24 results about "Automatic target recognition" patented technology

The invention belongs to the field of radar automatic target recognition, and relates to an SAR target recognition method and device based on projection features and improved self-learning. The method comprises the following steps: respectively processing each sample parameter of a training sample by a classifier T respectively trained by each prototype set to obtain integrated projection features of the sample parameters; initializing an SVM classifier based on the labeled sample feature set; taking out part of samples from the unlabeled sample feature set, and constructing a temporary sample pool; classifying the samples in the temporary sample pool by using an SVM classifier; evaluating the classification confidence of the samples in the temporary sample pool by using the classification posterior probability; adding a plurality of samples with high confidence into the labeled sample feature set, and retraining the SVM classifier; repeating for multiple times until an optimal SVM classifier is obtained; and target identification is carried out based on the optimal SVM classifier. The target recognition performance under the condition of a small number of labeled samples and a large number of unlabeled samples is effectively improved.

The application discloses a kind of SAR image's adversarial sample generation method and application, belong to synthetic aperture radar automatic target recognition security technical field;The present application is based on the average value of using integral integrated gradient, by randomly selecting part of model in integrated model multiple times, and calculating the difference of the temporary adversarial sample gradient generated in inner and outer loop, to correct the outer loop integrated gradient, so that its update direction is more accurate, can generate high-quality adversarial samples in real black box scene, and then can improve the success rate of attack in black box attack, and then affect the accuracy of security detection, in addition, it can also be used to reinforce depth learning model, such as in the process of training model, by adding the generated adversarial sample in data set can further construct better model with robustness, and then can effectively defend the attack of adversarial sample.

The present application relates to the technical field of rescue vehicles, in particular to a visual recognition intelligent mobile vehicle and a use method, comprising: a mobile bearing assembly, comprising a vehicle chassis frame provided with four Macadam wheels, and a top chassis frame installed above the vehicle chassis frame, used to provide a mobile and bearing basis; a temporary storage assembly installed on the top chassis frame, used to bear and collect clamped objects; a clamping jaw part, comprising a mechanical arm rotatably installed on the top chassis frame, and a clamping jaw unit installed at the front end of the mechanical arm and used to clamp target objects; and a visual assembly installed at the front end of the top chassis frame, used to collect images for visual recognition; the device can integrate visual recognition, mechanical arm grabbing and temporary storage functions, realize automatic target recognition, accurate grabbing and safe transfer, and overcome the defects of single function, low operation efficiency and dependence on manual control of existing equipment.

The invention relates to a recognition method with an automatic target recognition function, a recognition camera and a measurement device, and the recognition method comprises the steps: sequentially transmitting a first light beam and a second light beam which is different from the first light beam, an imaging element is used for receiving the first light beam and the second light beam reflected by the reflecting object and capturing a first image matched with the first light beam and a second image matched with the second light beam, and the first light beam and the second light beam are emitted by a target light source; the first image comprises a first light spot formed on the imaging element by a first light beam reflected by the target, and the second image comprises a second light spot formed on the imaging element by a second light beam reflected by the target; acquiring a third image based on the feature difference between the first image and the second image; and acquiring at least part of the first light spot and / or at least part of the second light spot as a target light spot based on the third image. Therefore, the recognition precision and the recognition efficiency of the target light spot can be improved.

The invention discloses an unmanned aerial vehicle inspection target automatic identification method using active learning. The method comprises the following steps: obtaining a high-quality inspection image sample set; outputting unified high-dimensional feature representation, and constructing an initial fine-grained recognition model on the basis of the unified high-dimensional feature representation; generating a generative adversarial active learning sample candidate set; accurate category labels are obtained, and an expert labeling key sample set is formed; boundary enhancement samples are obtained and generated, key sample sets are marked with experts, and a sample enhancement pool is established; updating to obtain an iterative fine-grained recognition model; performing defect detection and sub-category identification on the unmanned aerial vehicle inspection original multi-source image data collected in real time, and outputting an identification result and a confidence score. According to the method, the misjudgment probability and the missed detection probability between similar defects are effectively reduced in the actual unmanned aerial vehicle inspection reasoning stage.

Embodiments of the present application disclose a sample labeling method and device, electronic equipment and computer readable storage medium. Compared with the existing automatic target identification method, the method has the ability to evaluate the prediction reliability of unknown new samples in an open environment, can timely discover prediction error samples and unknown new samples, and update the identification model through an efficient labeling update mechanism. Through the present application, the technical problems of low identification reliability and update efficiency of the existing automatic target identification method for incremental samples are solved, and the technical effects of improving the identification reliability of incremental samples and the update efficiency of unknown new samples in an open environment are achieved.

