8 results about "Rotational invariance" patented technology

The application discloses a point cloud 3D object detection method not influenced by rotation transformation, comprising the following steps: (1) regarding the network weight of the first layer of the neural network as a vector set distributed in a feature space with the same dimension as the point cloud feature; (2) performing seed point sampling and neighborhood aggregation on the input point cloud data to obtain the local point cloud around each seed point; (3) performing principal component analysis on the network weight and the local point cloud; (4) aligning the weights of the network weight and the local point cloud to obtain a feature with rotation invariance; (5) inputting the local point cloud feature of step (4) into the neural network for feedforward transmission, detecting the prediction of the head output 3D object frame of the network; (6) training the neural network through gradient back propagation; and (7) after the training is completed, performing a 3D object detection task on the point cloud. According to the application, the classification accuracy of the point cloud under arbitrary rotation transformation can be greatly improved, and thus the accuracy of the 3D object detection task can be improved.

The application provides a two-dimensional DOA estimation method for arbitrary array single-base MIMO radar based on data rearrangement, and relates to the technical field of array sensor direction finding. The rank structure of a coherent signal covariance matrix is recovered by rearranging a received data matrix through arbitrary array element positions of a transmitting array and a receiving array; S2: eigenvalue decomposition is performed on the signal covariance matrix to obtain a signal subspace; S3: reliable rough estimation is obtained through rotation invariance; S4: an integer ambiguity vector corresponding to spatial phase ambiguity caused by array element spacing greater than half a wavelength is solved; S5: the phase of an array flow pattern with phase ambiguity is compensated, and an accurate estimation value of the 2D-DOA of the target is calculated; the high-precision 2D-DOA estimation problem in a complex scene of a coherent signal source, arbitrary array geometry and phase ambiguity caused by part of array element spacing greater than half a wavelength is solved, unambiguous and high-precision angle parameter estimation is realized, and the estimation accuracy and algorithm reliability are improved.

This invention discloses a polarization-agile interference suppression method based on a polarization-frequency-controlled array radar, belonging to the field of radar anti-jamming and array signal processing technology. Under the PFDA-MIMO radar system, this invention achieves spatial position parameter estimation and globally optimal self-pairing for polarization-agile interference by utilizing the array spatial manifold rotation invariance and cyclic measure algorithms. Furthermore, it combines spatial decoupling to eliminate crosstalk between multiple interferences and uses the least squares criterion to obtain instantaneous polarization information under a single snapshot. It further reconstructs the instantaneous interference plus noise covariance matrix under a single snapshot, obtains the optimal weight vector, and performs joint space-polarization adaptive filtering. In the scenario of polarization-agile interference, it achieves high-fidelity target extraction and deep cancellation of interference energy, effectively improving the system's output signal-to-interference-plus-noise ratio and target detection capability.

The application discloses a pollen image feature extraction and recognition method based on deep learning, relates to the technical field of computer vision and image recognition, and comprises the following steps: an image preprocessing step: obtaining a target pollen image and converting the target pollen image into a single-channel grayscale data matrix; a manifold field reconstruction step: inputting the matrix into a surface reconstruction encoder to generate a pseudo-three-dimensional geometric feature map representing surface space structure; a topological feature extraction step: using a geodesic line convolution unit to perform non-Euclidean space sampling and extracting surface geometric curvature features with rotation invariance; and a Riemann classification recognition step: inputting the features into a classifier based on Riemann metric, calculating geodesic line distance in a manifold space, and determining a category label; the application solves the problem of topological structure information loss caused by rotation of a two-dimensional image by constructing a pseudo-three-dimensional geometric feature map and applying geodesic line convolution, and high-precision recognition of close species is realized.

The application discloses a kind of based on graph semantic redirection module's remote sensing image ship target detection method and system, comprising: input remote sensing image to target detection model, carry out feature extraction by stacked convolution block, obtain shallow, middle and deep feature map and utilize feature pyramid network to carry out feature fusion;The improved detection head of feature map after fusion is introduced into graph semantic redirection module, graph semantic inference is carried out by introducing multilayer perceptron, pixel-level semantic similarity is used to construct graph structure to extract key features with rotation invariance, global semantic consistency integration is carried out using the feature integration mechanism based on exponential moving average, decoupling feature map obtains target class and position information;Target detection model is trained and tested for target detection task, and ship target detection is carried out on the remote sensing image to be measured;The application realizes task feature decoupling in detection head stage and enhances the extraction of rotation invariance semantic features, so as to realize high-precision remote sensing ship detection.

This invention discloses an euploidy prediction method based on the fusion of blastocyst developmental dynamics and spherical harmonic decomposition. The method acquires multifocal plane image sequences of blastocysts using a time-difference imaging system and extracts dynamic parameters. After semantic segmentation and depth estimation, a three-dimensional surface model is obtained through surface reconstruction. The trophoblast cell instances are segmented to obtain three-dimensional centroid coordinates. These centroid coordinates are radially projected onto a unit sphere, and a spherical density function is constructed after excluding the inner cell mass mask. Spherical harmonic decomposition is then normalized using Monte Carlo zero-model normalization to obtain normalized power spectra at each degree. The basic morphological features, dynamic parameters, and spherical harmonic features are fused and filtered before training an ensemble learning model to output prediction results. This invention is the first to introduce spherical harmonic analysis into blastocyst assessment, capturing multi-scale spatial distribution information that global statistics cannot obtain. The power spectrum exhibits rotational invariance, and zero-model normalization eliminates cell number confounding.

The application provides a MIMO radar extended target adaptive detection method based on information geometry, and belongs to the technical field of radar target detection. Firstly, the statistics of three kinds of adaptive detectors are derived based on GLRT, Wald and Rao detection criteria under the condition that the clutter covariance matrix is known; then, the information geometry theory is used to model the clutter covariance matrix estimation problem on the matrix manifold; secondly, based on a geometric measure with rotational invariance, an optimization solution method is designed to obtain the clutter covariance matrix estimation result; finally, the clutter covariance estimation value is substituted into the GLRT, Wald and Rao detection statistics to obtain the adaptive detector, which is compared with the threshold to make a judgment, so that target detection is realized. The application depends on less auxiliary data, can effectively realize target detection under non-Gaussian clutter, and is suitable for practical application.

The application discloses an unsupervised PET image reconstruction method based on a learnable gradient descent, a regularization term is designed by using an L2,1 norm of a convolutional neural network, and the regularization term is subjected to a smoothing approximation process, so that an explicit solution of the regularization term gradient descent can be obtained. Based on this, the application designs an LDA network composed of multiple modules, which has strong prior learning ability and strong interpretability. At the same time, the application proposes a double-domain unsupervised loss composed of an image domain loss and a measurement domain loss, wherein the image domain loss is designed by using rotational invariance, and the measurement domain loss is subjected to a random noise data enhancement operation, thereby solving the problem that a large number of labeled images are required for training in mainstream methods, and thus the application has better clinical feasibility. The application can reconstruct a high-quality PET image from low-count Sinogram projection data, and excellent reconstruction effect is verified on clinical data.