16results about How to "Suppress artifacts" patented technology

The invention relates to the technical field of sewage treatment, and discloses a sewage biochemical treatment process abnormity intelligent early warning method and system based on image recognition, and the method comprises the steps: periodically collecting a sludge image and water quality sensing data; extracting floc form, bubble distribution and color texture features of the image, and performing space-time alignment fusion with the sensing data to construct a multi-source fusion feature matrix; inputting the matrix into a pre-trained digital twin evaluation model, calculating a dynamic deviation degree between a current state and a normal working condition, and identifying an abnormal mode; and generating a graded early warning signal according to the deviation degree and the abnormal mode and pushing the graded early warning signal. The system comprises an online image acquisition unit and a water quality sensor array which are used for data acquisition, and a feature processing module used for feature processing and fusion. According to the scheme, early-stage, accurate and intelligent early warning of abnormality in the sewage biochemical treatment process is realized, the false report and missing report rate is reduced, and the accuracy and the reliability of the system are improved. And the operation stability and the regulation and control efficiency of the system are improved.

The present application belongs to the field of CT tomographic reconstruction technology and artificial intelligence, and discloses a limited-angle CT reconstruction artifact suppression method based on a multi-domain feature fusion network. In view of the problem that the reconstruction result of traditional CT scanning under limited-angle conditions is prone to artifacts and structural distortion, thereby affecting the image quality and defect detection accuracy, the present application constructs a multi-domain feature fusion artifact suppression network, takes the limited-angle reconstruction result as input, and realizes artifact suppression and detail recovery through the synergistic effect of the encoder part, the decoder part, the feature enhancement part and the feature conversion part. The present application can obtain high-quality tomographic images under limited-angle conditions, effectively reduces the scanning angle and time of industrial CT detection, improves the imaging clarity and reliability without increasing the radiation dose, is suitable for industrial detection of complex structure workpieces, and has important industrial application value.

The application discloses a point cloud driven thoracic cavity whole organ dynamic reconstruction and respiratory monitoring method and system, and the method implementation comprises the following steps: based on the thoracic medical image data of a target object, constructing three-dimensional geometric models corresponding to each target structure in the thoracic cavity, and converting the three-dimensional geometric models into static point cloud data; based on the respiratory motion law of the target object, driving the static point cloud data to perform dynamic deformation simulation in a respiratory cycle, and obtaining dynamic point cloud data synchronized with the respiratory phase; synchronously associating the dynamic point cloud data with electrical characteristic parameters in time and space, and constructing a digital twin thoracic cavity model; deploying a virtual sensor assembly that dynamically deforms with the thoracic cavity in the digital twin thoracic cavity model, and simulating and monitoring the dynamic respiratory process of the target object, and generating a virtual physiological signal; and through signal processing on the virtual physiological signal, a dynamic reconstruction image reflecting the respiratory motion is obtained. The continuous simulation of the whole respiratory motion process is realized, and the time and space resolution and diversity of the virtual database are improved.

The invention belongs to the field of computer vision, and relates to an alternating iteration straight line matching and image splicing method. After pre-registration based on matched feature points is carried out by using an existing algorithm, firstly, global line-guided grid deformation is carried out to obtain a preliminary splicing result; secondly, obtaining line matching pairs through a new straight line matching method which does not depend on descriptors; then, point matching pairs are obtained from the line matching pairs according to the invariant feature number, and the line matching pairs and the point matching pairs return to the previous step to jointly guide global line guide grid deformation. Global line guide grid deformation and line point matching are continuously and alternately iterated. Finally, the optimal line point matching and the best splicing result are obtained, more accurate matching feature pairs can be generated through the result, clearer textures can be displayed, a remarkable natural structure in a spliced image is reserved, and the effect is superior to that of a traditional image splicing method and a latest method based on deep learning.

The application relates to a precondition matrix-based fast limited-angle fan-beam CT image reconstruction method and belongs to the technical field of CT image reconstruction. The method comprises the following steps: S1, acquiring projection data; S2, considering the sparsity of an image under tight wavelet frame transformation and image gradient transformation, establishing a double-regularization limited-angle fan-beam CT image reconstruction model to inhibit noise and artifacts and protect a boundary; S3, according to an analytical reconstruction algorithm, designing a precondition matrix, introducing the precondition matrix into the double-regularization limited-angle fan-beam CT image reconstruction model established in the step S2, and solving the model by using an alternating direction multiplier method; and S4, outputting a reconstructed image. The precondition matrix technology is used to integrate the analytical reconstruction algorithm into the regularization iterative reconstruction algorithm, fast convergence of the algorithm is realized, artifacts and noise of the limited-angle fan-beam CT reconstructed image can be effectively inhibited, and therefore the quality of the CT reconstructed image is improved.