The invention relates to calibration equipment and method for a servo total station, and belongs to the field of intelligent robots. The method comprises the following steps: S1, fixing a servo total station on one side of the top of a first tripod, and fixing a prism on one side of the top of a second tripod; s2, leveling an electronic bubble of the servo total station through a leveling module; s3, switching the servo total station to a positive mirror state through a positive mirror compensation module; s4, acquiring positive lens parameters and compensating the positive lens parameters; s5, switching the servo total station to an inverted mirror state through an inverted mirror compensation module; s6, acquiring mirror reversing parameters and compensating the mirror reversing parameters; s7, checking whether the electronic bubble is stable or not through a positive and negative mirror automatic target identification test module, and if the electronic bubble is stable, completing the calibration process; and if the electronic bubble is unstable, returning to S2. According to the method, the problem that the automatic target identification sighting range is inconsistent with the eyepiece view field range due to the fact that errors exist in motor rotation in a normal and reverse mirror calibration method is solved.

The invention discloses a polarized SAR image generation method for automatic target identification, and relates to the technical field of polarized synthetic aperture radars. The method comprises the following steps: acquiring two actually measured polarimetric SAR images with different azimuth angles and a polarimetric SAR image with a middle azimuth angle; inputting the actually measured azimuth angle and the polarized SAR image into a low-dimensional manifold sparse perception feature encoder to obtain the GAS feature of the azimuth angle; inputting the GAS features of the azimuth angles, the two actually measured polarized SAR images of different azimuth angles into a residual topology generator based on a generative adversarial network to obtain the polarized SAR images of the generated azimuth angles; in the training process of the residual topology generator, an actually measured polarized SAR image of a middle azimuth angle is used as a training reference image, and a multi-stage constraint loss function is adopted for control training. According to the invention, high-resolution and physically consistent polarized SAR image generation can be realized.

The utility model discloses a U-shaped rotary table three-light-in-one optical axis fine tuning device, which belongs to the field of automatic target identification and tracking of electronic equipment and comprises a mounting plate, one side of the mounting plate is fixedly connected with a fixing plate, the top of the fixing plate is provided with four springs, the four springs are distributed at four corners of the top of the fixing plate, and the four springs are arranged on the mounting plate. The top ends of the four springs are connected with the bottom of the same movable plate. According to the light-in-one optical axis fine adjustment device for the U-shaped rotary table 3, through the arrangement of the fixed plate, the spring, the movable plate and the ball head connector, the visible light camera main body inclines and displaces under the guidance of the middle ball head connector through the movable plate provided with the visible light camera main body, so that the light-in-one optical axis fine adjustment of the U-shaped rotary table 3 is realized; and the visible light camera main body can realize translation and angle change in the up-down and left-right directions, so that the optical axes of the infrared camera main body, the laser range finder main body and the visible light camera main body can be conveniently and quickly intersected at an infinite distance, and the processing efficiency of the electronic equipment is further improved.

The invention relates to the technical field of computer vision, and discloses a deep neural network model design method for infrared and visible light image fusion automatic target recognition in a complex environment, which comprises the following steps: S1, data enhancement fusion operation; step S2, constructing an attention convolution module; s3, constructing a self-adaptive feature extraction network; step S4, constructing a multi-scale fusion prediction network; step S5, constructing a multi-modal fusion module; s6, designing a multi-task loss function and a dynamic non-maximum suppression operation; according to the method, the problem that the recognition effect is poor due to the fact that a precise guided weapon is generally affected by insufficient multi-modal feature extraction and fusion strategies, complex low-illumination scenes, dense scenes and large model sizes when carrying out a multi-modal image fusion automatic target recognition task in a complex battlefield environment is solved; and an existing precise guided weapon is poor in infrared and visible light multi-mode image automatic target identification effect.

This invention discloses a multimodal fusion automatic target recognition system and method for total stations. It acquires three types of raw data—infrared images, visible light images, and laser echo point clouds—of a target scene in parallel. From the preprocessed infrared, visible light, and laser point clouds, it extracts unique features of the prism compared to other objects in the target scene, generating candidate target masks for each, and initially delineating the candidate target region. It then evaluates the confidence levels of the infrared, visible light, and laser modes under the current environment and dynamically assigns fusion weights to each mode based on these confidence levels. The candidate target masks are projected onto a unified space and then weighted and fused to generate a fusion response map, extracting a candidate target set. A lightweight convolutional neural network is used to classify and discriminate each candidate target in the candidate target set to eliminate interference, and Kalman filtering is combined for spatiotemporal consistency tracking verification, ultimately determining the unique true measuring prism.

