3556results about How to "Improve clarity" patented technology

The invention provides an infrared and visible light image fusion method based on multi-scale feature interaction enhancement, belongs to the field of image processing, and aims to improve the image fusion quality in a complex environment. The method comprises the following steps: firstly, performing feature extraction on infrared and visible light images through a double-flow multi-scale encoder to generate multi-level features of different scales; and then, cross-modal information interaction is carried out by adopting a feature interaction enhancement module, fusion features are optimized through a space and channel attention mechanism, and clear expression of key areas such as target heat radiation and image edges is ensured. In the decoding stage, multi-scale jump connection and layer-by-layer up-sampling are utilized to reconstruct a high-resolution fusion image. And finally, training the fusion network through a content-gradient joint loss function to realize balance between detail reservation and global information. The method is suitable for tasks such as monitoring and target identification in a complex scene, and can effectively improve the image fusion effect especially under the conditions of low illumination, smoke shielding and the like.

The invention discloses a photovoltaic cleaning robot based on automatic path planning and an operation method thereof, relates to the technical field of robots, and solves the problem of cleaning blind spots in an existing cleaning robot and the endless loop risk of a path planning algorithm. According to the invention, the surface information of the photovoltaic panel is accurately captured through the multispectral acquisition module, a high-resolution and high-definition comprehensive image is generated through the data fusion processing module, then dirt is accurately identified on different scales by using a multi-scale pyramid structure and an adaptive feature fusion mechanism, and an attention mechanism and environmental perception data are combined, so that the real-time performance of the system is improved. The recognition threshold is dynamically adjusted, and misrecognition and missing detection are reduced; and finally, enabling the robot to autonomously learn in a complex environment and optimize an obstacle avoidance strategy through a dynamic obstacle avoidance algorithm based on deep reinforcement learning. The intelligent level and the cleaning effect of an existing photovoltaic cleaning robot are greatly improved.

The invention relates to the technical field of medical data processing, in particular to an advanced nasopharynx cancer treatment effect prediction system based on deep learning. According to the invention, a data acquisition module acquires multi-modal medical images and clinical data of a patient from a hospital information system; the data preprocessing module is used for cleaning, registering, normalizing and standardizing the data; the feature extraction module extracts image features by using a convolutional neural network and fuses the image features with clinical features to form a comprehensive feature vector; the model construction and training module constructs a deep learning model, and is adjusted and optimized through training; the treatment prediction module inputs the comprehensive feature vector to obtain a treatment effect prediction result; and the result evaluation and feedback module compares prediction and actual results, calculates indexes and collects and feeds back the improved model. The system comprehensively integrates data, accurately mines features, constructs an effective model, provides personalized prediction to assist clinical decision, continuously improves the model performance, and has important significance in improving the prediction accuracy of the treatment effect of advanced nasopharynx cancer, promoting personalized medical treatment and the like.

The invention provides a braid quality detection system and method based on industrial camera detection, and relates to the technical field of data processing. A multi-angle acquisition module; an image fusion module; an optical correction module; a preprocessing module; a preliminary defect detection module; a fine defect analysis module; a self-adaptive filtering module; the dynamic coordinate mapping module is used for mapping the physical position of the defect to obtain actual coordinate data; the data storage and analysis module is used for storing and analyzing historical actual coordinate data and related detection data and generating a quality report and a trend analysis report; the alarm and control module is used for automatically sending out an alarm signal and adjusting production parameters or pausing production when the defect number or the trend change rate in the trend analysis report reaches a preset threshold value; according to the invention, the accuracy and efficiency of braid quality detection are improved.

