283 results about "Binary segmentation" patented technology

The invention belongs to the technical field of alkane gas leakage detection, and particularly relates to a leaked gas detection method based on deep learning and an optical flow method. The method comprises the following steps: simulating image data of real gas leakage by using a Gaussian plume model, and making a data set; a deep learning network YOLOv8-seg model is used for training; an intermediate infrared camera is used for monitoring an area where leakage may occur; when leakage occurs, a leakage gas area is detected and segmented, and a binary segmentation mask of leakage gas is generated; and analyzing the motion direction of the gas plume in the binary segmentation mask by using an optical flow method. The problem that leakage points are difficult to find in an existing gas detection means is solved, the diffusion track of leaked gas can be tracked more accurately, and rapid tracing is achieved.

The invention discloses a remote sensing image cultivated land segmentation method and system fusing context and boundary perception, and belongs to the technical field of remote sensing image processing and agricultural information. Constructing a cultivated land segmentation initial model composed of a backbone network, a feature enhancement module, a multi-scale feature fusion de-wharf module and a mask prediction module; training set data are input into the initial model, a composite loss function value is calculated, back propagation is executed, and a cultivated land segmentation model with boundary sensing ability is obtained through multi-round iterative optimization; and inputting the remote sensing image into the trained cultivated land segmentation model, and outputting a binary segmentation image representing the cultivated land position. Visual state space modeling and large receptive field convolution are combined, deep and shallow layer information is fused through feature injection, boundary perception supervision and composite loss are introduced, cultivated land boundary discrimination is improved, remote sensing image cultivated land high-precision extraction is achieved, and the method is suitable for agricultural interpretation and monitoring.

The invention discloses an SAM2 multi-task perception binary segmentation method based on a hybrid expert adapter. The method is specifically implemented according to the following steps: step 1, constructing a data set and an encoder; step 2, constructing a hybrid expert adapter module; and step 3, constructing a task awareness gating module. According to the method, a pre-trained SAM2 is taken as a main network, on the basis of freezing the main body weight, a standard adapter and a MoE-Adapter are respectively deployed on odd and even layers of an encoder, and a lightweight expert sub-network and a dynamic gating strategy are combined, so that unified processing of multiple tasks such as salient target detection, camouflage target identification, marine animal segmentation and the like is realized.

The invention relates to a multi-modal image threshold segmentation preprocessing method based on a convolutional neural network, and the method comprises the steps: unifying an image into a standard space, carrying out the pixel value mapping, carrying out the resampling, generating high and low frequency sub-bands, carrying out the soft threshold denoising of the high frequency sub-bands, enhancing the contrast of the low frequency sub-bands, and carrying out the fusion; an optimized VGGnet framework is constructed; a noise adversarial network is generated to carry out active learning loop training on a convolutional neural network model; the image is input into the model for prediction; a local entropy and a gradient magnitude are calculated based on a prediction result; optimal segmentation is realized by setting a double-layer matrix of a feature tag-segmentation method; grey matter Dice calculation is carried out on the segmented images, and preprocessing parameters of unqualified images are optimized through a dynamic parameter adjusting module based on a Gaussian process regression model. The segmentation precision and the processing efficiency of the multi-modal image are effectively improved, and the adaptability of the model to a complex image is enhanced.

The invention relates to the technical field of image analysis, in particular to a concrete arch bridge crack semantic segmentation method based on multispectral imaging, and the method comprises the following steps: collecting a multi-period multispectral image through an unmanned aerial vehicle, segmenting a crack region through a K mean value of the multispectral image, extracting a multiband spectral vector sequence, and carrying out the sliding normalization to generate a principal axis spectral vector; and calculating a spectral vector included angle and a change rate mark jump point, screening boundary points by combining direction consistency and gradient scoring, extracting spectral lines Z-score standardization by time sequence alignment, evaluating discrete fluctuation, correcting an abnormal output concrete arch bridge crack binary segmentation map. According to the method, the reflection spectrum sequence is constructed through the multi-period multi-spectral image, the crack recognition precision is improved by combining the main shaft features and the sliding window, the texture interference is eliminated by using the spectral vector included angle change and the gradient score, the standardized spectral line time sequence model is generated, the boundary positioning accuracy is enhanced, the noise interference is reduced, and the long-term monitoring of the structural damage is supported.

