Air target recognition in real-world scenarios has become an important part of the military offensive and defensive systems of various countries. By identifying aerial targets captured by image acquisition equipment and utilizing the obtained information to achieve effective identification of friend or foe, identifying enemy sources, combat capabilities, and intentions, important references are provided for tactical decision-making. With the continuous development of military technology, traditional manual based recognition methods are no longer capable of identifying aerial targets. The role of deep learning related algorithms, which gradually replace traditional image processing algorithms, in the field of target recognition is becoming increasingly prominent. This article will use deep learning methods to study aerial target recognition. Provide a detailed description of the design and implementation process of the Faster R-CNN training model, including model structure, network layers, and feature selection. This model mainly includes a Region Proposal Network (RPN) and a set of convolutional neural networks for extracting target features and predicting target bounding boxes. Build a basic environment for training models, and establish a dataset of air targets related to military equipment. The dataset samples are annotated according to the VOC2007 dataset format. Train the model using a dataset, conduct testing and detection after training, and finally analyze the calculation results of Eval evaluation indicators to further optimize the algorithm model and improve recognition rate. The experiment verified the effectiveness and feasibility of the Faster R-CNN training model, and applied it to aerial target recognition tasks. The experimental results show that the model can quickly and accurately identify aerial targets.
KEYWORDS: Target recognition, Hyperspectral imaging, Education and training, Detection and tracking algorithms, Deep learning, Target detection, Image processing, Data modeling, Small targets, Hyperspectral target detection
Hyperspectral image target recognition, as an important component of computer vision, has broad application prospects in the military field. We conducted research on hyperspectral image target recognition technology based on deep learning, with typical military targets as the main recognition objects. A target recognition algorithm suitable for hyperspectral images is proposed based on the YOLO network model. To address the issue of the YOLO algorithm not being sensitive enough to small targets, the calculation formula for the loss function in the algorithm has been adjusted, and a multi-scale training method has been added. In response to the phenomenon of missed detections when the YOLO algorithm faces multi-objective clusters, five anchor boxes are added to each grid. In addition, the network convergence speed is accelerated by adding a BN layer; Introducing residual networks effectively prevents network degradation; Adding 1*1 convolutional layer can better extract image detail features and improve network recognition accuracy. The hyperspectral image target dataset is obtained by using hyperspectral imaging instruments to shoot multiple targets from multiple angles and directions under different lighting conditions, and labeling them. Divide the data into training and testing sets for training. The trained model was used for recognition experiments on vehicles in hyperspectral images. The experimental results showed that the model met the real-time requirements for target recognition in hyperspectral images, and performed well in the recognition performance of small targets at long distances.
Hyperspectral image refers to a three-dimensional data cube image containing both spatial shape information and spectral reflectance information of ground objects. In recent years, with the rapid development of land-based platforms and near ground flight carriers, the production and application of land-based hyperspectral imaging equipment are gradually mature. Compared with hyperspectral images under remote sensing imaging conditions, hyperspectral images under land-based imaging conditions have higher spatial resolution and temporal resolution, saving a lot of time and use costs. The land-based hyperspectral imaging technology provides a solution to the problem of "similar shape is the same object" existing in the current two-dimensional space target detection. Therefore, a method of target location and recognition based on joint spatial and spectral information is proposed in this paper. This paper first introduces the basic principles of space target detection and spectral target detection, then establishes the spatial dataset of land-based hyperspectral images, and carries out experiments to verify the proposed methods. Use the ground imaging spectrometer to obtain the hyperspectral image data of the area to be measured, and then use the spatial target detection model to frame the specific target; The similarity test between the specific target set out in the frame and the prior spectral information shows that the matching degree between the pixel in the target area to be measured and the target spectrum is 88.48%, and the matching degree between the pixel in the non target area and the target spectrum is only 3.12%, thus successfully completing the target positioning and recognition task. The experiment shows that the method of specific target location and recognition based on spatial and spectral information can solve the problems of simple spectral dimension detection and simple spatial dimension detection at the same time, which is of great significance in specific target location, identification of true and false targets, and provides a new idea for specific target location and recognition in the future.
The spectral information data of ground objects refers to the relationship between spectral reflectance and wavelength. At present, the field imaging spectrometer is mainly used to obtain the image and spectral information of objects at the same time. However, the spectral reflectance of the same object in different directions is different, which seriously affects the accuracy of subsequent classification and target detection based on spectral data. In order to solve this problem, a method of spectral data expansion of ground objects based on semi empirical kernel driven model is proposed in this paper. A small amount of spectral data of ground objects under the condition of known directions are substituted into the model, and the spectral data under the condition of other arbitrary directions are inverted, which not only reduces the cost of sample collection, but also expands the spectral data of ground objects. Experiments prove the effectiveness of this spectral data expansion method and use the expanded spectral data as a priori sample for ground object classification. Compared with the classification method based on a small number of original spectral samples, the experiments show that this method can effectively improve the accuracy of ground object classification.
