35 results about "Graph cut algorithm" patented technology

The invention discloses a distributed power supply load model establishment method and system, and relates to the technical field of load model establishment, after an establishment system obtains power system information through an API interface, a topological graph model of a power system is established, each component in the power system is abstracted as a node and an edge in the topological graph model, and the topological graph model is established; after simulation software is used for conducting simulation detection on the topological graph model, a target value of the topological graph model is calculated through an objective function, an optimal path between nodes is updated through a path algorithm, current distribution of edges is adjusted through a graph cut algorithm, power loss is reduced, overload is avoided, the steps are repeated for multiple iterations, and when convergence conditions are met, the optimal path is obtained. And outputting the final topological graph model as a comprehensive load model. The system can optimize the targets at the same time through a graph optimization algorithm, improves the overall benefits of the system, and overcomes the defect that the complexity of the power system cannot be comprehensively considered in single target optimization.

The invention discloses a food waste detection method and system based on image processing, belongs to the field of image recognition, and aims to realize efficient and automatic recognition and quantification of kitchen waste. According to the method, residual food images are collected at multiple periods and multiple angles in a kitchen garbage can or a dinner plate recovery area through high-resolution and multi-spectral imaging equipment, preprocessing is carried out in combination with an improved Retinex algorithm and a space self-adaptive denoising technology, and the image quality is improved. Afterwards, fine segmentation of a food area is achieved through a multi-scale super-pixel segmentation and graph segmentation algorithm, and multi-category intelligent recognition is conducted on remaining food through a recognition network fused with multi-modal features. The system further combines stereoscopic vision and Monte Carlo sampling to dynamically and accurately count the volume or weight of various residual foods. The method has the advantages of high adaptability and accurate statistical result, and can provide data support for catering management, resource recovery, nutrition evaluation and the like.

The invention relates to an artificial intelligence image segmentation processing method and device, equipment and a medium. The method comprises the following steps: acquiring pixel intensity distribution of an input image and extracting local features by using a convolutional neural network to generate a pixel intensity change feature map; on the basis of the feature map, boundary area probability distribution is calculated by using a conditional random field model, and a zigzag edge is smoothed by using a graph cut algorithm to generate a continuous segmentation boundary; splicing the continuous segmentation boundary and an input image, inputting the spliced continuous segmentation boundary and the input image into a Transform enhanced U-Net network, balancing edge alignment loss and region overlapping precision loss through a dynamic weighted loss function, and generating an optimized segmentation mask; and performing threshold binarization processing on the optimized segmentation mask, removing isolated noisy points in combination with morphological closed operation, and generating a segmentation image matched with the physical defect form. By adopting the method, the continuity of segmentation boundaries and the matching precision of defect forms can be improved, and the problems of edge sawteeth and noise interference in traditional segmentation are solved.

The invention discloses a complex slope runoff simulation method based on multi-source data fusion. Comprising the following steps: fusing an unmanned aerial vehicle LiDAR, a ground penetrating radar and a remote sensing image, and considering physical coupling relationships such as vegetation root systems to generate a high-precision and physically consistent initial earth surface parameter field; the method comprises the following steps: identifying micro-topographic depression through a graph cutting algorithm and the like, and constructing a dynamically evolved slope overflow network based on a depression filling-overflow mechanism; establishing an overflow-scour-infiltration dynamic feedback regulation and control mechanism, wherein the scour intensity calculated according to overflow data is used for correcting the saturated hydraulic conductivity of the soil in real time, and corrected parameters are immediately fed back to an improved Green-Ampt infiltration model; through coupling solution of a multi-time scale algorithm, spatial and temporal distribution of slope runoff is output, and early warning and inversion of critical rainfall conditions can be triggered. The underlying surface dynamic feedback process can be simulated, the simulation reliability is improved, and technical support is provided for mountain torrent early warning.

The invention relates to the field of digital modeling, and particularly discloses a mine modeling method and system based on point cloud data, a terminal and a medium, and the method comprises the steps: obtaining original point cloud data of a target mine area; preprocessing the original point cloud data, including voxel filtering, to obtain target point cloud data; calculating the curvature and the normal vector of each point in the target point cloud data; inputting the coordinate, the curvature and the normal vector of each point as input values into a pre-trained geological category classification model to obtain the probability that each point belongs to various geological categories; constructing an energy function of a graph cut algorithm based on the probability that each point belongs to various geological categories and the point cloud features of the target point cloud data, and obtaining the final geological category to which each point belongs through the graph cut algorithm; and constructing a digital model of the target mine area based on the geological category to which each point belongs. According to the method, geological category prediction is carried out based on the multi-scale features of the point cloud, optimization is carried out in combination with the graph cut algorithm, and the classification efficiency and robustness are improved.

