction model of UUV system is given based on state space model. According to the given performance index, the Taylor series expansion and Lie derivative are used to solve the optimal control law under the continuous time state, the underactuated UUV vertical path tracking control is achieved. A simulation experiment was carried out, the result proves the effectiveness of the design of the vertical path tracking controller.

Abstract:A SINS/CNS/SAR integrated navigation system suitable for cruise missile is designed in this paper. Aiming at the requirements of high precision and autonomy of the navigation system and the characteristics of uniform speed and constant height flight of cruise missile, the engineering practicability of the project is highlighted. The nonlinear modeling is adopted for the SINS/CNS/SAR integrated navigation and the nonlinear UKF method is used to perform the filtering of the subfilters. The nonlinear model and filter method are more practical, so the system accuracy is improved. In the information fusion process among the subfilters, the federal filter based on the information allocation factor is selected to adjust the system in realtime, which greatly enhances the fault tolerance and realtime performance of the system. A new method for computing observability is proposed and the observability computation is simplified. The digital and hardware in the loop simulation experiments show that the integrated navigation system is very suitable for the high altitude long endurance flight environment of cruise missiles. The integrated navigation system is not only suitable for cruise missiles, but also has great engineering value for some other long endurance aircrafts, such as UAVs, and has great engineering application value.

Abstract:In order to solve the difficulty of high precision measurement and evaluation of the surface morphology of laser fusion target, a laser differential confocal sphericity measurement and evaluation method for the laser fusion target is proposed. This method uses the property that the zerocrossing point of the axial response curve of the laser differential confocal measurement system corresponds to the focus point position of the objective lens to realize the target surface focusing. Through rotating the target in two orthogonal directions respectively, a number of crosssections of the target surface are measured and acquired; and the whole morphological structure measurement for the target is achieved. The minimum zone sphericity evaluation algorithm is used to establish the threedimensional model for the target sphericity evaluation, the quantitative evaluation of the target morphological structure is conducted. Based on the method, an experiment system was built to perform the morphological structure measurement of the target surface. The evaluation results show that with the proposed method the measurement repeatability reaches 0.15 μm, which provides a feasible scheme for the morphological structure measurement of fusion target.

Abstract:Aeromagnetic compensation accuracy is one of the key factors affecting the accuracy of geomagnetic navigation. A component compensation method is proposed based on improved Fuzzy Adaptive Kalman Filtering. The observation matrix in Kalman filter is modified to effectively improve the low speed convergence. By monitoring the variance ratio between the residual output and theoretical residual at each moment, the measurement noise covariance is modified step by step and the optimal parameters are obtained. Simulation results show that this presented method has strong adaptability to timevarying measurement noise. The total field compensation error is reduced from 175.8 to 6.7 nT, component error is reduced from 59.2, 110.2, 122.2 nT to 6.0, 1.7, 1.5 nT, respectively. Experiment is carried out by constructing a vector system and its results demonstrate that the proposed method performs better in convergence speed and compensation accuracy. The proposed method can provide a good reference to distortion magnetic field compensation.

Abstract:The intrinsic magnetism calibration of the marine vector magnetometer is the core technology for magnetometer application system, especially when the magnetism of the carrier changes; the insitu calibration algorithm is the key to influence its application effectiveness. This paper analyzes the main magnetic interference sources of the underwater exploration system carrying three axis fluxgate magnetometer firstly, describes corresponding magnetic calibration model, and introduces the ellipsoid fitting algorithm, particle swarm optimization algorithm, genetic algorithm and particle swarm optimizationgenetic hybrid algorithm briefly. These algorithms are realized with MATLAB respectively. Finally, the field operation data of the underwater vehicle in the southwest Indian ocean were used to verify the functions of the above algorithms. From the results of the data fluctuation before and after calibration, rootmeansquare error and relative error, it can be seen that the hybrid algorithm can get better calibration effect and improve the measurement effectiveness of the magnetometer sensor in the insitu magnetic calibration.