The embodiment of the invention provides a target segmentation method and device for a multi-view image, electronic equipment and a storage medium, and the method comprises the steps: obtaining scene images of different views obtained by a plurality of image collection devices which are deployed in a target scene in advance, and carrying out the three-dimensional Gaussian reconstruction, and obtaining a three-dimensional Gaussian point cloud corresponding to the target scene; in response to a segmentation prompt instruction corresponding to a target segmentation article in the scene articles, performing two-dimensional article segmentation on the scene image, and generating a corresponding binary segmentation mask; carrying out projection processing on the three-dimensional Gaussian point cloud to obtain projection planes of a plurality of visual angles, and carrying out Gaussian point filtering on each projection plane based on the binarization segmentation mask to obtain a filtered three-dimensional Gaussian point set; and based on the pose point cloud estimation information, the filtered three-dimensional Gaussian point set is sequentially splashed to device planes corresponding to a plurality of image acquisition devices to obtain a corresponding target segmentation result, so that the accuracy of object two-dimensional image segmentation in a target scene is greatly improved.

The invention provides a dermatoscope image segmentation method, and relates to the technical field of medical image processing, and the method comprises the steps: inputting a to-be-segmented image into an improved encoder module, carrying out the multi-scale feature information extraction, and obtaining a deep advanced semantic feature map; inputting the deep advanced semantic feature map into the bridging module, and performing context information modeling and cross-scale feature fusion to obtain a fused feature map; inputting the fused feature map into the decoder module, and carrying out layer-by-layer up-sampling and feature integration to obtain a reconstructed feature map; inputting the reconstructed feature map into the boundary sensing double-branch module, and performing collaborative feature processing; wherein the main branch outputs a binary segmentation mask, and the auxiliary branch is used for measuring a sign distance function diagram to provide boundary perception supervision; obtaining a binary segmentation mask output by the main branch as a segmentation result; compared with an existing model, the optimized dermatoscope image segmentation model is higher in recognition accuracy.

The invention discloses a large-size grinding wheel concentricity error measurement method and system based on machine vision, and relates to the technical field of grinding wheel concentricity error measurement, and the method comprises the steps: calibrating an industrial CCD camera through PyCharm software, collecting a part of images of a grinding wheel, and carrying out the RGB three-channel extraction, gray conversion, filtering and binary segmentation processing of the collected images; secondly, a Canny edge operator is adopted to carry out coarse positioning on the edge of the image, and then a novel edge judgment criterion is established to construct an edge detection model with sub-pixel-level precision; and finally, carrying out feature fusion on sub-pixel contours of RGB channels, clustering extracted discrete point sets by using an improved DBSCAN algorithm, and then carrying out ring fitting by using improved RANSAC to realize accurate measurement of the inner and outer diameters and concentricity of the grinding wheel.

The invention relates to the technical field of image processing, in particular to an automatic snow removal operation method and system based on image enhancement processing, and the method comprises the following steps: collecting an original accumulated snow image of a road surface, decomposing an image brightness component into a low-frequency component and a high-frequency component through a multi-scale Retinex algorithm, and generating an enhanced image; on the basis of the enhanced image, calculating a joint gradient of saturation and brightness in a pixel point HSV space, and generating a snow area binary segmentation image by adopting an improved Otsu dual-threshold method; the accumulated snow compactness grade is judged according to the texture uniformity; and driving the execution mechanism to spray the snow removing agent according to the path coordinates, and adjusting the spraying concentration of the snow removing agent in combination with the compactness grade. According to the method, the snow boundary definition and the plaque internal texture identifiability are improved, a high-quality image foundation is laid for subsequent segmentation and analysis, and the adaptive capacity and stability of the system under the natural light condition are enhanced.