The invention provides a pipeline defect detection method, electronic equipment and a storage medium, and relates to the technical field of computer vision. The method comprises the following steps: performing coordinate conversion on an initial fisheye video stream to obtain a multi-frame plane image; for any frame of plane image, determining a binarization mask of the frame of plane image based on a semantic segmentation network, and extracting feature points of a pipe wall region in the frame of plane image by taking the binarization mask as a constraint to form a feature point set; determining an optimal suture line between any two adjacent frames of feature point sets by adopting a graph cut algorithm based on energy minimization, and carrying out splicing to obtain a pipeline panorama; and identifying the pipeline panorama by using a deep neural network model to perform defect identification to obtain a target defect. According to the method, the binarization mask is used as a constraint, and non-pipe wall area interference is eliminated; an optimal suture line is determined and splicing is completed based on an energy minimization graph cut algorithm, so that inter-frame splicing gaps and artifacts are effectively eliminated, and rapid and accurate detection of target defects can be realized.

The invention discloses a circuit board welding spot detection method based on surface structured light, and the method comprises the steps: installing an electric rotatable polarizing film in front of a lens of a binocular camera, and synchronously obtaining stripe images at four orthogonal polarization angles in a structured light collection stage. A Fresnel equation and a micro-surface model are utilized to establish an analytic model of polarization state and phase error, pixel-by-pixel reflectivity suppression is carried out on original stripes, and active physical compensation of high reflection error is realized. A micro-area self-adaptive phase unwrapping algorithm based on welding spot topology priori is provided, a Markov random field with process priori constraint is constructed, a Graph Cuts algorithm is used for solving, and edge period dislocation is eliminated fundamentally. And a dynamic graph convolutional neural network is adopted to extract depth features of the welding spot cloud, and few-sample intelligent classification of each type of defects only needing 5-15 samples is realized through contrast quantity learning and a dynamic memory bank.

The application relates to a CAD vector data intelligent preprocessing method based on a multiple recognition mechanism, which comprises the following steps: parsing a DXF file to construct a CAD heterogeneous information graph; extracting text features for a layer node and generating a layer embedding vector through adaptive fusion of a gate unit by aggregating high-order semantic structure features through a meta path; introducing a dynamic threshold decision mechanism to generate a dynamic threshold according to global statistical features of a drawing to determine a background layer; constructing a global dependency graph containing cleaned dependent edges and semantic constraint edges, dividing the global dependency graph into weakly coupled connected subgraphs as independent optimization transactions by using a graph cut algorithm, and performing parallel execution to complete cross-dimension collaborative optimization; setting a multi-level logical checkpoint and generating an atomic operation log, and realizing tracing and accurate recovery based on an incremental storage mechanism. The application has the effects of significantly improving background layer recognition accuracy and scene adaptability, improving collaborative optimization efficiency, and guaranteeing operation safety and traceability.

The application discloses a kind of method and device for microorganism counting using combined image, it is related to microorganism counting technical field, including, using convolutional neural network model to extract the core feature point of preprocessed microscopic image, and the similarity between core feature point is calculated, obtain the microorganism microscopic image containing splicing position.Utilize adaptive splicing algorithm to the splicing of microorganism microscopic image, identify and correct the error that appears, obtain the panoramic image after splicing.Cut the panoramic image after splicing into graph algorithm, construct the energy function of panoramic image, find out the lowest energy segmentation path using minimum cut algorithm, generate the panoramic image of microorganism region.The application greatly improves the matching precision between adjacent images and the stability of geometric transformation, and the use of breadth-first search algorithm and minimum cut algorithm improves the segmentation accuracy of image segmentation, avoids the situation that image is misaligned and overlaps inconsistently.

The invention relates to the technical field of image processing, in particular to a large-range water area time-space dynamic change monitoring method fusing remote sensing and GIS, and the method comprises the steps: obtaining the hyperspectral remote sensing image data of a to-be-monitored water area in real time; aiming at each moment, constructing a feature vector of each pixel point and a center vector of each area label in the water area to be monitored by combining a normalized water index and a normalized vegetation index of each pixel point through gray difference and texture feature difference between each pixel point and an adjacent pixel point in the hyperspectral remote sensing image data and by combining the normalized water index and the normalized vegetation index of each pixel point; and constructing an energy function of a graph cut algorithm for segmenting the to-be-monitored water area, segmenting the to-be-monitored water area at each moment through the energy function, and carrying out spatial-temporal dynamic change monitoring on the to-be-monitored water area through segmentation results at different moments. The invention aims to improve the accuracy of monitoring the spatial-temporal dynamic change of the water area by improving the accuracy of area segmentation.

