Search

When generative adversarial networks (GAN) have just become popular, many GAN models have room for expansion. Firstly, this paper introduced document text localization generation adversarial network (TL-GAN), a deep neural network used to perform text localization from digital images. Furthermore, this paper conducts corresponding research on the TL-GAN algorithm, innovating the application of TL-GAN from two aspects. By analyzing the TL-GAN framework, it can generate any image. At the same time, based on its fuzzy characteristic axis, the internal logic algorithm of TL-GAN is studied through the analysis of the results.

At present, how to use low-cost and superior algorithms to obtain high-resolution 3D range image is the focus of lidar research. In this letter, the low-resolution Gm-APD lidar is combined with the high-resolution ICCD lidar to obtain the registered low-resolution range image and high-resolution intensity image. This letter proposes an improved image guidance algorithm. The algorithm uses a Markov random field model to define a global energy function. This function combines the distance fidelity term and the regularization term to obtain a high-resolution 3D range image by solving the optimization model. The experimental results show that compared with the traditional algorithms, the algorithm improves the resolution of the range images, the edge of the reconstructed image is sharper than the regional similarity guidance algorithm, and the image quality evaluation index has the better value.

Recognition and analytics at the edge enable utility companies to predict and prevent problems in real time. Clearing the voltage sag disturbance source by the positioning method is the most effective way to solve and improve the voltage sag. However, for different grid structures and fault types, the existing methods usually achieve a sag source location based on the single feature of monitoring data extraction. However, due to the effectiveness and applicability of the existing method features, this paper proposes a multidimensional feature of the voltage sag source positioning method of the matrix. Based on the analysis of the characteristics of the voltage sag event caused by the fault, this paper proposes a multidimensional feature matrix for voltage sag source location, based on the convolutional neural network to establish the mapping relationship between the feature matrix and the voltage sag position, thus achieving multiple points based on multiple points. The voltage sag source orientation is identified by the monitoring data. Finally, the voltage sag event caused by the short-circuit fault is simulated in the IEEE14 node model, and the effectiveness of the proposed method is verified by simulation data. The simulation results show that the proposed method has higher accuracy than the traditional method, and the method can be applied to different grid structures and different types of faults.

In an optical phased array (OPA) free space laser communication system, one of the main issues is to control the phase of each beam to acquire fast and stable phase-locking for phase consistency, in which the number of the elements in the array is of vital importance. In this paper, we studied the method on improving the performance of phase-locking about speed and stability of optical phased array with large number of elements. The performance of OPA with different number of emitting elements was analyzed with the simplified stochastic parallel gradient descent (SPGD) algorithm for closed-loop phase-locking. And the algorithm showed a poor performance when the elements of the array were in large amounts. An optimal algorithm, the distributed SPGD (DSPGD) algorithm, was proposed to improve the performance of the array, which revealed fast speed and stability in phase-locking of a large number of emitting elements of optical phased array, suggesting the algorithm is effective.

Aiming at the problem of difficulty in feature extraction and identification of wake echo signals detected by underwater lidar due to instability, an underwater bubbles recognition algorithm based on PCA feature extraction and elastic BP neural network was proposed. First, slice preprocessing was carried out on the echo signals collected continuously. Then the PCA algorithm was used to extract the main features of the spliced high-dimensional samples to determine the number of feature values. Then the parameters of the elastic BP neural network was selected to determine the number of hidden layer node and the number of features that can achieve optimal classification. Finally, an indoor wake detection simulation platform was used to realize the identification of bubbles and interfering targets. The experimental results show that when the hidden node is 12, the increment factor is 1.15, and the decrement factor is 0.55, two eigenvalues can be selected to classify the bubbles, non-bubbles and interfering targets. The recognition rate increases by 13.4% with the increase of bubbles density. At low density, the average recognition rate increases by 6.3% with the increase of laser energy. The recognition rate first increases and then decreases with the increase of distance. When the bubbles distance is 2.2 m, the target peak characteristics are obvious, and the average recognition rate is improved by 3.5%. Compared with adaptive and additional momentum BP, this method can reduce recognition time and achieve 99.1% accuracy. It is proved that this algorithm can be effectively and widely used in the recognition of bubbles in the ships wake by lidar.