The invention relates to the technical field of industrial defect detection, in particular to a wire harness defect detection method and system based on visual identification. The method comprises the following steps: acquiring wire harness image data, and performing image segmentation according to the wire harness image data to obtain wire harness image segmentation data; performing morphological feature extraction according to the wiring harness image segmentation data to obtain wiring harness morphological feature data, and performing three-dimensional reconstruction according to the wiring harness morphological feature data to obtain wiring harness three-dimensional reconstruction data; performing microstructure defect detection according to the wiring harness morphological characteristic data to obtain first wiring harness defect detection and first wiring harness defect detection data, and performing microstructure defect detection on the wiring harness three-dimensional reconstruction data to obtain second wiring harness defect detection data; and marking the wire harness image data according to the first wire harness defect detection data and the second wire harness defect detection data to obtain wire harness image defect marking data. According to the method, the potential defect can be identified in the fine form change of the wire harness.

The invention discloses a cancer image denoising method and system based on a multi-scale neural network, and the method comprises the steps: carrying out the adaptive segmentation of a cancer image, carrying out the noise estimation and statistical analysis of each segmented region, and automatically adjusting the denoising parameters; performing multi-scale feature extraction by adopting wavelet transform and pyramid decomposition, and analyzing and processing the image after parameter adjustment; performing quality evaluation and distortion detection on the image after multi-scale processing, automatically identifying and positioning artifacts and distortion, and correcting and compensating interference by adopting a point spread function inversion and Wiener filtering method to obtain a corrected image; and after histogram equalization and contrast-limited adaptive histogram equalization are applied to the corrected image, denoising processing is carried out by using a convolutional neural network in combination with sparse representation and a dictionary learning algorithm, and a denoising result is obtained. According to the method, noise in different types of cancer images can be effectively removed, meanwhile, key diagnosis information is reserved, and the image quality and definition are improved.

The invention relates to the technical field of medical data processing, in particular to a risk assessment system for chronic diseases. The system is used for performing physiological index analysis according to a medical physiological monitoring signal to generate physiological index data; performing individual physiological index weight empowerment on the physiological index data based on genomics data to obtain individual physiological index data; performing quantitative analysis on pathological influence factors of chronic disease types according to the electronic health record data and the individual physiological index data, and generating individual chronic disease influence factor quantitative data; performing composite interaction risk factor analysis on the individual chronic disease influence factors based on the quantitative data of the individual chronic disease influence factors to generate individual chronic disease composite interaction risk factor data; and performing individual chronic disease risk assessment based on the individual chronic disease composite interaction risk factor data to generate individual chronic disease risk assessment data. According to the invention, chronic disease health management and accurate evaluation are realized through collaborative evaluation work of chronic disease risks.

The invention provides a methane emission remote sensing intelligent detection method and system based on sparse spectrum reconstruction. The method comprises the following steps: performing first-order Taylor expansion on a Beer-Lambert law; introducing a spatial continuity constraint and a physical diffusion constraint term based on a wind field, and deducing an optimization objective function for inversion of atmospheric methane increment; simulating remote sensing observation hyperspectral radiation brightness values under different methane increment concentrations, and calculating to obtain a unit absorption characteristic spectrum; carrying out dimension reduction processing and clustering operation on the remote sensing hyperspectral image; performing matrix decomposition and sparse reconstruction on the spectrum value according to a clustering result; the hyperspectral image obtained after matrix sparse decomposition reconstruction is equivalent to a background spectral image, and atmospheric methane increment is obtained through calculation; generating an atmospheric methane plume sample based on a large eddy simulation program; constructing a training sample set of methane emission rate machine learning estimation; constructing a machine learning model for estimating the atmospheric methane emission rate; and based on an atmospheric methane increment inversion result, completing methane point source emission rate estimation of a real scene.

