29results about How to "Avoid artifacts" patented technology

The invention provides a light field microscope three-dimensional image reconstruction system and method integrated with an event camera, and belongs to the technical field of optics. According to the invention, original light field data is synthesized into event data for training a network, the event data is input into an event image reconstruction network module capable of converting continuous event streams into a series of intensity images to obtain light field event images, and then three-dimensional reconstruction of a target area is obtained through a three-dimensional image reconstruction network module. And a new technical reference is provided for solving the problem of how to improve the three-dimensional reconstruction quality and resolution of the light field microscope.

The invention provides a laser confocal ocular surface image stitching method based on deep learning. Efficient and accurate image stitching is realized through unsupervised learning. Firstly, an ocular surface image is preprocessed, and a data set is constructed; then, a multi-scale feature pyramid is constructed by using a convolutional neural network, image features are automatically extracted and pre-aligned, and image alignment is optimized through global motion aggregation and local motion correction technologies. Thirdly, constructing a content fusion generative adversarial network which comprises a generator and a discriminator to smooth the fused image and generate a spliced image conforming to human eye perception; a staged training strategy is adopted, image alignment is optimized firstly, and then image smooth fusion is achieved. And finally, automatically finishing image alignment and fusion by using the trained network to obtain a seamless large-view confocal ocular surface image. According to the method, the splicing precision and robustness are improved, the method is suitable for image splicing in different environments, and the splicing success probability is remarkably improved.

The invention discloses a turbine blade surface field rapid prediction method based on a multi-channel self-attention network, and relates to the technical field of data prediction. The technical problems that in the prior art, an existing turbine blade surface field prediction technology is long in consumed time, high in calculation cost and insufficient in precision, and obvious temperature and pressure distribution nonuniformity exists are solved. According to the technical scheme, the turbine blade surface field rapid prediction model building method based on the multi-channel self-attention network comprises the steps that turbine blade surface field data are collected; collecting a preset multi-channel self-attention network architecture; performing pretreatment; performing feature extraction and modeling on the turbine blade surface field according to the processed multi-channel self-attention network architecture; and according to the feature extraction and modeling results, establishing a'mapping relation from boundary conditions to a temperature field and a pressure field on the surface of the turbine blade '. The method is suitable for predicting the temperature field and the pressure field on the surface of the turbine blade.

The invention discloses a method for remotely measuring gas concentration two-dimensional distribution based on an absorption spectrum technology, and belongs to the technical field of gas monitoring and remote measurement. On the basis that the unmanned aerial vehicle is used for mounting a telemetering laser receiving and transmitting device to inspect the natural gas pipeline along the line, as the natural gas pipeline is buried underground, when a leakage point appears on the natural gas pipeline, a leaked natural gas mass is formed on the ground surface where the natural gas pipeline is buried; a remote measuring laser receiving and transmitting device carried by the unmanned aerial vehicle is used for detecting an area where the natural gas concentration exceeds a threshold value, and therefore the natural gas pipeline leakage point position is judged. According to the method, the high natural gas area can be more accurately positioned, so that the pipeline leakage point is deduced, the local constraint variable strategy is adopted, the resolving stability is improved, and artifacts or invalid data are avoided. And real-time generation and updating of a concentration distribution diagram are realized through simultaneous scanning and resolving.

The application discloses a kind of autonomous ultrasonic robot multimode drag control method and device.The method comprises: obtaining real-time ultrasound image and the force data value of mechanical arm end;Determine whether the pixel value of ROI region of the ultrasound image of continuous preset frame number is greater than preset pixel threshold value: if yes, according to force data value and first initial control parameter, it is calculated to obtain the target displacement amount in X axis direction and Y axis direction, simultaneously, according to force data value, preset expected force and second initial control parameter, it is calculated to obtain the target displacement amount in Z axis direction;Tool coordinate system under target displacement amount is converted into world coordinate system under target displacement amount;Zero force control is implemented in the horizontal direction of probe contact surface, and constant force control is implemented in the vertical direction of probe contact surface, until probe moves to target position.The method improves the operation flexibility and scanning safety by the synergistic effect of zero force control and constant force control.