The invention discloses a chip surface defect detection method, electronic equipment and a readable medium. The method comprises the following steps: aligning an ROI (Region of Interest) of a reference image with a test image through affine transformation; obtaining a gray average value of the ROI of the reference image and the ROI of the aligned test image, and carrying out gray stretching normalization processing on the test image based on the gray average value; respectively carrying out mean filtering on the reference image and the test image after normalization processing, and carrying out differential operation to obtain a differential image; performing threshold-based binary segmentation on the difference image, extracting surface detection candidate areas of bright defects and dark defects, performing morphological corrosion operation on the surface detection candidate areas to suppress noise, and generating a final defect mask; and mapping a defect area in the final defect mask back to the original test image coordinate system through inverse transformation of affine transformation. According to the method, high-precision smudginess defect extraction of the IPM chip area under the background of complex brightness can be realized, and the method is suitable for various complex packaging working conditions.

The invention discloses a low-altitude remote sensing ground feature element extraction method fused with a DSM side adapter. According to the method, visual features are extracted by taking a pre-training visual large model of frozen parameters as a trunk, and a trainable DSM side adapter is constructed in parallel to extract geometric features of elevation of a ground object target; a cross-modal attention fusion mechanism is utilized, visual features are used as queries, elevation geometric features are used as key values, height information injection is dynamically guided, and content-aware adaptive modal fusion is achieved. Besides, a parallel prompt decoding strategy is adopted, a virtual task batch is constructed by using a batch stacking technology, and multi-category semantic segmentation is reconstructed into a high-dimensional parallel prompt-driven binary segmentation task, so that the limitation of a large model native normal form is broken through on the premise of less structural modification, and the real-time performance of the system is improved. And end-to-end fine tuning is carried out by combining a channel mutual exclusion and competition mechanism. According to the method, the ground feature element extraction precision and the boundary integrity in a complex scene are remarkably improved, and the method has the advantages of low training cost, high generalization and flexible deployment.

The invention discloses a multi-attention mechanism network model and a thyroid ultrasound contrast focus segmentation method and system based on a multi-attention mechanism network. The multi-attention mechanism network model adopts an RFM-ResNeXt module to extract spatial and temporal feature information of an input video; aggregating the three-dimensional feature maps into a two-dimensional feature map through a bidirectional convolution LSTM module; inputting the two-dimensional feature map into a double attention module (DAM) to obtain an attention feature map; and carrying out convolution and up-sampling on the attention feature map to obtain a binary segmentation result image. The time-space information of the ultrasound contrast video can be effectively extracted by utilizing three attention mechanisms of channel attention, time attention and space attention, and the ablation area in the thyroid ultrasound contrast image can be automatically and accurately segmented.

The invention discloses a regional farmland water-saving irrigation and conventional irrigation classification and high-precision mapping method based on machine learning. The method comprises the following steps: S1, acquiring climate, geography, crops and remote sensing image information data related to a farmland in a research area; s2, selecting seven indexes including ELE, LS, ET, LAI, CI, DLST and LTH to represent water-saving irrigation (WI) and conventional irrigation (FI), performing binary segmentation by using a threshold method, obtaining classification of WI and FI under each index, and generating WI candidate samples; s3, WI and FI typical pixels are determined, and a training sample library is formed; s4, synthesizing a characteristic wave band as a classifier input characteristic; s5, based on the training sample library and classifier input features, training a classifier through a machine learning algorithm; and S6, applying the classifier to the characteristic wave band in the research area to obtain the classification and classification graph of the farmland WI and FI in the research area.