Because of its imaging characteristics, large field of view infrared images have low resolution and few high-frequency details. And it is difficult to obtain high-resolution large field of view infrared images as training sample database. Therefore, some super-resolution reconstruction algorithms that achieve better results in visible images may not be suitable for large field of view infrared images. Based on Convolutional Auto-encoders, a shallow Convolutional Auto-encoders which consists of four convolution layers, one maxpooling layer and one upsampling layer was constructed. And deep Convolutional Auto-encoders composed of 10 convolution layers, 2 maxpooling layers and 2 upsampling layers was provided. shallow Residual Convolutional Auto-encoders and deep Residual Convolutional Auto-encoders also were concluded. In order to make up for the shortcomings of large field of view infrared images, this paper adopts the combination of some large field of view infrared images and ordinary field of view infrared images as the image training library, and uses the large field of view infrared image data as the test set. The reconstructed image quality and the objective indexes of MSE, PNSR and SSIM show that the reconstruction effect of the shallow Residual Convolutional Auto-encoders is slightly better than that of the deep Residual Convolutional Auto-encoders, while the shallow s Convolutional Auto-encoders and the deep Convolutional Auto-encoders without residual model are not good. However, the reconstruction effect of deep Convolutional Auto-encoders is the worst due to the loss of more shallow information.
Facing the large amount of optical reconnaissance images, how to improve the reconnaissance efficiency and accuracy has become an urgent problem to be solved. In our designed scheme, the joint transform correlator is utilized to achieve the aim of recognizing military target in dual band. By monitoring the distribution of correlation points from correlation output results, our designed scheme can judge whether there are targets in visible band and infrared band reconnaissance images. This recognition method can not only improve the recognition ability for camouflaged targets, but also greatly improve the target recognition speed with the help of the recognition speed advantage of the joint transform correlator. Theoretical analysis and simulation experiments both verify the feasibility of this method.
Ranging distance was one of the most important index of laser rangefinders. However, it could be affected by the atmospheric turbulence. Effect on maximum distance critical value of laser rangefinders with atmospheric turbulence was discussed. Firstly, the atmospheric structure constant of refractive index (C2n) of specific area was monitoring and the variable regularity was analyzed. With the assumption that the target was the small target, the power of small target was calculated. Then, the receiving power of detector was obtained. Lastly, the maximum distance critical value with the same visibility and different turbulence intensity was got. When the refractive index (C2n) was 5×10-15m-2/3, 5×10-14 m-2/3, 5×10-13 m-2/3 and 5×10-12 m-2/3, the maximum distance critical value was 1900m, 1850m, 1850-1810m and 1810m. It could be generalized that the maximum distance critical value of laser rangefinders could be affected by the atmospheric turbulence.
Laser is used for remote charging of UAV instead of sunlight to improve charging power, continuous charging performance and endurance of UAV. Laser alignment is one of the key technologies for laser charging. On the basis of briefly describing the composition and working process of the laser charging system, the method, module composition, working process and key process based on beacon laser alignment are emphatically analyzed. Combined with the whole working process of the laser charging system, the error factors affecting the charging accuracy are analyzed, and the reasonable error distribution is carried out to provide theoretical basis for the system realization.
KEYWORDS: Camouflage, Infrared radiation, Thermography, Metals, Semiconductors, Data processing, Temperature metrology, Temperature sensors, Infrared imaging, Control systems
In order to improve the target's antagonism to the infrared reconnaissance and recognition, an infrared dynamic camouflage array based on the electro-thermal material is designed. The overall design scheme and the structure of the regional temperature control drive module are described. The main dynamic factors of external environment and internal controllable factors of temperature control are analyzed. Through processing, assembly and debugging, the camouflage array prototype is completed, and the controllable dynamic camouflage effect is achieved. Then the temperature control accuracy and radiation uniformity are tested, and the error is analyzed.
Dynamic infrared camouflage can need to change according to the background, and quickly from a state of camouflage into another state of camouflage. The image characteristics of various infrared camouflage states weak correlation, can make the infrared surveillance and tracking, target infrared characteristics of the real difficulty of the guidance system, unable to complete the target reconnaissance and combat, so as to improve the battlefield survivability of the target. According to different background conditions, infrared camouflage algorithm is adopted to realize infrared camouflage for equipment, so that equipment and background can be integrated. The evaluation of camouflage effect needs to be carried out through corresponding methods. Based on the principles of evaluation index selection, and given to evaluation object, evaluation purpose and evaluation method, constructs the evaluation index system, comprehensive selection of brightness contrast and color features, texture features three evaluation indexes. Combined with the specific test, the infrared camouflage effect was evaluated, the indicators were analyzed, and all the indicators were comprehensively analyzed according to the weight value. Finally, the optimal state for the comprehensive evaluation of the camouflage effect was given.
