24 results about "Scale invariance" patented technology

The invention discloses a multi-modal large model three-dimensional perception method based on Riemannian manifold prior guidance, and relates to a computer vision technology. The method comprises the following steps: firstly, constructing a multi-modal large model and point cloud sensing network coordinated cross attention reinforcement learning joint training framework, so that the large model obtains a three-dimensional scene space topology understanding capability; secondly, a conformal property on a Riemannian manifold is utilized to construct a three-dimensional contour and topological association, and multi-scale invariance of homotopy mapping learning contours is introduced, so that the challenges of shape diversity and great size difference of objects are solved; in addition, a lightweight three-dimensional attention gate filtering mechanism is introduced into a point cloud encoder, and more effective global and local point cloud geometric semantic association is established. And finally, when the motion pose quality is evaluated, fusing prior knowledge of the three-dimensional physical relationship to form mixed physical measurement so as to overcome label noise caused by single evaluation scale. And universal understanding and high-precision perception of the robot on complex three-dimensional environments and objects can be realized.

The invention relates to a multi-modal image matching method and system based on learning features and epipolar geometric constraints. The method comprises the following steps: carrying out edge enhancement processing on an input image through wavelet transform; extracting a multi-scale dense feature map based on the transformed convolutional neural network, and generating a feature descriptor with rotation and scale invariance in combination with principal direction normalization; adopting an FLANN algorithm and dynamic distance constraint to realize preliminary feature matching; and introducing a basic matrix construction and epipolar geometric consistency verification mechanism, and eliminating mismatching point pairs in combination with an RANSAC affine constraint model. According to the method, image enhancement, deep learning and geometric verification strategies are fused, the problems of radiation nonlinearity and geometric distortion caused by imaging mechanism differences among multi-modal images are effectively solved, the matching precision and robustness are improved, and the method is suitable for remote sensing application scenes such as optical-SAR registration, multi-source image fusion and earth surface change detection.

A point cloud denoising method based on a denoising autoencoder. First, the point cloud data is processed, and the point cloud denoising problem is treated as a local problem. The neighborhood of each point is taken and randomly sampled. Secondly, the Transform layer appropriately destroys the input data to create obstacles for subsequent feature extraction. Then, the point cloud is aligned using the rotation matrix calculated by principal component analysis, rotating the point cloud to the same angle. Then, the Encoder layer extracts potential features from the damaged data through a multi-layer perceptron and uses maximum pooling to enhance translation invariance, rotation invariance, and scale invariance. Finally, the Decoder layer of the network decodes the potential features through full convolution and outputs the predicted displacement of the noise point to complete the denoising. The present invention removes noise as efficiently as possible while maintaining the geometric characteristics of the point cloud data.

The invention discloses a network topology reconstruction-based adaptive sampling method for a CAE (Computer Aided Engineering) agent model of a numerical control machine tool, and relates to the research field of sampling methods for agent models in multi-dimensional complex scenes such as the numerical control machine tool. S12, a sample point preprocessing and validity checking stage; s2, a network topology construction stage; s23, a network topology validity verification and weight normalization stage; s3, a topology reconstruction stage; and S4, a dynamic optimization stage. According to the network topology reconstruction-based adaptive sampling method for the CAE agent model of the numerically-controlled machine tool, a sampling process is converted into a dynamic network reconstruction process, power-law distribution, small-world effect and scale invariance characteristics in a complex network theory are fused, system key information is captured by using a network topology structure, and global search capability of a simulated annealing algorithm is combined, so that the network topology reconstruction-based adaptive sampling method for the CAE agent model of the numerically-controlled machine tool is realized. The adaptive optimization distribution of sampling points is realized, the coverage rate and representativeness of samples to different working condition characteristics are improved, and high-quality training data are provided for a numerical control machine tool CAE agent model.