Abstract:GNSS time transfer technology is an important application of satellite navigation technology in time and frequency domain, and is also one of the main means of time transfer. With the continuous development of China′s Beidou satellite navigation system, the applications in the fields of time and frequency based on Beidou also gradually increase. In his work, time transfer algorithm is focused and software package (PPTSlo) is developed using Beidou common view and carrier phase technology, based on specific IGSO, MEO, GEO satellite constellation in Beidou satellite navigation system. The time transfer link accuracy and stability is analyzed by qualitative and quantitative analysis. Experimental results show that Beidou common view time transfer accuracy is better than 6ns, Beidou carrier phase time transfer accuracy can reach subnanosecond to nanosecond.

Abstract:With the developing of global navigation system, the high quality of positioning navigation and time service are gradually provided by Beidou navigation satellite system (BDS) in China. In this paper, the data of BDS receiver operated in time keeping laboratory of national time service center (NTSC) are used, the BDS range signal evaluation and precise point positioning (PPP) are studied. Based on the measurements data, the signaltonoise ratio and multipath effect of B1 and B2 frequencies are analyzed. And the algorithm of BDS PPP is discussed in detail. The precision positioning and timing information is calculated by using the data and the products of precise orbit and precise satellite clock from MGEX of IGS. The result show that reception performance of B2 frequency point is better than B1 frequency point, and the BDS PPP error of X, Y, Z kept at cm level. The short term frequency stability of time difference between local time and IGST is in 10-14, and the performance is approximately the same as global positioning system (GPS). It is proved that the BDS can be used in ns level time transfer.

Abstract:Optical fiber acoustic sensor has attracted a significant amount of research attentions in recent years. A key issue for all optical fiber acoustic sensors is to improve the quality of detected signal. In this work, a novel wavelet packet speech signal retrieving method is proposed for improving detection performance of optical fiber acoustic sensors. The experiment system of optical fiber acoustic sensing is set up based on the principle of optical fiber interferometer. The experimental results show that the new threshold processing method improves the signaltonoise ratio compared with hard threshold and soft threshold processing method. Furthermore, the continuity of speech signal is guaranteed and the speech quality is improved. The proposed method can be widely used for a variety of speech detecting applications, especially for realtime speech detecting.

Abstract:In order to improve the ability of acoustic emission modal recognition of Micro/nono Cupling Fiber Sensor (MCFS), based on the theory of modal acoustic emission, the influences of the effective sensing area structure, length and symmetry of micro/nono coupling fiber on the modal identification ability of the sensor has been studied in experiment, and the orientation properties of the optimized sensor have been analyzed. Experiment results show that: When the effective sensing area structure is “X” type, including separation zone, exact taper zone and waist zone, the sensor can`t identify the S0 modal and the integrity of the identified A0 modal is poor; when the effective sensing area structure is “” type, including waist zone and all the exact taper zones, the sensor can identify the A0 modal and identify S0 modal completely. When the effective sensing area is symmetric, the integrity of the measured signal is good; when the effective sensing area is asymmetric, the integrity of the measured signal is poor. For the“X” type structure, in the separation zone partial sound wave energy is attenuated, the sensitivity of the sensor increases as the length of the effective sensing area decreases; however, for the “”type structure, the sensitivity of the sensor decreases as the length of the exact taper zone decreases due to the strain magnification effect of the exact taper. The modal identification ability of the sensor decreases as the detecting angle decreases, the effective detecting angle range is from 45° to 135°. This study has great significance to the application of this sensor in modal acoustic emission.