Judging from the information war in recent years, the future battlefield UAV will play a huge role. The length of the endurance has become one of the main factors limiting its performance. Using laser beam to charge the flying UAV in real time is a powerful means to greatly extend its endurance and improve its operational efficiency. Accurate target tracking of the charging panels on the aerial flying UAV is a key technology for the subsequent laser beam charging. Aimed at the target tracking module of UAV laser charging system, hardware and software are designed respectively. The structure of each part of the hardware system and the related parameters are analyzed. The software design process and key algorithms are described. In this paper, the target tracking aiming test is carried out by combining the charging panel of the UAV, and various error factors in the process of target tracking aiming are analyzed.
Using laser features of high brightness and good directivity, can undertake directional transmission of energy, laser charging technology transforms the energy supply way with its own carrying to charge at any time. The key to laser alignment is to identify the target of the charging panel and locate the center. Through the control of the servo module, the battery charging plate can be adjusted to the center of the field of view to complete laser alignment charging. Object recognition and center location algorithm are mainly divided into three steps: image preprocessing, rectangle panel identification and center positioning. Through this, the center positioning is realized, and the center positioning of the panel charged by laser is carried out for rotorcraft with different distances and angles, so as to achieve a better effect.
According to the working principle of the binocular photoelectric instrument optical axis parallelism digital calibration instrument, and in view of all components of the instrument, the various factors affect the system precision is analyzed, and then precision analysis model is established. Based on the error distribution, Monte Carlo method is used to analyze the relationship between the comprehensive error and the change of the center coordinate of the circle target image. The method can further guide the error distribution, optimize control the factors which have greater influence on the comprehensive error, and improve the measurement accuracy of the optical axis parallelism digital calibration instrument.
Low-light level night vision device and thermal infrared imaging binocular photoelectric instrument are used widely. The maladjustment of binocular instrument ocular axises parallelism will cause the observer the symptom such as dizziness, nausea, when use for a long time. Binocular photoelectric equipment digital calibration instrument is developed for detecting ocular axises parallelism. And the quantitative value of optical axis deviation can be quantitatively measured. As a testing instrument, the precision must be much higher than the standard of test instrument. Analyzes the factors that influence the accuracy of detection. Factors exist in each testing process link which affect the precision of the detecting instrument. They can be divided into two categories, one category is factors which directly affect the position of reticle image, the other category is factors which affect the calculation the center of reticle image. And the Synthesize error is calculated out. And further distribute the errors reasonably to ensure the accuracy of calibration instruments.
A variety of environmental detecting and perception of equipment are equipped in high integrated systems such as unmanned vehicle, unmanned aerial vehicles, and intelligent robot and so on. The equipment include laser radar, television observation device, infrared observation device and low light level observation device and so on. In order to ensure the effective fusion of multi-spectral information of multiple devices, all the optical axis of different spectrum equipment must be parallel, or the initial zero of the optical axis must be parallel. The locations of different spectral equipment in system are different, some optical axis are far apart. To achieve the goal of all-weather, arbitrary geographical conditions detecting multi-optical axis parallelism of different spectral, multi-spectral multi-axis parallel calibration system is designed. In the calibration scheme, the infrared axis is used as the first benchmark axis, and adjusts the multi-spectral parallel light pipe to parallel to the infrared axis. Then multi-spectral parallel light pipe is used as second benchmark axis to detect the parallelism of light axis of TV, low light level observation device, and laser radar. In the process of system design, on the premise of guarantee accuracy, the indicators of the parts are fully demonstrated, and the maneuverability and adjustability of the system are considered. Through the steps of design, processing, assembly and debugging, the multi-spectral multi-axis calibration system is realized. After by testing the multi-optical axis parallelism in integrated system, it is reflected that the multi-spectral multi-axis calibration system has the characteristics of reasonable design, easy to operate. And it is able to achieve rapid and high resolution multi-spectral multi-axis parallel calibration under all-weather and arbitrary geographical environment.
The maladjustment of photoelectric instrument binocular optical axis parallelism will affect the observe effect directly. A binocular optical axis parallelism digital calibration system is designed. On the basis of the principle of optical axis binocular photoelectric instrument calibration, the scheme of system is designed, and the binocular optical axis parallelism digital calibration system is realized, which include four modules: multiband parallel light tube, optical axis translation, image acquisition system and software system. According to the different characteristics of thermal infrared imager and low-light-level night viewer, different algorithms is used to localize the center of the cross reticle. And the binocular optical axis parallelism calibration is realized for calibrating low-light-level night viewer and thermal infrared imager.
As one of the main weapons, impulse laser rangefinders have become the main object of the electro-optical countermeasures. So its real maximum range (defined as utmost operating range in the paper) becomes the most concerned index to evaluate the performance of electro-optical countermeasure weapons. A method for calculating laser rangefinders′ utmost operating range by its sensitivity in different weather is obtained. Then a method by experiment for getting the sensitivity is supplied. By analyzing the experiment data which the detectivity is 40%-60%, the laser rangefinders′ sensitivity is in the range of 1.7×10-5 W to 9.8×10-5 W. For the reason that in order to get an exact utmost operating range, the experiment accuracy of sensitivity is very important, in the last part of paper, the factors which influence the experiment accuracy of sensitivity are analyzed, such as circuit of automatic gain control, the fluctuation of laser power, incident angle of laser.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.