The invention discloses a remote sensing surveying and mapping image enhancement processing method for an unmanned aerial vehicle, and relates to the technical field of image enhancement processing. During operation of a system, flight attitude and position information are recorded through GNSS and IMU equipment, the flight path of the unmanned aerial vehicle is controlled, and collected original images are primarily processed; calculating and obtaining a flight stability coefficient Fs, an image acquisition quality coefficient Iq and a topographic feature retention coefficient Tf, removing random noise in a remote sensing image by using an improved non-local mean NLM algorithm, optimizing and processing an edge region, improving edge definition and texture details of a topographic image through a Laplacian pyramid algorithm, strengthening landform features of mountains and hills, and obtaining a remote sensing image; according to the method, color distortion caused by complex mountain illumination conditions is corrected, image colors are enabled to truly restore topographic features, important landform information is prevented from being lost due to over-brightness or over-darkness, topographic feature extraction is optimized, a final index Zfinal is generated, and quality evaluation is performed on an enhanced and corrected image.

The invention relates to the field of image processing, in particular to a welding quality comprehensive evaluation method and system, and the method comprises the steps: obtaining an ultrasonic image of a welding joint, carrying out the contour detection, and extracting a defect region of the welding joint; gradient features and gray features of pixel points in the defect area are calculated, feature vectors of target pixel points in the defect area are constructed, and the feature vectors serve as the expression degree of the defect area; edge information of the defect area is obtained so as to determine defect types, the defect types comprise cracks and bubbles, the shape and size of a window are set according to the defect types, bilateral filtering processing is carried out, defect information is extracted, and comprehensive evaluation is carried out on the welding quality. According to the method, the appropriate window shape and size are selected according to the defect type, the detail information of the defect is effectively reserved, the definition of the ultrasonic image is remarkably enhanced through the self-adaptive filtering strategy, and therefore the welding quality is better evaluated.

The invention relates to the technical field of camera adjustment, in particular to a camera adaptive adjustment method and system and a storage medium. The method comprises the following steps: acquiring a multi-source environment sensing data set and an initial environment image; performing environment feature recognition on the initial environment image according to the multi-source environment sensing data set to obtain a scene environment feature map; performing semantic scene segmentation on the scene environment feature map to obtain a semantic scene macroscopic feature map; performing time sequence attention target tracking on the scene environment feature map to obtain a target vehicle motion feature map; performing feature constraint fusion on the semantic scene macroscopic feature map and the target vehicle motion feature map to obtain a target vehicle-scene motion semantic map; and performing parameter adjustment on the camera according to the target vehicle-scene motion semantic map to obtain a camera parameter adjustment strategy set. According to the invention, the performance and the image quality of the camera in a changeable environment are improved.

The invention relates to the technical field of image classification, in particular to a fetal heart monitoring image classification method and system based on a neural network. The method comprises the following steps: acquiring a fetal heart monitoring image, generating a standard monitoring image through Gaussian noise elimination and spatial domain normalization, then performing multi-channel convolution on the standard image to extract image convolution features, obtaining standard fetal heart features through compression regular optimization, performing morphological reconstruction to generate a fetal heart dynamic model, and finally obtaining the fetal heart dynamic model. Baseline features are identified based on historical data, then periodic variation quantification and safe floating range derivation are performed on fusion features, and finally a multi-scale discrimination framework is constructed to realize intelligent classification of fetal heart monitoring images. The fetal heart monitoring image classification method is more stable and more accurate.

The invention relates to a backside illuminated image sensor and a preparation method thereof, and electronic equipment, and relates to the technical field of integrated circuits, the backside illuminated image sensor comprises a substrate, and photodiodes and laminated target isolation structures which are located on a first surface of the substrate, are alternately arranged along a first direction parallel to the first surface, and extend towards the substrate; the substrate internally comprises a plurality of trench isolation structures which extend into the substrate through the first surface and are arranged at intervals in the first direction; the trench isolation structures and the laminated target isolation structures right above the trench isolation structures are arranged in a one-to-one manner; the laminated target isolation structure comprises isolation parts and insulation parts which are sequentially and alternately laminated along a second direction deviating from the substrate; the photodiode comprises at least two photosensitive layers which are sequentially stacked in the second direction, the longitudinal sections of the photosensitive layers are T-shaped, one photosensitive layer makes contact with the first surface of the substrate, substrate damage caused by ion implantation can be avoided, and meanwhile the photoelectric conversion efficiency and the quantum efficiency are improved.

