13 results about "Weighted median filtering" patented technology

The invention discloses a marine radar oil spill target segmentation method and system, and relates to the technical field of target segmentation. The method is characterized by comprising the following steps: preprocessing a shipborne radar image; the preprocessing comprises the step of carrying out convolution on the shipborne radar image by adopting a Laplacian operator; an Otsu algorithm is adopted to segment the image after convolution, and same-frequency interference noise is extracted; smoothing the same-frequency interference noise by means of mean filtering to obtain a preliminary smoothed image; segmenting the preliminary smooth image by using a gray threshold algorithm and a pixel area threshold method, and extracting block noise; smoothing the block noise by adopting a dynamic window and distance weighted median filtering to obtain a denoised image; and segmenting the denoised image by using an improved Phansalkar local threshold method, and extracting an oil film. According to the method, the oil spill target features are more prominent by effectively suppressing noise, so that the accuracy of oil spill detection is remarkably improved, and the actual requirements of effective region extraction in the field of oil spill detection in marine radar images are met.

The invention discloses an oil leakage detection and traceability method and system for an oil-immersed power transformer, and belongs to the technical field of transformer detection. Irradiating the surface of the transformer by adopting an ultraviolet light source with controllable power to obtain an oil leakage characteristic image of the surface of the transformer; carrying out oil leakage feature image preprocessing through graying, 3 * 3 convolution kernel Gaussian filtering, 3 * 3 weighted median filtering and a contrast-limited adaptive histogram equalization algorithm in sequence; an improved double-threshold maximum between-class variance algorithm is adopted to determine a background and a fluorescence region, and oil stains are accurately extracted through connected region analysis and texture feature screening; a diffusion main direction is determined through a gradient vector weighted voting algorithm, candidate sources are screened from an easy leakage part database in combination with a K nearest neighbor algorithm, and a final leakage source is determined through gray profile analysis and verification. Through multi-algorithm cooperation, the oil leakage detection precision and traceability accuracy are effectively improved, the method is suitable for daily operation and maintenance of the transformer, and the equipment fault risk is reduced.

The invention relates to the technical field of data processing, in particular to an anti-vibration multi-group camera motion capture system and a data correction method thereof, and the method comprises the following steps: collecting image data of a falling body target through a plurality of infrared motion capture cameras on a plurality of groups of mutually independent metal frame supports; based on spatial reconstruction consistency, performing multi-view cross validation processing on the image data acquired by the plurality of groups of cameras, and identifying and eliminating abnormal parallax data caused by camera vibration; performing trajectory filtering processing on the data subjected to cross validation processing, and performing smoothing processing on trajectory data in combination with Kalman filtering and adaptive weighted median filtering; dividing the filtered trajectory data into a plurality of overlapped time windows according to time; accurate trajectory tracking can be realized in a high dynamic impact scene, and the authenticity and reliability of captured data are improved.

The present invention relates to a depth estimation method for microscopic discrete noise scenes. The method comprises the following steps: 1, using a micron-level stepping motor to collect a multi-depth image sequence of the microscopic discrete noise scene; 2, using a multi-directional Laplacian operator to perform a convolution operation with the image sequence to obtain multiple focal volume results; 3, obtaining multiple initial depth maps based on the location of the maximum focal volume result; 4, screening the initial depth maps based on constraints proposed from the perspective of statistical data stability; 5, fusing the screened depth maps; 6, combining the image sequence with the fused depth map to obtain a fused image; and 7, performing weighted median filtering on the fused depth map and the fused image to obtain a final depth map of the microscopic scene. The method proposed in the present invention can accurately estimate the depth information of microscopic discrete noise scenes.

The present invention relates to the field of data processing technology, and specifically to a vibration-resistant multi-group camera motion capture system and a data correction method thereof. The method comprises the following steps: collecting image data of a falling target by using multiple infrared motion capture cameras on multiple independent metal frame brackets; performing multi-perspective cross-validation processing on the image data collected by the multiple groups of cameras based on spatial reconstruction consistency, identifying and eliminating abnormal parallax data caused by camera vibration; performing trajectory filtering processing on the data after the cross-validation processing, and combining Kalman filtering with adaptive weighted median filtering to smooth the trajectory data; dividing the filtered trajectory data into multiple overlapping time windows according to time; and ensuring accurate trajectory tracking in high-dynamic impact scenarios, thereby improving the authenticity and reliability of the captured data.

The invention provides a GNSS imaging method and device for detecting short-wavelength deformation, and the method comprises the steps: firstly determining the weight of each to-be-estimated point based on a space structure, then extracting an estimated value of a vertical speed, and generating a GNSS image; the step of determining the weight based on the space structure comprises the steps of firstly obtaining a known observation station connected with a to-be-estimated point based on a Delaunay triangle, then obtaining a preliminary weight by using Gaussian process regression, and obtaining a final weight of the connected known observation station after re-weighting based on velocity uncertainty; the step of extracting the estimated value of the vertical velocity comprises the step of extracting the estimated value of the vertical velocity of the to-be-estimated point by adopting generalized weighted median filtering processing based on the final weight of a known observation station connected with the to-be-estimated point. The invention provides a Gaussian process regression GPR-VU scheme considering speed uncertainty, a space structure between a known point and an unknown point is constructed by using a covariance matrix, an earth crust deformation GNSS image in a research area is generated, and detection of short wavelength deformation in an area with complex geology and strong structural activity is realized.