This invention discloses a multi-channel electromagnetic induction detection method and system based on uniform magnetic field constraint. A uniform magnetic field is constructed within the detection area of ​​the structure under test. Electromagnetic induction response signals are acquired through multiple channels under the constraint of the uniform magnetic field, and the spatial coordinates of each acquisition channel are marked. The raw electromagnetic induction signals acquired by each channel are preprocessed, including noise reduction filtering, envelope extraction, and amplitude normalization. A deep learning model is then used to extract structural features from the preprocessed multi-channel electromagnetic response signals. Based on the spatial position information of each acquisition channel, magnetic field uniformity parameters, and response characteristics, a channel correlation weight matrix is ​​constructed to describe the spatial consistency and physical coupling relationships between channels. A multi-channel joint optimization model is constructed, and the equivalent electromagnetic parameter distribution inside the structure under test is reconstructed through iterative solution. This invention improves the reliability, repeatability, and engineering applicability of electromagnetic induction detection results.

This application discloses an image processing method and apparatus in the computer vision field of the artificial intelligence field. The image processing method includes: obtaining a channel image corresponding to a to-be-processed raw (raw) image, where the channel image includes at least two first channel images; performing image restoration processing on the first channel image to obtain the restored first channel image; and performing image restoration processing on the to-be-processed raw image based on the restored first channel image to obtain a raw image obtained through joint demosaicing and denoising processing. The technical solution of this application can improve image quality of the raw image obtained through joint demosaicing and denoising processing.

The present application relates to the technical field of image processing, and more particularly to a projection image optimization method based on image processing, which first extracts image block samples from a target scene, uses an industrial camera to collect data and combines a texture separation model to accurately extract features to determine key texture distortion anchor points. Secondly, according to these anchor points, the non-flat wall surface is divided into multiple distortion transmission areas to realize local fine modeling. Then, the preset imaging parameters are independently optimized in multiple channels for different areas to generate projection correction parameters adapted to each area. Finally, the optimized parameters are used to control the imaging of the projection equipment. This method effectively decouples the inherent texture of the wall surface and the interference of the projection content, solves the problems of non-linear distortion, uneven brightness and blurring caused by complex curved surfaces through adaptive correction by area, significantly improves the clarity, contrast and visual consistency of the projection image, and does not require expensive three-dimensional scanning hardware, with high robustness and low cost advantage.

The present invention discloses a multi-exposure image fusion method based on a multi-dimensional collaborative refinement network, including: constructing a collaborative extraction module CEM to obtain shallow features and deep features of an image; obtaining an attention feature map through a self-attention module SAM; correcting the extracted edge texture of image features using an edge correction module ERM; and using a combination of two loss functions to train an image fusion network, wherein the image fusion network includes a collaborative extraction module CEM, a self-attention module SAM, and an edge correction module ERM. In order to improve the quality of the fused image and avoid artifacts, the present application designs a collaborative extraction module to extract and optimize image features. With the cooperation of the self-attention module, the texture, brightness, and color of the multi-exposure fused image are more prominent; the edge correction module complements CEM and SAM, and it can reduce the impact of noise on features at different scales, enhance the texture details of the feature map, and ensure the edge effect of the fusion result.

The utility model provides a drawstring type close-up net bag set for breast reconstruction surgery, which comprises prosthesis beads, a bag body, a bag mouth and a drawstring embedded in the bag mouth, and the bag body is latticed. According to the utility model, the silica gel prosthesis bead is stored through the draw-cord type closing-up net bag, so that the edge of the prosthesis is smooth and stable, the size of the prosthesis can be dynamically adjusted in an operation, and the device has the advantages of simplicity and convenience in operation, short operation time and few complications.

The invention relates to a method for controlling a computed tomography system (1), comprising the following steps: defining an examination region (U) of an examination object (6); defining a number of interfering structures, wherein the interfering structures are structures located within the beam path of the computed tomography system (1) and reflecting or absorbing X-ray radiation having an intensity beyond the tolerance range; calculating a beam path of the X-rays (R) of the CT examination of the examination region (U) and determining a disturbance angle range (W) of the beam path through the disturbance structure (S) and a gantry (2) of the computed tomography system on which the examination region (U) is located; cT recording is executed while the rack (2) rotates, and CT data (D) is recorded; reconstructing the image data set (B) from the CT data (D) without the determined disturbance angle range (W); and outputting the image data set (B). The invention also relates to a control device and a computed tomography system.