The invention discloses a finished pastry appearance analysis method and device based on AI image analysis and recognition, and relates to the technical field of food detection. Comprising the following steps: S1, acquiring a multispectral image, and preprocessing the multispectral image to obtain a preprocessed multispectral image; s2, recognizing the preprocessed multispectral image through a set rotatable anchor frame, obtaining a pastry position posture detection frame, and cutting the pastry position posture detection frame through a graph cutting algorithm to obtain a cutting result; and S3, obtaining feature data according to the separation boundary, and determining a confidence degree according to the feature data. The problems that traditional manual detection is low in efficiency and difficult to cope with random postures and background interference are solved, and automatic classification and grading treatment of pastry fermentation states, foreign matter mixing and appearance defects are achieved.

The application discloses a kind of multi-source ship target identification methods based on two-stage cooperative fusion, it is related to signal processing technical field, first with the optical camera on the plane and radar sensor acquisition target sea area data, including original optical image and radar image data, then using Gaussian filter to the data collected is preprocessed;Next, using deep learning algorithm is carried out sea-land segmentation, constructs loss function and optimizes network parameter to obtain segmentation mask, then according to the mask result removes land background area, obtains the optical and radar image after processing;Afterwards, using the registration method that improved SURF feature and graph cut algorithm are combined, by screening feature point pair and solving minimum value of energy function, optimal transformation parameter is obtained, image registration is realized;For the image after registration, based on adaptive fusion rule to fuse image;Finally, corresponding feature is extracted by hand feature extraction and based on deep learning feature extraction and is fused to obtain final target recognition result.

The present invention relates to the technical field of data clustering and integration, and discloses a multi-view financial data clustering and integration method and device. The method comprises: processing financial multi-view data using a K-Means clustering method to generate an ensemble pool, and then using a fuzzy membership function to generate a binary partition matrix; learning the global structure of the label space through the binary partition matrix to obtain an initial co-correlation matrix, and optimizing it using an LSR model with a Frobenius norm to obtain a label co-correlation matrix; learning a subspace projection matrix, an affinity matrix, and a sample adjacency matrix of each view, optimizing the co-correlation matrix from a feature space, and obtaining a feature co-correlation matrix; combining the label co-correlation matrix and the feature co-correlation matrix to obtain an optimized co-correlation matrix; projecting the optimized co-correlation matrix into a constraint space for constructing an undirected graph; and partitioning the undirected graph using a graph cut algorithm to obtain a final clustering integration result of the financial multi-view data.

The present application relates to the technical field of plate identification, in particular to a plate identification method based on geometric range discrimination. Mainly including the following steps: using image acquisition equipment, laser scanner and BIM model to obtain the geometric data of the plate, including boundary contour, surface morphology, spatial position and size information; extracting the boundary and key geometric feature points of the plate through graph cut algorithm; generating a three-dimensional geometric model by using multi-scale shape analysis algorithm; using Markov random field algorithm for geometric range discrimination to identify abnormal areas; comparing the BIM model design data by using iterative closest point algorithm to generate an identification report. The present application realizes accurate identification and range discrimination of building structure plates, effectively improving the monitoring and management efficiency of construction quality.

This invention discloses an intelligent analysis method for low-altitude remote sensing data from low-altitude unmanned aerial vehicle (UAV) hangars, belonging to the field of intelligent control technology. This scheme first collects multi-source heterogeneous data through timestamp alignment and hardware-triggered synchronization mechanisms; then, geometric correction is achieved through joint optimization of the RPC model and GCP, and image stitching is completed by combining SIFT feature matching and graph cut algorithms; subsequently, an improved YOLOv8 model is constructed to achieve multimodal fusion target detection, and semantically guided change detection is completed based on the Transformer algorithm; a high-precision 3D model is generated by fusing multi-source data using an enhanced NeRF algorithm, and vegetation analysis and trend prediction are completed using a dynamic weighting algorithm and an LSTM network; finally, a comprehensive report is generated based on a knowledge graph and interactive visualization output is provided. This invention effectively solves the problems of inefficient multi-source data processing and insufficient core detection accuracy in existing technologies, significantly improving data processing efficiency, detection accuracy, and analytical practicality, and is applicable to multiple scenarios such as ecological monitoring and resource exploration.