This application provides a method for measuring and analyzing the accuracy of a stereotactic radiotherapy system. The method involves sequentially performing geometric transformations on film images to calibrate the S-L and S-A directions, adaptively correcting positioning errors, purifying the target region through multi-dimensional interference removal, and denoising using normalization and adaptive weighted median filtering. Then, improved fuzzy C-means clustering with fused dual images is used to segment and output label images. Gray-level classification is used to extract the radiation field and tungsten sphere contours. After morphological optimization, the centroid position deviation is calculated. The system accuracy analysis is then completed by combining the S-L and S-A dual film deviations. This application solves the problems of inaccurate contour recognition and large centroid calculation errors in traditional detection methods, significantly improving the accuracy, stability, and robustness of the detection. The steps are clear, the parameters are quantified, and the method is repeatable, providing efficient and accurate technical support for AQA detection in radiotherapy systems.

The invention discloses a fluorescent image-based neural immune antibody screening method and system, which are characterized in that hierarchical features of a fluorescent image are automatically learned through a deep learning model, complex information such as antibody form and distribution is extracted, and various antibodies can be simultaneously detected through learning of a large number of samples; after the existence of the antibody is determined, a preset regression analysis layer is combined with training data, the concentration is accurately inferred according to image features, and manual quantization errors are avoided; an adaptive weighted median filtering and wavelet threshold denoising combined algorithm is adopted in preprocessing, denoising is effectively carried out, image details are reserved, the image quality is improved, and the influence of experimental conditions is reduced; the detection and recognition process is based on model automatic analysis, human interference is reduced, the result is more stable and reliable, the whole process from image acquisition to antibody information determination is automatically completed by a computer, the processing speed is high, automation and high efficiency are achieved, and the detection period is greatly shortened.

The present invention provides a GNSS imaging method and device for detecting short-wavelength deformation. For each point to be estimated, a weight is first determined based on the spatial structure, and then an estimated vertical velocity is extracted to generate a GNSS image. Determining the weight based on the spatial structure includes first determining known stations connected to the point to be estimated based on the Delaunay triangle, then using Gaussian process regression to obtain preliminary weights, and then re-weighting the connected known stations based on velocity uncertainty to obtain the final weights. Extracting the estimated vertical velocity includes extracting the estimated vertical velocity at the point to be estimated using a generalized weighted median filter based on the final weights of the known stations connected to the point to be estimated. The present invention proposes a Gaussian process regression (GPR-VU) scheme that accounts for velocity uncertainty. This scheme uses a covariance matrix to construct the spatial structure between known and unknown points, generating GNSS images of crustal deformation within the study area, and enabling the detection of short-wavelength deformation in areas with complex geology and intense tectonic activity.

The application relates to the technical field of parameter adjustment, and discloses a robot parameter adjustment optimization system and method applied to appearance detection. First, an initial product image set to be detected is acquired, an improved weighted median filtering algorithm and a hybrid detection algorithm are used for target region segmentation and extraction, and a processed product image set to be detected is obtained; second, pixel point coordinates in the processed product image set to be detected are converted into a world coordinate system, spatial coordinates of robot joints in the world coordinate system are determined, and an associated matrix of a robot connecting rod is generated; third, a robot connecting rod end parameter data sequence is determined, a target function is established based on a shortest time principle; and finally, an education competition optimization algorithm is used to solve the target function, optimized robot parameters are generated, and robot parameter adjustment optimization is completed. The application realizes the purpose of robot parameter adjustment optimization by processing and analyzing robot parameters, and the method is objective and accurate.

The invention relates to the technical field of parameter adjustment, and discloses a robot parameter adjustment optimization system and method applied to appearance detection. The method comprises the following steps: firstly, acquiring an initial to-be-detected product image set, and performing target region segmentation and extraction by using an improved weighted median filtering algorithm and a hybrid detection algorithm to obtain a processed to-be-detected product image set; secondly, pixel point coordinates in the processed to-be-detected product image set are converted into a world coordinate system, space coordinates of robot joints in the world coordinate system are determined, and an incidence matrix of robot connecting rods is generated; determining a robot connecting rod tail end parameter data sequence, and establishing a target function based on a shortest time principle; and finally, solving the objective function by using an educational competition optimization algorithm, generating optimized robot parameters, and completing adjustment and optimization of the robot parameters. According to the method, the robot parameters are processed and analyzed, the purpose of adjusting and optimizing the robot parameters is achieved, and the method is objective and accurate.

The present invention discloses a method and system for segmenting oil spill targets of a marine radar, which relates to the technical field of target segmentation. The technical key points of the present invention include: preprocessing the shipborne radar image; the preprocessing includes convolving the shipborne radar image with a Laplacian operator; using the Otsu algorithm to segment the convolved image and extract the co-frequency interference noise; using mean filtering to smooth the co-frequency interference noise to obtain a preliminary smoothed image; using a gray threshold algorithm and a pixel area threshold method to segment the preliminary smoothed image and extract the block noise; using a dynamic window and distance-weighted median filtering to smooth the block noise to obtain a noise-reduced image; using an improved Phansalkar local threshold method to segment the noise-reduced image and extract the oil film. The present invention effectively suppresses noise, makes the characteristics of the oil spill target more prominent, thereby significantly improving the accuracy of oil spill detection and meeting the actual requirements for extracting effective regions in the field of oil spill detection in marine radar images.