Abstract:Aiming at the plane angle insite measurement requirement, a noncontact measuring device for plane angle is developed based on the diffuse reflection measurement features of the laser displacement sensors. The spatial coordinate calibration method of the laser displacement sensors are proposed that combines the coordinate measuring machine and position sensitive detector, an accurate measurement model for plane angle is constructed. The Montel Carlo method is used to evaluate the uncertainty of the plane angle noncontact measurement device, and the result is U=0.044°~0.046° (k=2) for the angle measurement range of ±25°. The performance verification test, repeatability test and stability test were performed to examine the specifications of the noncontact measurement device. The results show that the absolute indication error, repeatability and stability of the device are less than 0.036°, 0.004° and 0.021° respectively in the measurement range of ±25°. The test results demonstrate that the noncontact measurement device for plane angle based on laser displacement sensors is accurate, reliable and has the application prospect for the insite measurement of plane angle.

Abstract:The magnetic memory method can effectively determine the stress damage areas of the ferromagnetic metal components. However, the formation mechanism and complicated influenced factors of the spontaneous magnetic flux leakage signal make it hard to quantitatively analyze the magnetic memory signals under different stress concentration, and limit the practical applications. According to electron spin, the electronic exchange model of the sd orbit is established, the changing rules of the electronic spin of density, the atomic magnetic moment and the crystal lattice size nearby the field of the crystal are calculated. Furthermore, the relationship between the magnetic memory signals and the degree of stress concentration is quantitatively analyzed. Results show that the magnetic memory signal varies linearly with the stress. Before the yielding stress appearing, the magnetic memory effect is determined by the electronic spin interactions of the d orbit, and the repeatability of the relationship between the magnetic memory signals and the stress is good. After the yielding stress appearing, the electronic spin interactions enhance, the electronic spin interactions of the d orbit weaken, the electronic action of the s orbit enhances, the amplitude of variation of the magnetic memory signal decreases, and the whole magnetic memory effect weakens.

Abstract:Support vector data description (SVDD) has been applied to fault detection of batch processes without any restriction of process data distribution. However, the SVDD based traditional fault detection method of batch processes just simply divides the phase of batch processes using cluster analysis and model identification, and leads to rough phase division results which may compromise the accuracy of fault detection in multiphase batch processes. To address this issue, a SVDD based fault detection method for multiphase batch processes is presented. Firstly, the different phases are divided according to the change of hypersphere radius and support vectors of SVDD. Then, the SVDDbased fault detection models of different phases are established. Finally, fault detection is achieved by judging whether the difference between hypersphere radius of corresponding phase and the distance from the sample point to center of corresponding hypersphere exceeds the control limit. The experimental results of the fedbatch penicillin fermentation process show that the presented method can achieve phase division of multiphase batch processes with a better accuracy, and further realize multiphase fault detection of batch processes with a higher fault detection rate.

Abstract:Based on the fundamental theory of highly nonlinear solitary waves (HNSWs), three transducers embedded with PZT circular materials are investigated, and designed for the nondestructive testing (NDT) of the Young’s Modulus of materials. Three different materials, including Teflon, granite and stainless steel, are tested by the transducers. The testing results show that the differences among transducers are below 0.5%, and the ratio between standard deviation and average value for single transducer is below 0.4%, which proves the high consistency and repeatability of the three transducers. The deviations from the theoretical analysis and the three materials are below 3%, which confirms the accuracy and reliability of the transducers. By comparing the results of Ultrasonic Pulse Velocity method, the HNSWs based transducers are more affordable, accurate and reliable.

Abstract:Semantic information can help the robot to better understand unknown environment and lay the foundation for more advanced humancomputer interaction and more complicated task. To enable mobile robot to build semantic map in real time, a light deep learning model is developed for object detection on embedded computer Jetson TX1. The interframe optical flow information in the video stream is used to reduce the missing rate of object detection algorithm, which is called motion guided propagation (MGP) algorithm. A realtime depth map restoration algorithm based on CUDA is utilized because the depth map generated by Kinect has black hole and black border. SLAM technology is employed in this paper for robot location, navigation and mapping. On this basis, Bayesian inference framework is integrated with measurement information of environment and object detection information to complete the building of semantic map. Experiments show that the proposed method can enable the mobile robot to build the semantic map in real time in the real, complicated indoor environment.