The invention discloses a corrugated pipe inner wall multi-type flaw detection method based on deep learning, and the method comprises the following steps: S1, collecting an initial corrugated pipe inner wall image and corrugated pipe inner wall images in different illumination environments, and generating an original image; s2, preprocessing the original image to generate a preprocessed image; s3, extracting surface geometric overall features of the inner wall of the corrugated pipe from the preprocessed image through a pre-trained ResNet101 network; s4, inputting the preprocessed image and the surface geometric overall features into an improved StyleGAN3 network to generate an enhanced image; s5, inputting the enhanced image into the multi-scale adaptive deep network, and outputting a defect detection result; and S6, performing post-processing on the flaw detection result, and outputting a final flaw detection result. According to the method, deep learning and geometric characteristic embedding technologies are adopted, so that the defects of the inner wall of the corrugated pipe are accurately detected, and the method has the advantages of high precision, high robustness and comprehensive adaptability.

The invention discloses a visual detection system for an automobile door panel assembly, and relates to the technical field of automatic detection, and the visual detection system comprises the steps: installing a high-precision industrial camera, and carrying out the synchronous collection of the multi-view image data of an automobile door panel; preprocessing the acquired multi-view image data of the automobile door panel; carrying out detail enhancement on the preprocessed multi-view image data of the automobile door panel by adopting a super-resolution algorithm; integrating the enhanced multi-view image data of the automobile door panel into an automobile door panel panorama through binocular stereoscopic vision matching, and constructing an automobile door panel three-dimensional model; the automobile door panel three-dimensional model is compared with a standard automobile door panel three-dimensional model, and the buckle defect in the automobile door panel is recognized; and carrying out position marking and defect type analysis on the identified defect points, and generating an automobile door panel detection report according to the defect type analysis. According to the invention, details are enhanced by using a super-resolution algorithm, defects are accurately positioned by using a three-dimensional model, and the detection efficiency and accuracy are improved.

The invention relates to the technical field of wastewater treatment, in particular to a generation method for a chemical production wastewater treatment system control model. The method comprises the following steps: collecting chemical wastewater data by using an online monitoring instrument, wherein the chemical wastewater data comprises tin-containing wastewater data and etching liquid waste liquid data; determining a wastewater recovery control stage according to the chemical wastewater data; constructing a wastewater control model by utilizing chemical engineering simulation software based on the wastewater recovery control stage; inputting the chemical wastewater data into a wastewater control model, simulating a wastewater recovery control process by using simulation software to obtain wastewater recovery control data, and performing control abnormality judgment on the wastewater recovery control data to obtain abnormal wastewater data; the high-definition industrial camera is used for collecting images of the waste water sedimentation tank, and a centrifugal pump image is recognized and obtained from the images of the waste water sedimentation tank. According to the method, wastewater recovery control is optimized based on a wastewater treatment technology, and the wastewater recovery efficiency and the anomaly detection rate are remarkably improved.

The invention provides a virtual fitting method and device, multi-target personalized virtual fitting is realized through a multi-target and multi-mode virtual fitting model, and the virtual fitting device comprises a multi-mode feature interlaced learning attention module (MFILA) and a parallel denoising UNet module. Wherein the MFILA module realizes multi-modal condition feature coding through a multi-modal feature fusion module (MFF) and a multi-reference target texture feature extraction module (MRSTF), and the parallel denoising UNet module is mainly responsible for denoising generation. A user uploads a model clothes graph, a target fitting clothes style graph and corresponding text cues, and the system gradually generates a denoised virtual fitting result from pure noise through an MM-VTON model by taking the inputs as conditions. According to the method, the reality sense and the accuracy of virtual fitting are improved by utilizing an advanced image generation technology and multi-modal learning, meanwhile, the demand of computing resources is reduced, and the method has relatively high practical application value and a wide market prospect.

