Search

At present, the existing laser line extraction algorithms need to be in good lighting conditions to get better results. However, in practical application, due to the uneven illumination of the scene or the inconsistent reflection degree of the scanned object surface, the laser line extraction results will be greatly reduced. In some conditions, in order to improve the scanning efficiency, the grid laser line is used, but the grid laser is used Lines usually cause uneven distribution of laser lines. This paper proposes a method of extracting grid laser lines under the condition of non-uniform reflection. In the first step, the influence of global light can be eliminated using the difference between global illumination intensity and its first moment, and the image blocks can be classified using the average illumination intensity of local area to obtain the Region of interest (ROI). In the second step, a constrained region can be calculated by using edge gradient variance of local area and edge gradient maximization. In the third step, the center coordinates of sub-pixel laser line are finally obtained by combining the constrained region and illumination intensity map with Steger algorithm. The method used in this paper needs less empirical parameters to be set manually, and has a wider range of application. The accuracy of extraction is consistent with that of Steger algorithm. From the experimental results of this paper, we can see that we get more effective data, lower error rate and stronger robustness.

Fixed-point attack on the key parts of small aerial vehicles is one of the important means of UAV (Unmanned Aerial Vehicle) countermeasure. Fixed-wing aircraft flying in the air has two characteristics: fast speed and fast attitude change of aircraft. However, the traditional detection method of key parts of fixed-wing aircraft in infrared images is not only slow in speed but also low in accuracy. This paper presents an improved target detection and motion tracking algorithm based on deep learning method. The algorithm uses a deeper neural network to enhance the feature extraction capabilities of CNN and obtain richer feature information. The detection algorithm obtains the position and motion direction of the key points of the target. Experimental results show that the detection algorithm proposed in this paper is up to 30 frame/s, with an average accuracy of 91.5%/. The method achieved good results in the experiment.

This paper presents a method for the identification and location of circular markers in 3D reconstruction. The input image is processed with gray scale and gaussian filter to remove noise interference and improve the accuracy of edge detection. The size and direction of the gradient of the image were calculated to obtain the edge features of the circular marker, and the circle degree of the closed edge was judged, and the center of the marker was obtained by fitting the edge features with the ellipse of the least square method. Hu moment similarity of each marker was obtained, and the non-marker was deleted and the real marker was retained by similarity judgment. Experiments show that the algorithm has good robustness, high precision, can effectively reduce the probability of false detection, and has high accuracy in complex environments.

In this work we study the hierarchical organization of technical documents, where given a set of documents and a hierarchy of categories, the goal is to assign documents to their corresponding categories. Unlike prior work on supervised hierarchical document categorization that relies on large amount of labeled training data, which is expensive to obtain in closed technical domain and tends to stale as new knowledge emerges, we study this problem in a weak supervision setting, by leveraging semantic information from concepts. The core idea is to project both documents and categories into a common concept embedding space, where their fine-grained similarity can be easily and effectively computed. Experiments over real-world datasets from the subject of computer science, physics & mathematics, and medicine demonstrated the superior performance of our approach over a wide range of state of the art baseline approaches.

High-precision time-of-flight (ToF) measurement is the key to flow rate measurement of ultrasonic gas flowmeters. At present, the measurement accuracy of cross-correlation method, which is one of ToF measurement methods, depends on the high-speed ADCs. However, these high-speed ADCs, which have high cost, cannot be applied to low-power embedded systems and meet the practical requirement for actual industrial applications. To improve the accuracy and resolution of difference of ToF(dToF) measurement under the premise of satisfying the low-power, this paper proposes a novel signal processing method based on cross-correlation and interpolation. The echo signal is sampled by ADC in low-power embedded system, and the cross-correlation calculation is performed after digital filtering. Cubic spline interpolation is applied to improve the accuracy and resolution. The effectiveness of the proposed cross-correlation combined with interpolation method was verified and discussed. Experiment results show that this proposed method can significantly improve the resolution of dToF and the accuracy of ultrasonic gas flow rate measurement. This method provides a new tool and illuminates a good potential for real-time and high-precision flow rate measurement in low-power embedded systems.