Abstract:The crack is a serious defect on the ice cream stick surface, which affects the processing and usage of the ice cream stick seriously. However, certain thin and light cracks are similar to the wood fiber textures on the ice cream stick surface, which results in the poor extraction performance of present detection algorithms. Aiming at this problem, a detection scheme based on the combination of the mainlobe and sidelobe texture gray features is proposed on the basis of detailed analysis of crack texture characteristic and wood fiber texture characteristic. Firstly, the basic model of the texture mainlobe and sidelobe gray characteristic extraction is built. Secondly, the whole texture edges on the ice cream stick head are extracted. Then, the mainlobe and sidelobe texture gray features of corresponding textures of the edges extracted in the previous step are extracted according to the established model, and the candidate crack edges (containing all the crack edges and certain wood fiber texture edges) are preliminarily determined according to the mainlobe characteristic quantity size. Lastly, the crack texture edges are recognized from the candidate edges derived from the previous step according to the numerical relationship of the sidelobe characteristic quantity and the preset threshold value; and the crack defect detection is achieved. The algorithm was tested on the selfbuilt image database SUTI2. The results show that the crack defect FAR of this method is as low as 6.07 percent on the premise that the missing detection rate is 0; the missing detection rate is decreased by at least 9.29 percent compared with other ice cream stick or wood fiber surface crack detection algorithms, which indicates that the proposed crack detection scheme is superior and has actual application value.

Abstract:An image crack detection method with multiscale downsampled normalized cut is proposed to address the problems of low precision for edge detection and timeconsuming for feature vector solution with the multiscale normalized cut. Firstly, the feature of halfreconstruction of the antisymmetrical biorthogonal wavelet is used to extract the multiscale edge features of the test image Then, combining the strength feature and location feature of each scale, the multiscale similarity matrix and multiscale normalized similarity matrix are obtained. Spectral segmentation method is utilized to calculate the downsampled feature vector of the multiscale similarity matrix after downsampling. Finally, by multiplying the multiscale normalized similarity matrix with the downsampled feature vectors, the segmentation result is obtained after discretization. Experimental results indicate that the proposed method improves the accuracy of detection and reduces the computational time on three image datasets, compared with other methods.

Abstract:Various surface defects such as scratch, scrape and etc. may occur on the high reflective metal parts with highlight surface during production and posttreatment, which seriously affects the performance and service life of the product. The surface of this kind of parts has the characteristic of specular reflection, which leads to the miss detection and wrong detection of the defect object in detection process. Aiming at this problem, this paper proposes a new highlight surface defect recognition method with global threshold adaptive adjustment capability based on digital image processing technology. Firstly the filtering pattern fully using the information of both spatial domain and value domain is constructed, which is used to process the original image and preserve the edge information of the object. Secondly, the first derivative of Gaussian function is used to construct the Canny optimal edge detector, which combines with the global threshold maximum betweenclass variance method (Otsu segmentation method) and morphological image segmentation method to complete the image segmentation and the adaptive adjustment of corresponding threshold, and achieve the identification of the defect object. Experiment results verify the effectiveness and reliability of the algorithm. The proposed algorithm could effectively identify the defect object while eliminating the influence of highlight interference. The method has great significance to the automatic and accurate defect recognition on the highlight surface of the metal parts.