The invention relates to multi-mode signal processing, and provides a signal processing method of a computer system. The method comprises the following steps: acquiring different modal signal data by using a plurality of sensors and performing timestamp synchronization; a depth time sequence alignment network and a multi-layer dynamic time warping algorithm are utilized to ensure signal time consistency; multi-modal signal features are extracted and fused through a multi-stream depth feature extraction network and a cross-modal attention mechanism; the system adopts a dynamic convolution kernel generation and meta-learning technology to carry out adaptive signal processing; noise suppression and signal optimization are carried out by using a multi-mode noise reduction auto-encoder and a generative adversarial network, and parameters are optimized and processed through closed-loop feedback control and reinforcement learning strategies. According to the method, synchronization, fusion, adaptive processing, noise suppression and real-time feedback control of multi-mode signals are realized, and the method has the advantages of high precision, high robustness and real-time performance.

The invention relates to the field of image enhancement, in particular to an image enhancement method and system of a vehicle-mounted image sensor. The method comprises the following steps: acquiring driving record images of a vehicle at multiple angles; dynamic scene self-adaptive exposure adjustment is carried out on the driving record images of the vehicle at the multiple angles, so that multiple self-adaptive exposure driving images are obtained; carrying out deep network noise identification on the plurality of adaptive exposure driving images, and carrying out dynamic filtering and noise reduction on sub-images one by one so as to generate a plurality of filtering and noise reduction optimized images; carrying out image texture analysis on the plurality of exposure sub-images, and carrying out texture distortion reconstruction to construct a plurality of texture reconstruction images; and road object visual identification is carried out on the plurality of texture reconstruction images, and adaptive contrast enhancement processing is carried out, so that a plurality of detail contrast enhancement images are constructed. According to the invention, a high-definition and stable panoramic vehicle-mounted image is realized.

The invention belongs to the field of image processing, and particularly discloses an imaging simulation method and device based on a point spread function and electronic equipment, and the method comprises the steps: building a light source model corresponding to an image shot by an optical system, and obtaining the point spread function of a point light source of the light source model after the action of the optical system; obtaining error point spread functions under different optical system models; performing weighted averaging on each error point spread function according to a preset weight to obtain a comprehensive point spread function; forming an array diagram by the comprehensive point spread function according to light distribution, and carrying out convolution on the array diagram and a shot image to obtain a degraded image simulating a blurring effect caused by aberration; and applying the distortion model of the optical system to the shot image to obtain a distortion-simulated deformation effect image, and realizing imaging simulation of the optical system based on the deformation effect image and the degraded image. According to the invention, the imaging accuracy and efficiency can be improved.

The invention discloses a client data management method and system, and relates to the technical field of client data management, and the method comprises the steps: collecting client data from a plurality of data sources, carrying out the preprocessing, and extracting the time sequence features of the client data; constructing a customer loss model to calculate the loss probability of customers, classifying customer groups according to the loss probability of the customers, and implementing corresponding measures according to a classification result; and constructing a visual interface to display the result of the loss probability of the customer in real time, and storing, collecting and analyzing the generated customer data. The method comprises the following steps: acquiring customer data from multiple data sources, preprocessing the customer data, and extracting time sequence features of the customer data; according to the method, the customer loss model is constructed to calculate the loss probability of the customer, the customer loss prediction accuracy is improved, the definition and accuracy of data are improved, and the problem of processing unstructured and time-varying data in the prior art is solved.