The invention relates to a high-precision crack segmentation method based on double-flow visual basic model collaboration, belongs to the technical field of computer vision and deep learning, and aims to solve the problems that an existing crack segmentation model is poor in generalization under a complex background, fine cracks are missed and disconnected under high resolution, and the fine tuning cost of a large model is high. The method comprises the following steps: firstly, preprocessing an acquired infrastructure surface image into standard input of 1024 * 1024 pixels, and extracting features by adopting a double-flow encoder containing parallel SAM3 and DINOv3 double branches; freezing double-bone intervention training parameters, and embedding a lightweight adapter to complete efficient fine tuning of the parameters; double-branch multi-scale features are extracted, and multi-scale splicing and dimension reduction fusion are completed after channel alignment; and outputting a pixel-level crack binary segmentation result through a lightweight decoder based on depth separable convolution. According to the method, segmentation precision, cross-domain generalization ability and training deployment cost are considered, and the engineering landing requirement of infrastructure health monitoring is met.

The invention discloses a discharge channel extraction method and system based on edge detection and texture feature fusion, and belongs to the technical field of power equipment fault diagnosis and digital image processing, and the method comprises the steps: obtaining an ultraviolet weak light image of power equipment, and carrying out the preprocessing; multi-scale edge detection is executed based on the preprocessed image, edge information is fused through a multi-scale voting mechanism, and a candidate edge graph is generated by combining direction consistency connection fracture edges; extracting texture features of the pre-processed image, wherein the texture features comprise a rotation invariant local binary pattern uniformity feature and a multi-direction gray level co-occurrence matrix feature; according to a dynamic distribution weight coefficient of a discharge type, carrying out weighted fusion on the edge intensity of the candidate edge graph and the texture features, and generating a discharge channel confidence score graph; and carrying out binarization segmentation and contour optimization processing on the confidence score graph of the discharge channel, and extracting morphology quantization parameters of the discharge channel.

The invention discloses an image fine structure intelligent detection algorithm based on a double-branch encoder. The method comprises the following steps: firstly, constructing an image segmentation system comprising a feature prior branch based on a lightweight pre-training model, a fine structure extraction branch based on a structure perception visual state space module, a cross-modal fusion module and a multi-scale feature complementary mapping decoder; secondly, respectively inputting the original image into a feature prior branch and a fine structure extraction branch of an image segmentation system for feature extraction and preliminary fusion, then performing second-stage fusion on features after preliminary fusion through a cross-modal fusion module, and finally inputting the features into a multi-scale feature complementary mapping decoder to obtain a binary segmentation image; through the design of a feature prior branch and a fine structure extraction branch, the generalization ability of the model in a few-sample scene is enhanced, and the long-distance dependence and complex irregular form of the fine structure are efficiently captured with lower calculation cost; and a two-stage cross-modal fusion mechanism improves the characterization capability of the model on fine structure features.

The invention relates to an automatic grinding track generation method, a medium and equipment. The automatic grinding track generation method comprises the following steps: acquiring a workpiece point cloud, uniformly segmenting the workpiece point cloud to obtain a plurality of point cloud modules, obtaining a mean square error of each point cloud module, combining the point cloud modules into a plane according to the mean square error to obtain a workpiece plane image, and carrying out binary segmentation on the workpiece plane image to obtain a grinding track. Carrying out gradient calculation on the workpiece binary image to obtain a plurality of feature points, sequentially connecting all the feature points according to a set sequence to obtain an initial grinding track, fitting a curve of the initial grinding track, and determining a final grinding track. The problems that in the prior art, workpiece recognition and positioning are difficult, track calculation precision is low, and tracks are not smooth enough are solved.