Abstract:Aiming at current situation of low efficiency and being affected seriously by the weather in photoelectric sunshine duration recorder calibration process, based on the technology of solar simulation, integrating sphere brightness smoothing, light absorption trap and multidimensional adjustment, a new indoor calibration method for photoelectric sunshine duration recorder is proposed and a photoelectric sunshine duration recorder calibration system is designed. Firstly, according to the composition and operating principle of photoelectric sunshine duration recorder, the overall design scheme of the photoelectric sunshine duration recorder calibration system is put forward. Secondly, the main technical specifications of direct radiation simulation system, scattered radiation simulation system and solarterrestrial relationship regulating system are studied and analyzed. Then, according to the correlation among the components of the calibration system, the calibration chain of the calibration system is established, and the calibration methods of the components of the calibration chain are studied, and the traceability of the calibration result is realized. Finally, the simulation error of the sunshine hours of the photoelectric sunshine duration recorder calibration system is verified with experiments. The experiment results show that the simulation error of the sunshine hours of the photoelectric sunshine duration recorder calibration system is 3.75% within 4 hours, which satisfies the calibration requirements of current photoelectric sunshine duration recorder.

Abstract:There are a large number of analogy meters due to complex electromagnetic environment in transformer substation and these meters need manual reading, which makes it difficult to automated manage transformer substation. Currently, most meter automatic reading methods rely on preacquired high quality image, in which meter targets are big in size and locate in the middle and surface of meter is parallel with the camera. This needs lots of prior meter measurement and camera calibration, which fails to meet the requirements of actual use in transformer substation. In order to solve the problem mentioned, this paper presents a complete meter detection and recognition method. First, meter location within current visual field is obtained through a convolutional neural network model. Then the difference between target center location and camera visual field center location, as well as size percentage of target, are calculated. Camera state, including camera location and camera scaling factor, is adjusted according to the calculation result. After that, high quality image of meter target is acquired through perspective transform, which eliminates the image distortion caused by nonparallelism between meter and camera. Finally, locations of dial and pointer of the meter are obtained by conducting Hough Transform to the meter image, and meter reading is achieved. Results of actual experiments with transformer substation indicate that, maximum of reading error is as low as 1.82%. The proposed method can obtain accurate and stable performance with multiple kinds of meters in complicated background, which meets the demand of practical application in transformer substation.

Abstract:The monocular vision pedestrian detection can not obtain the depth information, so detection efficiency and accuracy are limited. Firstly, in the image preprocessing, the disparity information optimization analysis is proposed to simplify the expression of dynamic programming stixelworld in complex scenes based on stereo vision. Then, at the stage of pedestrian target detection, this paper analyzes the influence of the block scale on the detection effect in the traditional HOG feature, and obtains the multiHOG feature which is suitable for the road environment using the fisher criterion. The multiHOG feature is integrated with the LUV color channel feature. Finally, the hikSVM is used for pedestrian target classification. The experimental results show that the improved Stixelworld algorithm for image preprocessing greatly reduces the computation time and reduces the candidate region of pedestrian detection, target detection algorithm based on feature fusion and hikSVM has good realtime and robustness under the premise of guaranteeing the detection accuracy.

Abstract:In order to further improve the positioning accuracy of AGV (Automated Guided Vehicle) and meet the requirements of highprecision positioning occasions, a precise positioning method based on binocular vision and laser integrated navigation is proposed in this paper. Firstly, the method proposes an AGV position and pose control technique based on binocular vision realtime measurement. The multiobject vision system is used to recognize a number of circular identification points on the guide wire side, and the forward looking prediction and accurate pose determination are completed. Then, the improved least squares fitting ellipse algorithm based on algebraic distance variance calibration is used to determine the coordinates of the identification centers. Combining laser scanning and visual positioning information, the Unscented Kalman Filter (UKF) algorithm is adopted to realize the multiple sensor data fusion; and finally, the accurate positioning is realized. Experiment results show that after using the proposed method, the positioning accuracy is improved obviously, the control curve becomes smoother, the positioning robustness becomes better; the attitude adjustment accuracy can reach to ±0.5 degrees and the comprehensive parking positioning accuracy can reach to ±1 mm.