The invention provides a large model and mapping knowledge domain fusion method, application method and system, and the method comprises the steps: carrying out the mapping knowledge domain embedding of an obtained mapping knowledge domain to generate mapping knowledge domain embedded information, and generating a natural language question based on the mapping knowledge domain embedded information; the knowledge graph embedded information and the natural language question sentences are input into an entity relation learning model, entity expressions and relation expressions corresponding to the natural language question sentences are output, then sub-graphs corresponding to the natural language question sentences are extracted based on the natural language question sentences, the entity expressions and the relation expressions, and the sub-graphs are converted into natural languages one by one. And combining the entity relationship learning model and the knowledge graph embedded information into the large model to generate a target reasoning model so as to realize fusion of the knowledge graph and the large model. Therefore, the model can extract key information and perform deep understanding and reasoning, the reasoning ability of the model in processing complex problems is enhanced, and the definition and interpretability of subsequent output contents of the model are further improved.

The invention discloses a PE pipe quality detection method and system based on a target image, and relates to the technical field of pipe quality detection. In order to solve the problems that multi-angle image data cannot be obtained, under the condition of complex surface textures or slight defects, false detection or missing detection possibly occurs, so that the detection accuracy is influenced, and the method cannot adapt to defect detection of various types of pipes. Through multi-angle image acquisition and preprocessing, the comprehensiveness and definition of image data are improved, false detection or missing detection caused by complex surface textures or slight defects is reduced, the adaptability of a detection system and extraction and matching of target features are enhanced, the key detection area of the PE pipe can be recognized more accurately, preset standards and key features can be compared, and the detection accuracy is improved. According to the method, the quality of the PE pipe is accurately evaluated, the detection accuracy and efficiency are improved, and an operator can quickly recognize problems through visual output and a detailed detection report.

The invention discloses a bridge welding quality detection method based on a gray scale-relative depth fitting visual technology, and relates to the technical field of bridge engineering quality detection.The bridge welding quality detection method comprises the following steps that S100, multi-dimensional image capturing is carried out, specifically, a high-resolution camera is used, and by means of illumination control and multi-angle shooting, a multi-dimensional image is obtained; according to the method, the image of the bridge welding part is comprehensively collected, rich original data is provided for subsequent detection, the high-resolution camera is used for capturing the multi-dimensional image, fine gray processing is combined, the image contrast and definition are effectively enhanced, key features of the welding part are highlighted, further, relative depth calculation is carried out by using the binocular vision principle, and the accuracy of the relative depth calculation is improved. The relative depth information of the welding part is accurately obtained, the problems that in a traditional method, the image quality is unstable, and the feature point extraction matching difficulty is large are solved, and the three-dimensional curved surface construction precision is remarkably improved.

The invention relates to the technical field of synchronous acquisition, in particular to a multi-camera synchronous data acquisition method and system, and the method comprises the following steps: based on a real-time video stream of a camera, analyzing a video frame, recognizing a moving object in a differentiated video stream, and carrying out the recognition of the moving object in the differentiated video stream; and synchronously recording the occurrence time and position change of the moving object in the differentiated camera to obtain view angle change data. According to the invention, the angle and focal length of the camera are synchronously adjusted to cope with the view angle change of different scenes, and the trigger time of the camera is uniformly configured and is synchronously started through the network, so that the time deviation is reduced, the analysis of the synchronous trigger data set and the timestamp adjustment are synchronously triggered, and the synchronization between the cameras is optimized; the consistency and synchronism of the video streams obtained from different angles are ensured, the data transmission parameters are dynamically adjusted according to the network condition, the efficiency and synchronization precision in the data transmission process are improved, and the influence of network delay and bandwidth change is effectively dealt with.