The invention discloses a double-student-teacher network semi-supervised crack segmentation method based on uncertain sensing, and the method comprises the steps: S, building a DSMT, and enabling a student model and a teacher model to be CrackUnet segmentation models based on an encoder-decoder structure; using labeled data to calculate supervision loss # imgabs0 # to execute cross pseudo-supervision training and consistency training on unlabeled data, calculating unsupervised loss # imgabs1 #, updating student model parameters according to total loss # imgabs2 # # imgabs3 #, and lambda (t) being a dynamic weight coefficient changing along with training iteration times t; performing dynamic uncertain patch adjustment (DUPA) to generate a new training sample; and retraining the network only by using a new sample, updating teacher model parameters as exponential moving average (EMA) of two student model parameters, outputting a segmentation result of the teacher model on the input crack image, and generating a binary segmentation mask.

The invention belongs to the technical field of water ecology risk early warning, and provides an algal bloom risk remote sensing intelligent identification method and system, and the method comprises the steps: obtaining remote sensing image data, meteorological data and water quality data of a target region, and carrying out the preprocessing; performing frequency domain feature extraction to generate a three-dimensional frequency domain feature vector; fusing the global feature representation obtained by modeling and the generated three-dimensional frequency domain feature vector by using a multi-task deep learning network to obtain a feature map; obtaining an algae bloom binary segmentation probability graph and a continuous value distribution graph of chlorophyll a concentration based on the characteristic graph; generating an image semantic embedding vector by utilizing a semantic embedding head, and performing semantic alignment on the image semantic embedding vector by adopting a pre-trained knowledge graph to generate a research and judgment information text; and performing cross validation on the study and judgment information text, the algae bloom binary segmentation probability graph and the continuous value distribution graph of the chlorophyll a concentration to obtain a final algae bloom risk judgment result. According to the invention, the algal bloom risk remote sensing intelligent identification is realized.

The invention discloses a radar target template signal establishment method based on depth model adaptive segmentation, mainly relates to the technical field of template signals, and is used for solving the problems that a target contour is easy to fracture or false detection by clutters is easy to cause when a time-frequency map is subjected to binary segmentation by a simple threshold slice in the prior art. And the threshold parameter is very sensitive to the signal-to-clutter ratio and the environmental change. Comprising the following steps: reading a one-dimensional radar echo sequence, and mapping the one-dimensional radar echo sequence to a two-dimensional time-frequency domain to obtain a time-frequency image; multi-scale features are obtained, and a matrix is prompted; obtaining a binary mask corresponding to the mask feature through the multi-scale feature and the prompt matrix; determining the binary mask corresponding to the highest confidence score as a final mask; screening time-frequency transformation data corresponding to the radar echo sequence by using the final mask to obtain output data; and recovering the output data into a time domain signal by using inverse short-time Fourier transform, and taking the time domain signal as a target template signal.

The invention provides a high-resolution image acquisition method, which can convert a low-resolution image and a label thereof into a high-resolution image and a label thereof by establishing a super-resolution model, the generated high-resolution image and the label thereof can be used for a subsequent segmentation task, and the high-resolution image and the label thereof can be acquired by pre-training a first model. The convergence effect and reconstruction performance of the super-resolution model on a specified task are improved, then a self-supervised learning strategy is designed, so that the super-resolution model only needs data in a 2D scanning mode to complete training, and simultaneous modeling of a gray level image and a binary mask image is proposed, so that the super-resolution model is obtained. Therefore, the super-resolution model can process the grayscale image and the binary mask image at the same time. The method has the advantages that the problem that an existing super-resolution model can only process a grayscale image and cannot directly process a binary segmentation mask image is solved, and the time cost and the labor cost for obtaining a large amount of high-resolution magnetic resonance image data are reduced.

The invention discloses an SEM image microdefect analysis method and system based on machine learning, and the method comprises the steps: obtaining an SEM image, and sequentially carrying out the image preprocessing and multi-strategy binarization segmentation, and obtaining a binary segmentation SEM image; based on a pre-trained random forest classifier, firstly classifying the binary segmentation SEM image and then carrying out defect correction processing to obtain a bridge path of the corrected SEM image defect; performing physical scale calibration processing and skeletonized analysis visualization processing on the bridge connection path of the corrected SEM image defect to obtain a visualized SEM image defect; based on the visualized SEM image defects, integration and report generation analysis are carried out, and an SEM image microdefect analysis report is obtained. According to the invention, comprehensive, accurate and efficient automatic analysis of microdefects can be realized. The SEM image microdefect analysis method and system based on machine learning can be widely applied to the technical field of material microstructure characterization analysis.