The present invention relates to a lightweight pest recognition method based on a Transformer structure, which belongs to the field of deep learning and comprises the following steps: S1: extracting shallow features of pest images using a focused fast downsampling module; S2: extracting global feature information in a deep feature map using a multi-head self-attention module; S3: adding local feature sensitivity and scale invariance information to the deep feature map using local convolution; S4: performing feature splicing on the global feature information with the local feature sensitivity and scale invariance information to obtain a pest image rich in semantic information, sending the global feature information to a multi-layer perceptron, and performing feature fitting on the fused feature tensor; S5: reducing the gradient vanishing problem of the network through residual connection, and integrating the information in the channel through point-by-point convolution; and S6: using a pooling mechanism to classify the finally calculated feature representation through a classification module.

This invention discloses a rotation- and scale-invariant feature extraction method and system based on dual-coordinate system collaboration, belonging to the field of image processing technology. Addressing the poor matching performance of existing image feature extraction methods under rotation and scale changes, this invention employs the following scheme: The input image undergoes a logarithmic polar coordinate transformation to generate a polar coordinate image; a dual-branch network is constructed, where the feature extraction branch extracts keypoint location information in a Cartesian coordinate system, and the feature description branch extracts a polar coordinate feature map in a logarithmic polar coordinate system; the keypoint locations in the Cartesian coordinate system are mapped to their corresponding positions in the polar coordinate feature map using a coordinate mapping function, and interpolation sampling is performed to generate feature descriptors that integrate precise location information and rotation / scale invariance information; based on the feature descriptors, matching point pairs are calculated, and the spatial transformation relationship between images is determined. This invention achieves highly robust feature extraction under rotation and scale changes while maintaining high real-time processing efficiency.

The present application relates to the field of computer aided design, in particular to a kind of three-dimensional model retrieval method and system based on spatial distribution projection image;The system includes processing module and output module, processing module is used to obtain the spatial distribution projection image corresponding to each three-dimensional model by the way of projecting the three-dimensional point cloud corresponding to three-dimensional model to typical surface;Again, the descriptor of three-dimensional model is obtained by spatial distribution projection image, finally the similarity of three-dimensional model is judged by descriptor;Output module is used to output retrieval result according to similarity;The method is realized by system, based on the typical surface of three-dimensional model with rotation and scaling invariance obtains the spatial distribution projection image corresponding to part surface, then the retrieval of three-dimensional model is realized by similarity comparison, reduces the complexity of comparison, solves the problem that the description data is complex due to view direction uncertainty in the existing three-dimensional model retrieval method, improves retrieval efficiency.

A monocular depth estimation method fusing continuity features comprises the following steps: preprocessing an original image to make implicit relations exist between images; inputting three images at the same time, and obtaining a depth map output through an encoder and a decoder containing four stages and four parallel subnets with different scales; and the loss function comprises a minimum photometric error loss, a smoothing loss, a cross-scale consistency loss and a continuity loss. The application effectively improves the depth measurement precision by combining scale invariance and continuity features on the basis of keeping lightweight, and the training threshold is low due to the self-supervised characteristics of the application, so that the requirement for a data set is low.

The invention discloses a distribution network insulator and accessory defect real-time detection method and system, and belongs to the technical field of intelligent inspection of power equipment. The method comprises the following steps: detecting an insulator image by using a pre-trained defect network model to obtain a defect detection result of a distribution network insulator and an accessory, introducing multi-granularity parallel convolution into a backbone network to obtain context features, and adding an adaptive space reconstruction module after backbone output to adapt to target deformation diversity. A topological receptive field migration mechanism is introduced into a feature fusion network to strengthen key feature expression, and a normalized Wasserstein distance (NWD) is introduced into a loss function to improve scale invariance. The method can accurately recognize various defects such as insulative cover deficiency, damage, contamination and infirm fixation, has the advantages of high detection precision, high real-time performance, flexible deployment and the like, and is suitable for intelligent routing inspection of the power distribution network in a complex environment.