The invention relates to an HRRP automatic target identification method based on data physical joint driving. Comprising the following steps: firstly, carrying out denoising and normalization preprocessing on obtained original HRRP data to obtain standardized HRRP input data; inputting the standardized HRRP input data into a data driving branch and a physical driving branch in parallel to obtain a data feature vector and a physical feature vector respectively, and reconstructing loss constraint physical consistency through a reconstruction module; then, inputting the data feature vector and the physical feature vector into a feature fusion module based on an attention mechanism to obtain a fusion feature vector, and inputting the fusion feature vector into a classification layer to output a target category probability; in the training stage, a multi-task loss function is adopted to carry out end-to-end optimization, so that the recognition precision, robustness and physical interpretability are considered under the complex conditions of low signal-to-noise ratio, view angle change and the like. According to the method, the interpretability of the model decision process can be effectively improved, and the accuracy and robustness of HRRP target recognition and the model interpretability are remarkably improved.

The application relates to the technical field of total station instruments, in particular to an automatic target recognition total station capable of optimizing engineering measurement processes and a use method thereof. The automatic target recognition total station comprises a measurement module, a protective shell with an inner cavity accommodating the measurement module, an entrance formed on one side of the protective shell, the measurement module being arranged at the entrance and being vertically retractable, and a total station stand comprising a central shaft, supporting feet and connecting rods, the central shaft being vertically and slidingly connected to the protective shell, the central shaft being fixed with the measurement module, the supporting feet being arranged around the central shaft and being rotatably connected to the protective shell, and the connecting rods being connected between the supporting feet and the central shaft. The folding or unfolding of the total station stand and the putting on or taking off of the protective shell on the measurement module can be simultaneously completed, so that the measurement process is simplified, and the measurement efficiency is improved.

The invention belongs to the technical field of radar automatic target recognition. The invention provides a radar micro-motion target identification method based on a semi-supervised support vector machine. According to the embodiment of the invention, spectrum data of a known target type and spectrum data of an unknown target type are obtained, and features are extracted from the data to obtain corresponding marked samples and unmarked samples; the weights of the features of the marked samples are calculated through a Fisher criterion, the features with the large weights are retained, and the features with the small weights are removed; performing same retention and elimination on the features of the unmarked samples; and training the features of the marked samples and the unmarked samples through SVM to train a semi-supervised support vector machine S3VM capable of stably performing a recognition task. The semi-supervised support vector machine S3VM can realize engineering application on a narrow-band radar and can identify a single-frame target spectrum signal, the accumulation time is short, the processing time is short, and an identification task can be carried out and completed while a target is detected.

The application provides a synthetic aperture radar image adversarial sample generation method and device, relates to the technical field of synthetic aperture radar automatic target recognition safety, and comprises the following steps: acquiring an image sample and initial parameters; transforming frequency components of the image sample, constructing a masking matrix through the size of the frequency components; predicting the image sample, updating the adversarial sample through a prediction result, and obtaining a current prediction sample; performing frequency domain feature conversion on the current prediction sample, dynamically adjusting frequency domain features through the masking matrix, respectively calculating gradients, and obtaining integrated gradients; updating the momentum gradient according to the integrated gradient, updating the current prediction sample through the updated momentum gradient, and outputting a final adversarial sample when the number of iterations exceeds a preset total number. The application solves the problem of systematic deviation of gradient estimation values.

This invention discloses a SAR image aircraft target recognition method based on simulated image semantic enhancement, belonging to the field of synthetic aperture radar image processing and automatic target recognition technology. The method includes constructing a simulated SAR image dataset with angle annotation information, training an angle semantic extraction network, constructing an angle semantic embedding encoder and a multimodal semantic fusion network, using the trained network to generate pseudo-angle semantic features for unlabeled measured SAR images, and fusing the measured image features with the pseudo-angle semantic features to output the aircraft target category recognition result. This invention, by explicitly modeling angle semantic modalities and performing multimodal feature fusion, reduces the angle sensitivity of SAR image aircraft target recognition, improves recognition accuracy and generalization ability under multi-angle conditions, and solves the problems of scarce measured angle samples and large differences between the simulated and measured domains.

The invention relates to a synthetic aperture radar automatic target identification method based on intelligent perception with a body, and the method comprises the steps: enabling an unmanned plane to interact with an environment at a moment t = 1, directly determining a state description at the moment t = 1 through a trained image encoder, inputting the state description into a trained motion decision network, and obtaining a new observation azimuth angle of the unmanned plane at the moment t = 2; at the moment t = 2-T, the unmanned aerial vehicle interacts with the environment after moving to a new observation azimuth angle, and the state description at the moment t is calculated by using the trained image encoder and the state description at the moment t-1 based on the maximum pooling operator and is input into the trained action decision network to obtain the new observation azimuth angle of the unmanned aerial vehicle at the moment t + 1. And inputting the T moment state description into the trained DST framework-based target classification model to obtain a final target recognition result. According to the method, recognition performance degradation caused by sample distribution difference between a new environment and a training environment can be quickly relieved, and recognition precision and reliability are improved.