Traditional gas compressibility factor estimation methods such as AGA8-92DC and SGERG-88 usually use overly complex theoretical derivation and corresponding estimation model. This will cost most of the operating memory of the low-power gas flowmeter. Therefore, the previous models are not suitable for application on the flowmeter using the low-power embedded chips. To solve this problem, this paper proposed a novel efficient soft computing model for natural gas compressibility factor based on Group Method of Data Handling(GMDH) neural network. First, the signal of working conditions such as temperature, pressure and gas mole fraction of components are used to calculate pseudo-critical pressure and pseudo-critical temperature. Second, the soft computing model based on GMDH neural network with Corrected Akaike’s Information Criterion (AICc) is utilized by using pseudo-critical pressure and pseudo-critical temperature as training sets. For the four common natural gas types, the estimated results show that the mean absolute percentage error is only 0.0168% and the computing time is effectively reduced. It also proved that the GMDH neural network can significantly reduce the computing time and improve the accuracy of the compressibility factor. Feasibility and effectiveness of this model was verified. Our work provides a very useful way and also make it possible to real-timely estimate the natural gas compressibility factor in low-power flowmeter under the premise of satisfying the accuracy.

This paper proposes a novel mixed-integer linear programming (MILP) model for distribution systems to respond to contingencies by forming multiple microgrids (MGs) energized by dispatchable distributed generators (DGs). The MILP model can model the outage propagation during faults, and minimize the load shedding while satisfying the radial topology and other operation constraints by controlling automatic switches and dispatchable DG outputs, thus, minimizing the outage areas. Case studies on a modified IEEE 123-bus test feeder are implemented to illustrate the effectiveness of the proposed MILP model.

Experimental studies reveal that the stator eigen-frequency is increasing along with the growing prestress load. However, the inherent mechanism is obscure and the equivalent physical model is not investigated especially. With introducing the external spring stiffness, the physical model involving the constraint boundary is developed and the corresponding finite element model is built. It is obvious that the computational results present the consistent variation trend with the measured data and optimize the traditional finite element modeling method. The proposed method is effective to calculate the frequency domain characteristics at the free or prestress case, which is used to optimize the structure or determine the driving frequency for traveling wave rotary ultrasonic motors.

Convolutional Neural Network (CNN) is one of the most promising deep learning technique that has achieved great success in many areas. In this paper, we implement a CNN for real-time monitoring of long-term voltage stability margin (VSM). To clarify our motivations of introducing CNN in this problem, we first discuss the essence and the complicity of VSM prediction, and summarize the limitations of existing methods, then point out the input structure of CNN inherently contains the topology information of power network, thus could have great potential in solving the problem. An CNN architecture, together with an input encoding method that strengthens topology information, is proposed and tested on IEEE 30-bus system. Preliminary results show that it can achieve better prediction performance comparing to some existing methods, and can be successfully employed in online voltage stability monitoring.

The realization of automatic frequency tracking is of great significance in ultrasonic transducer due to its wide application area in nowadays industrial manufacture. However, excessive overshoot current, serious overheating and long response time remain the urgent issue. To tackle these existing ultrasonic frequency tracking problem, this paper proposes a hybrid technique that integrates fuzzy control theory with PID control on the system. The technique involves coarse adjustment by fuzzy control and accurate adjustment by PID. Then, direct digital synthesizer (DDS) is used to generate required PWM wave with high accuracy. Initial driving frequency of the system and its corresponding phase difference of the feedback voltage and current can both be acquired during sweep frequency test before the tracking process, which corresponds to the parallel resonant point where the current is the minimum. This phase difference is also regarded as the target tracking value. During the tracking process, fuzzy control will be adopted when phase difference value error between feedback value and target value is greater than 10%, or PID will be used. Current difference AI and change rate AI/Af are used as two input values while driving frequency is the output value in fuzzy control when coarse adjustment is needed. Incremental PID arithmetic is used when accurate adjustment is needed. The result shows that the technique can reduce overshoot current of the ultrasonic transducer by approximately 8%, shorten the time for achieving stabilization of the system by 12%. In conclusion, this novel hybrid technique can quickly and stably track the frequency of parallel resonant point of the ultrasonic transducer system, meanwhile keep the whole system operating more efficiency and stable.

A novel approach for real-time monitoring of long-term voltage stability using local linear regression and adaptive training database is proposed in this paper. Comparing to previous methods, the proposed local predictive regression model could properly balance the simplicity/transparency and accuracy, and it is also adaptive to the changes of system and operating conditions. The approach is tested on IEEE 30-bus system. Preliminary results show that it can be successfully employed in online voltage stability monitoring.