The present invention discloses a method for liver tumor segmentation in CT sequence images based on deep learning. The method mainly comprises: (1) constructing a U-shaped 2D convolutional network based on dilated spatial pyramid convolution, and using the network to perform two-dimensional slice segmentation of CT sequence images from three viewing directions: sagittal, coronal, and transverse; (2) using a lightweight 3D convolutional network to fuse the segmentation results obtained from different viewing directions to obtain the probability that each pixel in the CT sequence belongs to the target and the three-dimensional segmentation results of the CT sequence liver tumor; (3) constructing a graph cut energy function based on the obtained probabilities and three-dimensional segmentation results to further optimize the segmentation results. By combining 2D and 3D convolutional networks and a graph cut algorithm, the present invention can effectively extract three-dimensional spatial information from CT sequences while using a lightweight network, thereby improving the accuracy of liver tumor segmentation.

The present invention discloses a solid-state hard disk storage optimization method and system, which relates to the field of data storage technology. The method comprises: partitioning storage units into regions based on behavioral layer data and physical layer data of the solid-state hard disk using a topological coloring algorithm and a graph cut algorithm; detecting abnormal behavior of storage units in each region using an isolation forest algorithm to obtain abnormal storage units; obtaining incremental change data of the abnormal storage units; generating an optimization strategy set based on a neural network model using the feature data and incremental change data of the abnormal storage units as input, so that different regions receive different optimization treatments; and optimizing the storage of the solid-state hard disk after obtaining the optimal execution strategy based on the optimization strategy set and under the principle of balancing the optimization strategy of the abnormal storage unit with the overall optimization goal through multi-objective tuning. The present invention can apply different optimization strategies to the internal region, boundary region, and external region of the solid-state hard disk.

The present application discloses a method, device and storage medium for segmenting large intestinal effusion. The method includes: extracting the target skeleton with the largest skeleton volume from the three-dimensional intestinal image and using it as the initial segmentation area; based on the initial segmentation area, using the active contour model to perform contour extraction to obtain the intestinal segmentation result; calculating the minimum enclosing circle of the intestinal segmentation result in the cross-sectional area; using the Gaussian mixture model to fit the probability distribution of the large intestine area and the probability distribution of the large intestine effusion area within the minimum enclosing circle; and inputting the fitting result into the graph cut algorithm model for effusion segmentation to segment the large intestine effusion area from the large intestine area. Using the solution of the present application, the large intestine effusion area can be accurately extracted, and subsequently by combining the large intestine effusion area with the large intestine area, the complete large intestine area can be segmented.

A volumetric based three-dimensional (3D) object is segmented. A training dataset of 3D training objects is acquired. A neural network is defined by multiple layers including a linear layer and a graph layer. The linear layer operates on 2D objects and provides inputs to the graph layer that performs convolution operations on vertices of the 3D objects. A model, that approximates a shape diameter function (SDF) determines neighborhood diameters including a distance from a vertex of the 3D objects to an antipodal vertex. The model is generated by iterating through the 3D objects using the neural network to converge on weights of SDF features. An input mesh is acquired and the converged weights are used to approximate SDF values. The SDF values are input to a graph cut algorithm that generates vertex clusters defining a segmented part of the input mesh. The segmented part is visually displayed or provided.

The application discloses a kind of narrow river detection methods suitable for spaceborne wide swath interferometric radar altimeter image, comprising: reading into spaceborne wide swath interferometric radar altimeter image;It is enhanced river linear feature by Gaussian difference preprocessing to suppress background;Curvature structure perception detector is constructed, curvature response map is calculated based on Hessian matrix eigenvalue, and feature fusion enhancement is carried out in conjunction with multi-direction structure consistency score;The binary image obtained by adaptive threshold segmentation based on local statistics is used as region label by carrying out adaptive threshold segmentation to enhanced feature map, maximum value is extracted in the region corresponding to enhanced feature map, seed point set is formed, and adaptive region growth based on queue priority is executed, to generate initial river center line;Markov random field model is introduced, global structure optimization and topological connectivity correction are carried out using graph cut algorithm, to obtain optimized river center line;Direction and radiation characteristic constraint is applied again, morphological dilation is carried out, and finally complete river mask is generated.