A system is configured to receive images of a target located down range, identify the target in the images and classify the target to determine the location of the scoring regions. The system is further configured to determine the likelihood of an actual impact from a projectile on the target and score an actual projectile impact. The system uses machine vision and a machine learning model to classify the targets and determine impact scores.

The application discloses a multi-view radar image ATR optimal observation path planning method, which is applied to the technical field of radar imaging and aims at the problem that the radar imaging platform of the prior art is difficult to explore in an unknown environment. According to actual task requirements, a geometric model and an optimization mathematical model are built. Based on a basic architecture of a deep convolutional neural network, a multi-view network architecture integrated by multiple base classifiers is constructed in combination with the characteristics of multi-view input, so that automatic target recognition of multi-view radar images is realized. Based on an artificial potential field path planning algorithm, the best viewpoint selection of the radar image ATR is realized in combination with the characteristics of the flight of the radar imaging platform in the unknown environment. The best path planning of the flight of the radar imaging platform is realized, and the constraint condition of the optimal recognition performance is also met.

A holographic display system for attaching to a firearm provides a user with enhanced target acquisition information, such as real time ballistic solutions, fused thermal imaging, extended zoom, and automatic target recognition. The holographic display system may include a waveguide, a light engine, holographic optical element, a casing and a coupler. The holographic optical element is positioned between the light engine and the waveguide and the coupler is attached to the casing, The light engine is configured to generate and transmit displayable information to the holographic optical element, and the holographic optical element includes a diffractive grating configured to guide the displayable information into the waveguide. The waveguide is configured to propagate the displayable information and transmit the displayable information as a distant image to a user of the holographic display system.

The application discloses a polarimetric SAR image generation method for automatic target recognition, and relates to the technical field of polarimetric synthetic aperture radar. The method acquires a measured polarimetric SAR image of two different azimuth angles and a polarimetric SAR image of an intermediate azimuth angle; the measured polarimetric SAR image of the azimuth angle is input into a low-dimensional manifold sparse sensing feature encoder to obtain a GAS feature of the azimuth angle; the GAS feature of the azimuth angle and the measured polarimetric SAR image of the two different azimuth angles are input into a residual topology generator based on a generative adversarial network to obtain a generated polarimetric SAR image of the azimuth angle; and in the training process of the residual topology generator, the measured polarimetric SAR image of the intermediate azimuth angle is taken as a training reference image, and a multi-level constraint loss function is used for control training. The application can realize high-resolution and physically consistent polarimetric SAR image generation.

The system is configured to receive images of targets located downrange, identify the targets in the images, classify the targets, and determine the location of a scoring region. The system is further configured to determine the likelihood of an actual hit from a projectile on the target and score the actual projectile hit. The system uses machine vision and machine learning models to classify the targets and determine the hit score.

A method includes receiving, from an image sensor, an image, identifying, by a first neural network, a plurality of locations-of-interest within the image, and generating, by the first neural network, a first classification label for each location-of-interest of the plurality of locations-of-interest. The method also includes extracting, from the image, a plurality of image chips derived from the plurality of locations-of-interest and generating, by a second neural network, a second classification label for each image chip of the plurality of image chips. The method further includes determining an identification of a set of targets within the image using the plurality of locations-of-interest, the first classification label for each location-of-interest of the plurality of locations-of-interest, the plurality of image chips, and the second classification label for each image chip of the plurality of image chips, and transmitting the identification of the set of targets within the image.

This invention discloses an automatic target recognition method and related equipment based on sonar images. The method includes: acquiring sonar monitoring images of a target area as raw images; preprocessing the raw images to convert them into target-sized images; extracting features from the target images using multi-layer convolution and residual structures to obtain multi-scale feature maps; inputting the multi-scale feature maps into a preset target recognition model for processing to obtain target recognition results at different scales, and then filtering to obtain the final recognition result. This invention constructs a complete and automated process from sonar image acquisition to final target recognition, systematically integrating advanced deep learning architectures with sonar image characteristics. It comprehensively improves the accuracy, robustness, and practicality of target recognition algorithms in complex underwater environments, providing a feasible path to solve the problems of "poor universality and difficulty in real-time application" in existing technologies. It can be widely applied in the field of image processing technology.