The present invention relates to the field of image analysis technology, specifically to a method for semantic segmentation of cracks in concrete arch bridges based on multispectral imaging. The method comprises the following steps: collecting multi-time multispectral images using an unmanned aerial vehicle (UAV), segmenting crack regions using K-means on the multispectral images, extracting and sliding normalizing a multi-band spectral vector sequence to generate a principal axis spectral vector, calculating the spectral vector angle and rate of change to mark transition points, screening boundary points by combining directional consistency and gradient scoring, extracting spectral patterns through time-series alignment and Z-score standardization, evaluating discrete fluctuations and correcting anomalies to output a binary segmentation map of concrete arch bridge cracks. In the present invention, a reflectance spectrum sequence is constructed using multi-time multispectral images, crack identification accuracy is improved by combining principal axis features and a sliding window, texture interference is eliminated by using spectral vector angle changes and gradient scoring, and a standardized spectral pattern time series model is generated to enhance boundary location accuracy, reduce noise interference, and support long-term monitoring of structural damage.

The invention provides a fault detection method and device, electronic equipment and a storage medium, and belongs to the technical field of image processing, and the method comprises the steps: obtaining an optical image of a target structure at a current moment, and extracting a multi-scale feature map of the optical image; based on the multi-scale feature map and prompt information, determining a preliminary binary segmentation mask map corresponding to the boundary of the target structure contained in the optical image, the prompt information indicating boundary information of the target structure contained in the image; and comparing the preliminary binary segmentation mask pattern with preset mask patterns, determining the fault type of the target structure, and enabling different preset mask patterns to correspond to different fault types. According to the technical scheme provided by the embodiment of the invention, the obtained optical image is subjected to preliminary binary segmentation mask image comparison, so that the fault type of the target structure can be quickly identified, and the fault identification precision is improved.

The invention discloses a driver direct view calculation model and a test method. Comprising the following steps: firstly, acquiring virtual test scene data corresponding to a to-be-tested vehicle; and inputting the virtual test scene data into the trained calculation model. The calculation model comprises a view region segmentation sub-model and a visible volume calculation sub-model, and a visible region and a shielding region in a standard evaluation volume in the virtual test scene data are accurately segmented through the view region segmentation sub-model to obtain a corresponding binary segmentation mask. And extracting spatial distribution characteristics strongly related to the visible volume from the segmented binary mask through a visible volume calculation model, fusing, screening and compressing the extracted spatial distribution characteristics, and finally mapping the spatial distribution characteristics into four types of visible volumes of the near side, the front side, the far side and the total of the tested vehicle. Therefore, a large entity test scene is not needed, and the test efficiency is improved. And when the standard is updated, only a data set needs to be updated and model parameters need to be calculated, and the overall architecture does not need to be redeveloped.

The invention provides a medical image segmentation method and device based on topological constraint enhancement, and the method comprises the steps: obtaining an original medical image data set, carrying out the preprocessing, generating a multi-scale random field according to the size of a preprocessed original image, carrying out the smoothing processing through a Gaussian kernel, and obtaining a smoothed multi-scale displacement field, performing weighted fusion operation on the multi-scale displacement field, and performing topological constraint forcing through Jacobian determinant analysis according to the fused displacement field to obtain an adjusted deformation field; calculating a transformed coordinate position according to the adjusted deformation field, and carrying out boundary clipping on a boundary crossing coordinate to obtain an enhanced data set; inputting the enhanced data set into a pre-trained convolutional neural network for processing so as to obtain a segmented probability image, and performing context-aware adaptive threshold optimization processing on the segmented probability image so as to obtain a final binary segmentation result; therefore, high-precision and low-complexity medical image segmentation is realized.