The application provides a scale-invariant image zoom scale mode perception method, comprising the following steps: training a data set of standard scale samples to obtain a trained convolutional neural network; zooming in and out each random scale sample in a test set to form a multi-layer image pyramid composed of multiple scale samples; using the trained convolutional neural network to construct a set of twin convolutional neural networks matched with the number of layers of the multi-layer image pyramid, inputting the multiple scale samples into the set of twin convolutional neural networks for parallel reasoning to obtain a multi-channel classification score matrix; and classifying and estimating the scale according to the classification score matrix. The method has good scale invariance, can classify and estimate the unknown scale at the same time, can be applied to scale-invariant mode recognition tasks of medical images, remote sensing images and the like, and has wider applicability.

The application discloses a kind of based on scale invariance collaborative salient target detection method, it is related to monitoring safety technical field, method includes: obtaining the image sequence converted from video stream in monitoring system, it is input to the main network of pre-training model to extract multi-scale feature;Multi-scale feature is enhanced and fusion processing is generated initial saliency map;By residual refinement structure, restore details edge, generate accurate saliency map;Scale consistency constraint and semantic feature feedback are applied to optimization, obtain target saliency map and output.The application can effectively solve the problem that model lacks semantic guidance, generalization ability is poor in prior art to different resolution images, significantly improve the accuracy and robustness of salient target detection in monitoring system, applicable to a variety of complex monitoring scenes.

This invention relates to a multimodal image matching method and system based on learned features and epipolar geometric constraints. The method includes edge enhancement processing of the input image using wavelet transform; extraction of multi-scale dense feature maps based on a modified convolutional neural network, combined with principal direction normalization to generate feature descriptors with rotation and scale invariance; preliminary feature matching using the FLANN algorithm and dynamic distance constraints; and the introduction of a fundamental matrix construction and epipolar geometric consistency verification mechanism, combined with a RANSAC affine constraint model to eliminate mismatched point pairs. This invention integrates image enhancement, deep learning, and geometric verification strategies, effectively addressing the radiation nonlinearity and geometric distortion problems caused by differences in imaging mechanisms between multimodal images, improving matching accuracy and robustness. It is suitable for remote sensing applications such as optical-SAR registration, multi-source image fusion, and land surface change detection.

This invention provides a method and system for classifying and processing multi-category data of online public opinion using artificial intelligence, belonging to the field of intelligent information processing technology. The method includes: collecting multimodal raw data, preprocessing the raw data and extracting metadata from non-textual data to form a spatiotemporally aligned initial data stream; extracting deep semantic features and sentiment characteristics of the text to form a basic visual feature map; and parsing geometric feature vectors with translation, rotation, and scaling invariance from keyframes of images or videos based on the salient spatial attention regions represented by the basic visual feature map. This invention achieves unified semantic understanding and real-time situational modeling of text, video, and audio data through multimodal spatiotemporally aligned data acquisition, dynamic attention cross-modal fusion, and multi-task fine-grained classification, thereby improving the accuracy and real-time performance of public opinion analysis and its adaptability to dynamic network environments.

This invention discloses a deep learning-based method for efficiently evaluating the magnification factor of gravitational lensing, comprising the following steps: establishing a diffraction integral model of the gravitational lensing system, defining dimensionless frequencies and dimensionless source positions, and converting the physical magnification factor into a universal magnification factor that depends only on the dimensionless parameters to achieve scale invariance of the model; obtaining a model training dataset containing dimensionless parameters and their corresponding ground truth values ​​of the magnification factor; constructing a sinusoidal representation network as the core computational model, and using the dataset to train the sinusoidal representation network; converting the frequency and parameters of the gravitational wave signal to be analyzed into a dimensionless form, inputting it into the trained sinusoidal representation network, and directly outputting the corresponding complex magnification factor for modulation and analysis of the gravitational wave waveform. Compared with traditional numerical integration, the neural network inference speed of this invention is improved by several orders of magnitude, which can meet the needs of real-time gravitational wave data analysis.