Abstract:Aimed at the problems of dynamic target tracking failure in the situation of fast motion and target occlusion, a spatiotemporal context tracking algorithm is proposed based on key points. Firstly, the key points of the target are extracted, and the predicted location of the object is obtained by combining key points matching with optical flow tracking. Then, the relationship model between key points change rate and spatiotemporal context model updating rate is established to control the update rate in realtime. In this way, the introduction of erroneous information can be prevented. Finally, a local context appearance model is constructed in the predicted location region, and the correlation between the spatiotemporal context model and the local context appearance model is computed to obtain the confidence map. Furthermore, the target is located accurately. The algorithm is validated in the test video, the highest average tracking success rate is 60% and the minimum average center error is 26.14 pixel. Compared to the 4 types of current major algorithms, the comprehensive performance of the proposed algorithm is superior to other methods, whose average tracking success rate is 90%, average center point error is 7.47 pixel and the average tracking rate is 25.31 frames per second. In the case of background interference, occlusion, target rotation and rapid motion, the mobile robot with binocular vision is used to track the random moving target. The success rate is 97.4%, and the average relative error of tracking distance is 4.05%.

Abstract:An improved monocular simultaneous localization and mapping method is proposed to solve a series of problems existing in the conventional monocular SLAM system, which is based on the classical EKF filter and FAST corners. To reduce the state estimation error resulting from deviation of expansion point when linearization, the cameracentered iterated EKF is applied to monocular SLAM system, which can minimize linearized error by iterative updating and representing all feature locations in the current camera frame. For robustness and efficiency of tracking features, and a homogeneous distribution of feature points, ORB features, which have the property of fast detection and matching, and invariance to scale and rotation, are selected as the feature points. Moreover, cell division method through detection to selection is applied. And the utilization of inverse depth parameterization for point features can efficiently avoid the problem of scalable ambiguity when map initialization. In addition, 1point RANSAC approach can ensure the stability and precision of filter by eliminating the wrong feature matching. The performance of system is evaluated by the ground truth. Experiments show that this new method is more robust and precise in comparison with the other monocular SLAM solutions, and meets the realtime processing requirements. The absolute trajectory positioning precision increases to 2.24 m and the mean error over the trajectory increases to 1.3%. To sum up, the proposed method is a practical solution to monocular localization and mapping.

Abstract:In order to study the biomedical effects in the treatment of tumors exposed to high pulse repetition frequency (PRF) nanosecond pulsed electric fields, a compact pulse generator which can provide high PRF ns pulses is needed. A new configuration of pulse generator combining Blumlein line and transmission line transformer (TLT) is proposed in this paper. The modular generator can provide high PRF ns rectangular pulses whose amplitude and PRF are adjustable with less switches. In this paper, the pulse forming processes of singlestage and multistage generator in matching impedance are theoretically analyzed. The changes of waveforms of multistage generator in mismatched loads are analyzed. Microstrip line is used to design the Blumlein in the generator. And the design processes of the Blumlein and TLT are introduced. The control strategy of solid state switches is also illustrated. A small prototype is developed to verify the proposed topology. Finally, the threestage generator can produce pulses on matched load （300 Ω） with amplitude of 0~3 kV, pulse width of 100 ns, and PRF to 10 kHz.

Abstract:The torque ripple of brushless DC motor (BLDCM) generated during the commutation process seriously limits the applications in high precision system, such as aviation and aerospace. By analyzing the cause of the torque ripple, a modified singleended primary inductor converter (SEPIC) topology of BLDCM drive system is presented, in which a capacitor in parallel with a diode circuit is embedded. Compared to classical SEPIC converters, high static gain and low switching voltage stress can be achieved. Meanwhile, the voltage is adjusted according to the speed of the motor during noncommutation period. As a result, the back electromotive force (EMF) is proportional to the DC input voltage, thereby, the torque ripple is suppressed. The experimental results show that the proposed topology can be effectively applied to BLDCM drive system, and significantly reduce the commutation torque ripple under different operating speeds.