The invention discloses a rice disease and insect pest early warning method and system based on image recognition, and particularly relates to the technical field of insect pest recognition, different areas of a rice field are comprehensively covered through multi-source high-resolution image collection, it is ensured that the image quality reaches a preset threshold value, median filtering and histogram equalization are adopted in the image preprocessing step, and the image quality is improved. Based on a geographic information system and a Kriging interpolation method, a spatial heterogeneity model of rice field disease transmission is established, a deep convolutional neural network is used for carrying out feature extraction on the image, the type, position and severity of diseases and pests are accurately recognized, the extracted features are combined with the spatial heterogeneity model, and the accuracy of the disease and pest detection is improved. A spatial autocorrelation analysis method is adopted to deeply evaluate the distribution mode of diseases in the rice field, and a support vector machine algorithm is applied to perform classification evaluation on the disease and pest risk, so that the problems of early warning lag and missing report caused by disease spatial heterogeneity are effectively solved, and the accuracy and timeliness of early warning are improved.

The invention provides an image restoration method and system based on multiple attention mechanisms, and relates to the technical field of image processing, and the method comprises the steps: obtaining a to-be-processed image; extracting shallow layer features of the to-be-processed image through the convolutional layer; according to the shallow features, attention features are calculated through a triple attention module, the shallow features are processed through a feature aggregation module, and deep features are extracted; fusing the attention features, the deep features and the shallow features to obtain fused features; and performing image restoration on the to-be-processed image according to the fusion feature, and outputting the restored image. According to the method, effective feature fusion and selection can be carried out among information of different scales and different types, so that the quality and generalization ability of image restoration are improved.

The invention discloses a capillary image evaluation method, device and equipment and a medium, and the method comprises the steps: obtaining IQ data of a to-be-detected region, and carrying out the beam forming of the IQ data; performing clutter filtering processing on the IQ data of each continuous data frame after beam forming, and extracting a blood flow signal; removing background noise from the blood flow signal to obtain a blood vessel structure image, and segmenting the blood vessel structure image to obtain a distribution position of a blood vessel region; screening a peak region based on the distribution position of the blood vessel region, and determining a peak center position; trajectory tracking is carried out based on the peak value center position, and all trajectories are superposed and displayed to generate a capillary image; and analyzing the generated capillary image, and extracting related parameters of the capillary. Therefore, the method is not limited by image resolution, and can be applied to microvascular images with limited resolution and super-resolution microvascular images, and vascular morphology and hemodynamic parameters can be extracted from the microvascular images with limited resolution and super-resolution.

The invention relates to an unmanned aerial vehicle image high-efficiency identification system. The system comprises an unmanned aerial vehicle, an image acquisition unit, an image preprocessing unit, a target detection unit, a target tracking unit and a control unit. The unmanned aerial vehicle has a high-speed stable flight capability, and the image acquisition unit consists of a high-resolution camera and is used for capturing image data in real time in a flight process; the image preprocessing unit carries out denoising and deblurring processing on the captured image data to generate preprocessed image data; the target detection unit adopts a pre-trained deep learning model to realize efficient recognition of a target object in the preprocessed image; the target tracking unit is combined with a Kalman filter and an optical flow method to lock and track a target object; the control unit adjusts the flight path of the unmanned aerial vehicle according to the target object position and motion information provided by the target tracking unit to ensure continuous tracking of the target object; according to the system, efficient and accurate identification and tracking of a target object are realized through cooperative work of all the units.

The invention relates to the technical field of digital image processing, and provides an infrared and visible light image fusion method for improving image details and textures, and the method comprises the steps: splicing the feature vectors of an original infrared image and an original visible light image; performing initial feature map extraction of different spatial scales on the obtained source image features by adopting a plurality of serial residual parallel visual Mama network models; taking the initial feature map obtained by the last iteration as a basic feature map, and identifying fusion feature information of the basic feature map by adopting a multi-channel parallel visual Mama network model; expanding the fusion feature information to the initial feature maps of different spatial scales to obtain initial fusion feature maps of different spatial scales; performing up-sampling on the initial fusion feature map on each spatial scale to the resolution of the original infrared image and generating a fusion image; and performing iterative optimization according to the feature loss of the fused image and the original image. According to the invention, the detail definition and texture expression in the fused image can be obviously improved.