The application discloses a kind of deep learning prediction frame matching methods based on feature descriptor, it is related to the identification and matching technology of target object in image recognition field.The present application determines the support area of deep learning identification prediction frame, obtains the feature descriptor with scale invariance by constructing regional gradient, and the recognition result is matched using the formed feature vector.Simulation results show that the present application has strong robustness to translation, rotation, scaling and scale transformation, and has important significance for the development of target tracking, visual navigation and other fields.

The invention discloses a self-balancing economic system based on a fractal network and scale invariance and a control method, and the system comprises a contribution value collection module which is used for obtaining a total contribution value generated by a platform every day; the point issuing calculation module is used for calculating the total daily warehouse locking point issuing amount through a nonlinear issuing formula based on scale invariance according to the total contribution value; the point distribution module is used for distributing the total issuing amount of the warehouse locking points to each account according to the contribution value proportion of each individual and recording the total issuing amount of the warehouse locking points as the warehouse locking points; the integral release module is used for gradually releasing the lock cabin integral into an available integral according to a set decreasing release rule; and the integral attenuation module is used for carrying out attenuation processing on all integrals according to a fixed attenuation rate every day and carrying out equivalent compensation on an available integral attenuation part. According to the invention, self-adjustment is realized based on a biological metabolism scale law and a fractal network theory, and the problems of expansion, unfairness, lack of self-adaptability, insufficient metabolic activity and the like in a traditional integral system are solved.

The invention discloses an asteroid surface feature extraction and matching method, which comprises the following steps of: firstly, extracting candidate feature points with scale invariance and rotation invariance by using an SIFT (Scale Invariant Feature Transform) algorithm, then filtering the SIFT feature points through angular point positions detected by a FAST algorithm, and reserving the SIFT feature points which are close to FAST angular points and are consistent in direction; therefore, the reliability and matching precision of the feature points are improved. According to the method, the problem of feature matching under the conditions of cross-scale, cross-view and violent illumination change can be solved, and the method has high engineering application value.

This invention discloses a method and system for reconstructing urban development boundaries based on Pareto optimization and scale-invariant graph neural networks, belonging to the field of land spatial planning technology. The method acquires multi-source heterogeneous spatiotemporal data of the target area and constructs a directed multi-scale attribute graph; it inputs this data into a pre-trained scale-invariant graph neural network to extract spatial topological heterogeneity representations with scale invariance; based on these representations, it constructs a Markov decision process and a multi-objective fitness function; it constructs a dynamic compliance mask based on preset spatial geometry and expansion capacity hard constraints; finally, it uses a Pareto spatiotemporal multi-criteria decision analysis algorithm to perform branch pruning in real time during heuristic search, outputting a non-dominated boundary vector scheme for the Pareto first front. This invention eliminates the interference of variable surface element problems from a mathematical perspective and significantly reduces computational overhead through early pruning, achieving efficient, automated, and compliant reconstruction of complex spatial boundaries.

The invention discloses a collaborative saliency target detection method based on scale invariance, and relates to the technical field of monitoring security, and the method comprises the steps: obtaining an image sequence converted from a video stream in a monitoring system, and inputting the image sequence into a backbone network of a pre-training model to extract multi-scale features; performing enhancement and fusion processing on the multi-scale features to generate an initial saliency map; a detail edge is recovered through a residual error refining structure, and an accurate saliency map is generated; and performing optimization by applying scale consistency constraint and semantic feature feedback to obtain a target saliency map and outputting the target saliency map. The method can effectively solve the problems that in the prior art, a model lacks semantic guidance for images with different resolutions and is poor in generalization ability, the accuracy and robustness of saliency target detection in a monitoring system are remarkably improved, and the method is suitable for various complex monitoring scenes.