The embodiment of the invention provides a bone microstructure parameter extraction method, computing equipment, a computer readable storage medium and a computer program product. The bone microstructure parameter extraction method comprises the following steps: firstly, carrying out super-resolution reconstruction on a CT (Computed Tomography) image to enhance the image quality; then, a first bone mask is generated through a global threshold value, a local adaptive threshold value is constructed in combination with local gray statistical characteristics, and a mask edge and a low-contrast region are subjected to reclassification to obtain a more accurate second bone mask; and then, noise connected domain elimination under volume constraint is performed on the second bone mask, and a fracture or a pseudo joint of the bone trabecula network is repaired based on morphological characteristics to obtain a high-fidelity bone tissue structure segmentation result. And finally, cortical bones are removed from the result to separate out cancellous bones, and bone microstructure parameters are extracted from the cancellous bones. According to the technical scheme provided by the embodiment of the invention, the accuracy of bone microstructure parameter extraction is improved.

The invention discloses a dual-camera image super-resolution method based on feature matching and confidence guidance attention fusion, which belongs to the technical field of dual-camera image processing and comprises the steps of wide-angle image feature extraction and up-sampling; feature coding and optical flow estimation; aligning the features guided by the optical flow; a multi-candidate feature matching module; a confidence coefficient guided deformation module; performing high and low frequency decomposition and feature modulation; a confidence degree guiding cross attention module; feature fusion and image reconstruction; according to the method, the robustness and accuracy of feature matching are effectively improved, and mismatching is effectively inhibited under the condition that the weak similarity and the resolution are different.

A robust demosaic bilinear interpolation method for polarization hyperspectral imaging is presented, addressing issues such as optical system platform vibration, frame loss in acquired images, and spectral aliasing in existing technologies. The method includes acquiring a set of polarization hyperspectral images and performing robust demosaic reconstruction on them. Image I is the acquired image of the Kth spectral band, and mosaic image II is an empty matrix used as the interpolation space for image I. Horizontal values ​​are assigned to mosaic image II, and boundary conditions are applied for filling to complete the horizontal interpolation. The intensity images of the four polarization channels in different directions in mosaic image II are extracted, and vertical interpolation is performed on these polarization channel intensity images. This completes the polarization demosaic reconstruction of the Kth spectral band image. This method is applicable to suppressing image artifacts generated during reconstruction.

The invention provides an infrared image enhancement method and device based on histogram matching and curvelet transformation. The method comprises the following steps: acquiring a to-be-enhanced infrared image and a reference image; respectively calculating a gray average value and a gray standard deviation of the infrared image and the reference image; calculating a correction coefficient and an average correction value based on the gray average value and the gray standard deviation; generating a histogram matching enhanced image of the infrared image based on the correction coefficient and the average correction value; curvelet transform decomposition is carried out on the histogram matching enhanced image and the infrared image to obtain corresponding multi-scale and multi-direction sub-band coefficients, and fusion is carried out on the corresponding sub-band coefficients under the same scale and the same direction to obtain a fusion sub-band coefficient; and performing curvelet inverse transformation on the fusion sub-band coefficient to obtain a fusion enhanced image. According to the invention, through the cooperation of the global gray correction of histogram matching and the local detail fusion of curvelet transformation, the contrast, definition and detail information of the infrared image are synchronously improved.

This invention discloses a method for detecting the compliance of clothing color in ID photos, belonging to the field of image processing technology. The method first performs portrait parsing and face detection on the input user image, generating extended candidate regions and cropping them to obtain local images and local portrait parsing mask images. Next, the clothing region is located and processed through erosion, partitioning, and exclusion of specific regions. Illumination normalization, automatic white balance, and optical anti-aliasing are then performed on the local image sequentially. Finally, based on the proportion of non-zero pixels in the accurate clothing mask image, color features are calculated in combination with the processed image and the mask, and a color feature vector is obtained according to a fusion strategy. In cases of occlusion, an empty vector is returned. This invention improves the accuracy and robustness of light-colored clothing detection, eliminates the influence of illumination and interference areas, adapts to different scenarios, and meets the requirements for compliance detection of clothing color in ID photos.

The invention provides a training method and device for determining a network model of underground reflectivity. The method comprises the following steps: acquiring observation system information and seismic source wavelets of seismic imaging; generating an initial migration image according to the Sigbee speed model, the observed seismic data and the initial data; determining an approximate fuzzy operator according to the smooth velocity model, the scattering point model and the initial data; performing deblurring processing on the initial offset image through a convolutional neural network model to obtain approximate subsurface reflectivity; performing non-stationary two-dimensional convolution processing on the approximate subsurface reflectivity and the approximate fuzzy operator to obtain an approximate offset image; and iteratively updating the network parameters of the convolutional neural network model to minimize the error between the approximate offset image and the initial offset image to obtain a target convolutional neural network model. By applying the method, the imaging precision of the underground shallow layer and the model boundary is improved, and artifacts between bottom layers in migration imaging are inhibited.