The invention relates to the technical field of artificial intelligence, and provides a face counterfeit segmentation method based on multi-condition input. The method comprises the following steps: extracting feature bottom noise in an image; and high-frequency information in the image is extracted. And constructing a multi-condition input segmentation neural network, and performing efficient fine tuning training on the multi-condition input segmentation neural network until the multi-condition input segmentation neural network converges. And the trained multi-condition input segmentation neural network model is utilized to carry out reasoning on the image, the segmentation neural network model finally outputs a binary segmentation image consistent with the original image in size, and a face counterfeit area in the image is detected and positioned. According to the method, feature bottom noise and high-frequency information in a forged image are extracted, the information serves as additional conditions and is input into a segmentation network together with an original image, a model is trained by means of a designed multi-condition input efficient fine tuning architecture, the model can predict a binary segmentation image, and the image segmentation efficiency is improved. Therefore, the face counterfeit area in the image is detected and accurately positioned.

The invention discloses a muck pile lumpiness calculation method based on three-dimensional laser point cloud context features, and relates to the technical field of smart mines, and the method comprises the steps: carrying out the preprocessing operation of muck pile point cloud data acquired by a three-dimensional laser scanner; based on the context features of the point cloud data in combination with a Euclidean clustering strategy, performing primary segmentation on the preprocessed point cloud data to obtain a first segmentation result; performing secondary segmentation on the first segmentation result by adopting a dynamic parameter optimization algorithm to obtain a second segmentation result; and obtaining a muck pile lumpiness calculation result according to the second segmentation result and the maximum point cloud distance. According to the method, high-precision three-dimensional scene information is obtained by utilizing a laser scanning technology, and accurate calculation of the blasting storage yard lumpiness is realized by combining a context feature analysis method, so that the precision and efficiency of blasting pile lumpiness calculation are improved, and reliable data support is provided for blasting engineering.

The invention discloses a prostate MR image segmentation method and system based on multi-level context aggregation, and the method comprises the steps: segmenting a prostate MR image through a trained prostate MR image segmentation model, and obtaining a segmentation result. The model is based on an improved U-shaped architecture and comprises a multi-level coding module, a jump connection processing module, a bottleneck layer, a multi-level decoding module and a segmentation head. The multi-level coding module extracts image features of different levels, and the jump connection processing module generates enhanced features through combination of hierarchical expansion convolution and an improved Transform architecture; the bottleneck layer decouples the bottom layer features into foreground / background features, and outputs a feature map through local window attention weighted fusion; and the multi-level decoding module gradually recovers the resolution through residual convolution and bilinear up-sampling, and finally outputs a binary segmentation result through a segmentation head. According to the method, the accuracy and robustness of prostate MR image segmentation are improved through hierarchical feature aggregation and attention fusion.

The invention discloses an infrared target detection method based on a multi-attention shape adaptive network, comprising: obtaining an infrared image to be detected; inputting the infrared image to be detected into a pre-trained multi-attention shape adaptive network, and obtaining a binary segmentation map corresponding to the infrared image to be detected output by the network; the multi-attention shape adaptive network comprises: a densely nested U-shaped feature extraction network and an output layer based on spatial feature fusion, a bidirectional shape adaptive extraction module and a multi-channel shape enhancement module connected in sequence are added to the lateral connection of the densely nested U-shaped feature extraction network, the densely nested U-shaped feature extraction network is used to spatially fuse visual features of different semantic levels, the extraction module is used to adaptively extract the shape information of the infrared target, and the enhancement module is used to enhance the shape information extracted by the extraction module. The invention can fuse visual features of different semantic levels, reduce the semantic gap in the feature fusion stage, and improve the accuracy of infrared target detection.