A volumetrically based three-dimensional (3D) object is segmented. A training dataset of 3D training objects is acquired. A neural network is defined by several layers, including a linear layer and a graph layer. The linear layer operates on 2D objects and provides inputs to the graph layer, which performs convolution operations on the vertices of the 3D objects. A model approximating a shape-diameter function (SDF) determines neighborhood diameters, which include the distance from a vertex of the 3D objects to an antipodal vertex. The model is generated by iterating through the 3D objects using the neural network to converge with respect to the weights of SDF features. An input mesh is acquired, and the converged weights are used to approximate SDF values.The SDF values ​​are fed into a graph-cutting algorithm that generates vertex clusters, defining a segmented portion of the input mesh. The segmented portion is then visually displayed or provided.

This paper discloses a distributed graph neural network training method based on heterogeneous devices. This method processes large graph data and uses a graph cut algorithm to obtain several subgraphs with balanced nodes and edges. A dynamic programming algorithm is then used to determine an optimal device placement plan. The training workload for each device is planned based on the training capabilities of the heterogeneous devices, allowing all devices to complete training tasks as simultaneously as possible within the same iteration, thereby improving edge device utilization. A fault-tolerance mechanism is also added to ensure that the network maintains a certain level of functional performance even when network nodes fail or are lost.

The invention discloses a panoramic image suture line searching method and device based on visual saliency guidance, and the method comprises the steps: eliminating distortion through preprocessing, separating a dynamic foreground, generating a saliency map through a lightweight model, and constructing a multi-component energy function fusing chromatic aberration, gradient and saliency. And searching an energy minimum path based on a graph cut algorithm, optimizing smoothness through a Bezier curve, and finally realizing natural splicing through multi-band fusion. The modular architecture comprises a preprocessing unit, a significance detection unit, an energy construction unit and the like, is realized by mixing C + + / CUDA (Compute Unified Device Architecture) and Python, is compiled into a lightweight dynamic library, and supports multi-platform deployment of a PC (Personal Computer), embedded equipment and the like. According to the scheme, the problems that the suture line passes through the salient region, the energy model is single and the like are solved, the visual comfort and the algorithm robustness are improved, and the method is suitable for industrial detection, mobile shooting and other scenes.

This invention relates to the field of smart fire protection technology, specifically to a method and system for dynamic assessment of building fire risks based on big data. It includes: S1, constructing a heterogeneous information network model, forming a heterogeneous information network graph of building fire protection based on multi-source data, and constructing a digital twin environment; S2, inputting the heterogeneous information network graph into a spatiotemporal graph convolutional network model, predicting the dynamic attribute weight changes of nodes and edges; S3, constructing a fire spread seepage model and a rescue failure seepage model based on evolutionary trends, and using a graph cut algorithm to identify critical seepage nodes / edges; S4, calculating fire resilience entropy based on the predicted network state; S5, extrapolating the runaway time window based on critical seepage nodes / edges and fire resilience entropy, generating risk assessment results, and mapping them to the digital twin environment for visualization. This invention can identify key weak points that lead to system collapse, predict the remaining time for fire spread to critical areas, and provide forward-looking decision support for fire rescue.

The invention relates to the technical field of remote sensing image processing, in particular to a cloud snow layer image segmentation method and system based on edge optimization and a storage medium, and the method comprises the steps: segmenting a foreground region containing a cloud layer and a snow layer on a remote sensing image through a ResNet network, and a background region; extracting the edge information of the cloud layer and the snow layer in the foreground region by using an edge detection operator to obtain the edge information of the foreground region; and fusing the edge information into a graph cut algorithm framework, and segmenting a cloud layer and a snow layer in the foreground region. According to the method, firstly, the ResNet stage is utilized to ensure that the cloud snow foreground is completely detected, then the edge optimization graph cutting stage is utilized to accurately guide the boundary by utilizing the mixed edge weight, so that the segmentation contour is highly matched with the real ground feature edge in the image, respective advantages of deep learning and the graph cutting model are fully utilized, and the image segmentation efficiency is improved. And high-precision and high-robustness automatic segmentation of the cloud layer and the snow layer in the remote sensing image is realized.