The invention discloses a Qinghai-Tibet Plateau global climate mode rainfall downscaling method based on double attention, and relates to the technical field of rainfall downscaling, and the method comprises the following steps: obtaining coarse resolution global climate mode CMIP6 rainfall data, high resolution reference rainfall data and multi-source auxiliary data, and carrying out the standardization preprocessing of all data; a generative adversarial network model is constructed, and a generator of the generative adversarial network model adopts a dual attention mechanism embedded with a channel attention module and a space attention module and is connected in series with a dense block to form a feature extraction main body; training the generative adversarial network model by adopting a two-stage sequential downscaling strategy; and inputting to-be-downscaled CMIP6 rainfall data into the final model, and outputting to obtain a high-resolution downscaled rainfall product. According to the method, the generative adversarial network fusing dense connection and a double attention mechanism is combined with the multi-scale discriminator, so that high-precision downscaling of the CMIP6 rainfall data is realized.

This invention discloses a high frame rate ultrasound plane wave tensor completion method based on T-SVD and angle regularization, belonging to the field of ultrasound imaging technology. The method includes: acquiring data; encoding selected and unselected angles into TCP codes; transmitting the encoded image along a specified angle; constructing a mask matrix Ω to form an incomplete tensor X containing partially known data; zero-padding the acquired RF data to expand it to the full size; reconstructing the data corresponding to the missing angles from the incomplete tensor; iteratively solving the problem using the alternating direction multiplier method; determining whether the convergence error convergence criterion has been met; and, after iteration, completing the tensor X to obtain the full angles. Composite imaging is then performed based on the obtained complete RF data. This invention solves the problems of slow convergence speed, high computational complexity, and unsmooth completion results in existing tensor completion algorithms. Under conditions of extremely limited angle sampling, this invention achieves a balance between high frame rate and high quality in ultrasound imaging through efficient low-rank tensor completion.

The application discloses a few-step sampling high-texture-fidelity remote sensing image super-resolution method based on a diffusion model, and relates to the technical field of image data processing. In the forward process, a residual scheduling parameter and a noise scheduling parameter derived therefrom are introduced, so that the noise intensity is first enhanced and then attenuated; in the reverse process, a low-resolution interpolation image is taken as initial input, and iterative reconstruction is performed based on the residual scheduling parameter, the noise scheduling parameter and a reconstruction result output by a prediction model. The prediction model adopts a UNet architecture, and the encoding layer, the intermediate layer and the decoding layer are all constructed based on a noise residual attention module. The module performs multi-scale feature extraction, convolution processing and attention enhancement through multi-branch parallelism. The application can obtain a high-texture-fidelity super-resolution reconstruction result under a small number of sampling steps.

The invention relates to a low-field magnetic resonance diffusion imaging reconstruction method. The method comprises the following steps: acquiring K space data of low-field magnetic resonance; mapping the acquired K space data to a standard Cartesian coordinate system by adopting a grid interpolation method of a Gaussian kernel, and weighting each radial sampling point through a Gaussian kernel function to ensure that the data is smoothly distributed on a standard grid; image reconstruction is carried out based on image prior constraints by using the interpolated K space data, and the image prior constraints comprise a low-rank structure and sparse prior information of the image; parameters of low-rank and sparse components are optimized step by step by using a mixed kernel function, and constraint is carried out by combining a low-rank structure and sparse prior information of an image, so that the image reconstruction effect is improved. The invention also relates to a low-field magnetic resonance diffusion imaging reconstruction system. According to the invention, the image quality and the anti-noise performance can be obviously improved.

The application provides an infrared image enhancement method and device based on histogram matching and curvelet transform. The method comprises: acquiring an infrared image to be enhanced and a reference image; calculating the gray mean value and the gray standard deviation of the infrared image and the reference image respectively; calculating a correction coefficient and an average correction amount based on the gray mean value and the gray standard deviation; generating a histogram matching enhanced image of the infrared image based on the correction coefficient and the average correction amount; respectively performing curvelet transform decomposition on the histogram matching enhanced image and the infrared image to obtain corresponding multi-scale and multi-direction sub-band coefficients, and fusing corresponding sub-band coefficients at the same scale and the same direction to obtain fused sub-band coefficients; and performing curvelet inverse transform on the fused sub-band coefficients to obtain a fused enhanced image. Through the cooperation of the global gray correction of histogram matching and the local detail fusion of curvelet transform, the application realizes the synchronous improvement of the contrast, the definition and the detail information of the infrared image.