The peak-picking method which is commonly used in Gm-APD laser radar 3D reconstruction always gets the wrong target position when there is an abnormal peak, and the reconstructed image has low signal-to-noise ratio and target missing because the threshold can only be integer. To solve these problems, a weighted Gaussian-like matched filtering algorithm was proposed. Fitting the echo firing histogram and normalizing can get the weight. Then the weighted window smoothing histogram was used and the peak position was selected again for reconstruction. According to the Poisson distribution of Gm-APD, the detection probability and false-alarm probability expression of the algorithm can be obtained, then compared with the peak method. The result show that the weighted Gaussian-like matched filter algorithm is better for the target in the middle of the gate. The theoretical derivation results are verified by Monte Carlo simulation. At last, by using the real experimental data and reconstructing data with two kinds of algorithms, the consequence shows that the weighted Gaussian-like matched filtering algorithm has a significant improvement on the restoration subjective and objective compared with the peak method. The results show that this algorithm has a good practical application prospect in dealing with low SNR and real-time 3D reconstruction.

Precision guided weapon is the main weapon in modern warfare. The precision guided weapon with photoelectric guidance has the advantages of strong resolution and high accuracy. It is widely used in the fields of sea-to-sea, air defense, anti-missile, space attack and defense. As optoelectronic guided weapons play an increasingly prominent role in modern warfare, their corresponding countermeasures are also constantly developing, which makes the combat environment more complex and requires them to continuously enhance adaptability to the increasingly complex battlefield environment. Based on the summary of the application of existing optical detection technologies such as infrared, visible light and laser semi-active in guided weapons, the shortcomings of the current technology in complex combat environments are analyzed, such as harsh natural weather, complex background and human interference. In view of the current development of multi-spectral detection, polarization imaging, laser three-dimensional imaging, quantum detection and other new technologies, the current research progress and results are introduced. It mainly includes multi-spectral multi-segment segmentation against true and false target camouflage recognition, polarization imaging against rain and fog, bad natural weather, laser three-dimensional imaging in the application of anti-complex background and penetration shielding and other research advantages and research progress. Finally, the problems in the application of the new detection technology and the key technologies that need to be broken through are systematically analyzed, and it is suggested that the guidance application of the new detection technology should be promoted through the joint attack of the basic technology and the application technology.

The method to realize the integration of laser communication and ranging is proposed in this paper. In the transmitter of two places, the ranging codes with uniqueness, good autocorrelation and cross-correlation properties are embed in the communication data and the encoded with the communication data to realize serial communication. And then the encoded data are modulated and send to each other, which can realize high speed two one-way laser communication. At the receiver, we can get the received ranging code after the demodulation, decoding and clock recovery. The received ranging codes and the local ranging codes do the autocorrelation to get a roughly range, while the phase difference between the local clock and the recovery clock to achieve the precision of the distance.

Laser beam's acquisition, pointing and tracking are crucial technologies of free space optical communication. Fine tracking system is an important component of APT (acquisition, pointing and tracking) system. Traditional fine tracking system use CCD or quadrant detector as the position detector of signal light. In order to simplify the system and improve accuracy, we propose a fine tracking system based on fiber nutation which don’t need a position detector and theoretically prove that the system is feasible. Meanwhile corresponding fine track system was built, experiment on position detection of the fine tracking system is done.

The out-of-plane rotation invariance is demonstrated for a recognition system for laser radar (ladar) range imagery. The key of our method is to transform an original ladar range image of scene data into a height-range image. We demonstrate experimentally that our method provides the out-of-plane rotation invariance property and preserves in-plane rotation invariant characteristics at the same time. We also show experimentally the relationship between the recognition rate and depression angles.

A method called local point density approach is presented for solving the difficulty of radius selection for point-based data integration in 3D model reconstruction by Ladar (laser radar) range images. After data integration, a set of points is properly selected for point density computing by setting the radius with different values. An optimal radius can be obtained when the variance of the point density is minimum. The algorithm is easy to adopt. The experimental results demonstrate the efficiency of the proposed approach.

Lidar intensity image is the foundation for precision target identification and attacking. How to enhance lidar intensity image is an important subject. Applied multriesolution contrast modulation method to 4-D image(range image and intensity image) fusion to enhance the intensity image. The experiments showed that the algorithm achieved a good image enhancement result. Contrast with intensity image without fusion, the outline of the targets in the image after fused was much more clear, and further improved arrangement also there was more efficient information.