The application discloses a star map recognition method based on a fuzzy neural network, which comprises the following steps: 1. screening navigation stars, a screening method for voting the navigation stars by using random visual axes is designed, and a navigation star library uniformly distributed in the whole sky is established; 2. the star closest to the center of a field of view is taken as a main star, a navigation star feature subgraph based on a minimum spanning tree mode is constructed by taking the main star as a root node and the angular distance between stars as the weight of an edge, and the navigation star feature subgraph has scale invariance and rotation invariance; 3. random noise is added to a feature vector to generate a training set; 4. a neural network is constructed and trained; 5. star map recognition is performed, and the binary number of the feature subgraph is output. Compared with the traditional star map recognition method, the star map recognition method based on the fuzzy neural network has high recognition accuracy and fast recognition speed, and has good robustness to star position noise, star loss and pseudo stars by using fuzzy rules for network learning and adjustment.

The application provides a cross-scale scaled model test design method for a transient strong nonlinear process, belongs to the technical field of structural impact dynamics model test, and is based on a scale similarity conversion model, introduces similarity conversion coefficients and an uncertainty control index therein, and constructs a scale similarity conversion equation from model test results to prototype responses.Combining self-similarity and scale invariance of the strong nonlinear system, discrete constraint is applied to a scaledown ratio of the cross-scale test, scale levels are defined, three types of cross-scale model conversion, i.e., large-middle-large, large-small and middle-small, can still realize cross-scale similarity conversion of strong impact response time histories such as acceleration and strain under the condition that medium parameters, material parameters and boundaries inevitably distort; and the method is suitable for model test data compensation deduction of prototypes of various strong impact response structures such as explosion impact, collision, drop, throwing into water / into soil and impact vibration.

The invention discloses a rotating target detection bounding box regression method based on Kullback Leibler divergence. The method comprises the following steps: step 1, carrying out feature prediction and decoding; step 2, mapping from a rotating frame to two-dimensional Gaussian distribution; step 3, KLD regression distance calculation; 4, performing nonlinear normalization and multi-task joint loss; and 5, in the reasoning stage, only performing'feature prediction and decoding ', and outputting a rotating bounding box by using the trained target detection network. The method has the following advantages: 1, the weight of each parameter of the KLD gradient is naturally and automatically adjusted according to the target size and angle, the index is remarkably improved, and the method is superior to GWD / Smooth L1 especially in AP75, large length-width ratio and dense scenes; 2, the KLD meets affine (including scale) invariance, and small target regression instability is avoided; and 3, when theta is equal to 0, the loss is degraded into common ln norm loss, so that the weight can be conveniently shared under a unified framework.

This application relates to the field of artificial intelligence technologies, and describes a classification model training method, a hyperparameter search method, and an apparatus. The training method includes obtaining a target hyperparameter of a to-be-trained classification model. The target hyperparameter is used to control a gradient update operation of the to-be-trained classification model. The to-be-trained classification model includes a scaling invariance linear layer. The scaling invariance linear layer enables a predicted classification result output when a weight parameter of the to-be-trained classification model is multiplied by any scaling coefficient to remain unchanged. The method further includes updating the weight parameter of the to-be-trained classification model based on the target hyperparameter and a target training manner, to obtain a trained classification model.

The invention discloses an efficient gravitational lens amplification factor evaluation method based on deep learning, and the method comprises the following steps: building a diffraction integral model of a gravitational lens system, defining a dimensionless frequency and a dimensionless source position, converting a physical amplification factor into a universal amplification factor which only depends on a dimensionless parameter, and carrying out the calculation of the universal amplification factor; the scale invariance of the model is realized; obtaining a model training data set containing the dimensionless parameters and amplification factor truth values corresponding to the dimensionless parameters; constructing a sine representation network as a core calculation model, and performing learning training on the sine representation network by using the data set; and converting the frequency and parameters of a gravitational wave signal to be analyzed into a dimensionless form, inputting into the trained sine representation network, and directly outputting a corresponding complex amplification factor for modulation and analysis of a gravitational wave waveform. Compared with traditional numerical integration, the reasoning speed of the neural network is increased by several orders of magnitude, and the requirement for real-time gravitational wave data analysis can be met.