The invention relates to the technical field of water supply transformation, and discloses a water supply dispatching frequency conversion regulation and control strategy optimization method based on pressure balance, which comprises the following steps of: obtaining terrain elevation data and pipe network topology data of a construction area, performing data preprocessing and generating a terrain gradient feature vector; calculating gravitational potential energy distribution by using a terrain flow field coupling analysis algorithm, generating a terrain influence coefficient matrix, analyzing a pipe network segmentation condition caused by construction based on a graph cut algorithm, calculating topological change characteristics of an affected sub-network, and integrating the terrain influence coefficient matrix and the topological change characteristics. According to the method, a quantitative relation between gravitational potential energy distribution and pressure propagation is established through a terrain flow field coupling analysis algorithm, so that the influence of terrain factors on pressure can be accurately quantified, and pressure compensation requirements at different elevation positions are differentially processed by introducing a terrain influence coefficient matrix; the defect that the terrain influence is simplified into a single parameter by a traditional method is overcome.

The application provides an ultrasonic probe shell structure damage positioning method based on a multi-channel echo consistency atlas, and five-layer progressive signal preprocessing is sequentially performed on the data of each channel to screen effective envelope wave bands; apparent damage distances and Shannon entropy weighted integrated energy attributes of each wave band are calculated, a cross-channel distance consistency atlas is constructed based on distance values of all channels, and wave bands are clustered into a plurality of suspected damage distance communities through a graph cut algorithm; the reliability of each community is quantified from three dimensions of normalized community energy, pulse width distribution compactness and cross-channel appearance frequency, a self-adaptive weighted harmonic mean model is used for fusion output of community comprehensive confidence and identification of effective damage echoes; distance values of each wave band in an optimal confidence community are weighted and fused in a nonlinear square weight manner to obtain a final damage positioning distance. The method has sub-centimeter positioning accuracy, strong noise robustness and wide structural and working condition adaptability in the detection of defects of plate shell structures.

The application provides a blood vessel segmentation modeling method based on topological knowledge in medical images, first, a segmentation arteriovenous vessel model is created based on a U-Net neural network, and the model is trained by using sample images and their labeled results, in the training, a partial supervision method is used, and partial labeled data is added in the data set; a graph cut algorithm is used as post-processing to optimize the prediction result of the neural network, and the accuracy of the blood vessel segmentation result is improved; Grow Cuts, layer traversal and other algorithms are used to construct the topological structure of the blood vessel, including segmenting the blood vessel, obtaining the radius of the blood vessel and extracting the arteriovenous sub-tree structure. The application uses the mixed training of complete labels and incomplete labels in the data set, introduces the optimization algorithm such as graph cut, guarantees the segmentation accuracy, improves the robustness of the algorithm, realizes the construction of the topological structure of the blood vessel, and forms a complete algorithm process for extracting the blood vessel from the medical image and constructing the topological structure of the blood vessel.

The invention relates to the technical field of image processing, in particular to a rock debris three-dimensional CT image segmentation method, device and equipment based on an image segmentation algorithm and a storage medium. After a target rock debris three-dimensional image is obtained, a target image model is constructed according to the target rock debris three-dimensional image, and image segmentation processing is carried out on the target image model; obtaining an original image segmentation result; according to graph nodes and adjacent edges of the target graph model, performing quality evaluation on the original image segmentation result to obtain segmentation quality evaluation data; and performing optimization processing on the original image segmentation result according to the segmentation quality evaluation data to obtain a target image segmentation result. According to the invention, the method can automatically achieve the segmentation of the rock debris image, improves the automation degree of operation, reduces the time consumed by the segmentation of the rock debris image, and shortens the segmentation time, so that the segmentation of the rock debris image is more efficient. The rock debris image segmentation method is suitable for a processing scene in which a large number of rock debris three-dimensional images are segmented.

The invention discloses a narrow river detection method suitable for a satellite-borne wide-cradling interference radar altimeter image. The method comprises the following steps: reading in the satellite-borne wide-cradling interference radar altimeter image; gaussian difference preprocessing is carried out to suppress background and enhance the linear features of the river channel; constructing a curvature structure sensing detector, calculating a curvature response diagram based on a Hessian matrix eigenvalue, and performing feature fusion enhancement in combination with a multi-direction structure consistency score; performing self-adaptive threshold segmentation based on local statistics on the enhanced feature map to obtain a second map as a region mark, extracting a maximum value in a corresponding region of the enhanced feature map, forming a seed point set, executing self-adaptive region growth based on queue priority, and generating an initial river channel center line; introducing a Markov random field model, and performing global structure optimization and topological connectivity correction by using a graph cut algorithm to obtain an optimized river channel center line; and then applying direction and radiation characteristic constraints, performing morphological expansion, and generating a final complete riverway mask.