The present disclosure proposes a light source device that generates illumination light to be applied to an object in order to enable light amount control to be executed quickly in response to a change during a rolling shutter period (pixel readout period) while avoiding the occurrence of distortion and artifacts caused by a rolling shutter of an image sensor, and includes a plurality of semiconductor light emitting elements that emit pieces of light having different wavelength bands, and a control unit that controls a light emission profile of the plurality of semiconductor light emitting elements and drives the plurality of semiconductor light emitting elements, in which the control unit controls the light emission profile of the plurality of semiconductor light emitting elements such that main light emission is performed in a pseudo global exposure period of an image sensor and preliminary light emission is performed in at least a part of a pixel readout period of the image sensor, and a light emission level of the preliminary light emission is lower than a light emission level of the main light emission

The present invention belongs to the technical field of high-energy industrial detection, and relates to a high-energy industrial CT detection device for a casting casing, wherein a conformable base is arranged on a CT equipment turntable in a CT room, the transmission chamber is a hollow rectangular parallelepiped, and the transmission chamber is vertically fixed on the upper surface of the conformable base; the power supply and the servo motor are arranged on the upper surface of the conformable base and close to the transmission chamber, the power supply is used to power the servo motor, the servo motor output shaft extends into the transmission chamber and a synchronous pulley is provided on the servo motor output shaft; the reducer is arranged at the upper end of the transmission chamber, the reducer is connected to the synchronous pulley on the servo motor output shaft through a transmission belt, the reducer is also coaxially connected to a clamping module, and the clamping module is used to fix the casting casing; the servo motor drives the reducer to rotate through the transmission belt, thereby driving the clamping module to rotate with the casting casing.

This invention discloses a target trajectory background modeling method and system based on bidirectional difference and grayscale correction, belonging to the field of image data processing technology. The method includes: S1, periodically acquiring basic image trajectory data and performing preprocessing on the acquired data; S2, constructing a frame index sequence, generating a frame pairing table and a grayscale coefficient table, performing grayscale correction and bidirectional difference calculation on the paired frame sequence to obtain the candidate motion mask for the current frame; S3, obtaining the initial background region of the current frame, evaluating the background confidence level of each pixel, and generating an effective background region; S4, comprehensively evaluating the stability level and fluctuation constraint capability of the background, dividing the stable and weak background regions, and outputting the frame-level motion mask to complete the background closed-loop output. This solves the problem that traditional background modeling methods directly use arithmetic mean modeling in a few-frame sequences, leading to the background grayscale being lowered by high-contrast moving pixels, thus generating motion artifacts and causing weak target detection failure.

The present invention discloses a method for fusion of infrared and visible light under weak registration and a binocular imaging device, the method comprising: obtaining a visible light image and an infrared image for the same target, using an image weak registration method to generate a target image pair; inputting the target image pair into a trained deep learning fusion model to obtain an output fusion image; the deep learning fusion model comprises a significant feature extraction module, an encoder, a fusion module and a decoder; in the binocular imaging device, a visible light camera and an infrared camera capture visible light images and infrared images respectively, an AI processor first corrects and crops the infrared image and the visible light image, and then uses a deep learning fusion model to perform fusion processing, and obtains a fusion image to be displayed on a terminal display. The present invention improves imaging quality and allows the fused image pair to have registration deviations, effectively highlights significant information, reduces visual artifacts, and is more conducive to the all-weather operation of the imaging system.

The invention relates to an image classification high-concealment back door sample generation method in combination with space-frequency domain features, and belongs to the field of artificial intelligence safety. The method comprises the following steps: firstly, extracting a high-frequency component of a trigger by using wavelet transform, embedding inverse transformation information of the high-frequency component into an original image, and converting into an initial trigger image spectrum; extracting phase spectrums and amplitude spectrums of the trigger and the original image by using FFT, and recombining the phase spectrums of the trigger and the amplitude spectrums of the original image to generate a reconstructed trigger image spectrum; performing linear addition and inverse transformation on the initial trigger image spectrum and the reconstructed trigger image spectrum to generate a primary trigger image; and finally, taking the original image and the primary trigger image as training samples, and training by using a conditional GAN model to generate a back door sample generator. Aiming at the problem that the concealment of the backdoor sample is influenced by image blurring and distortion easily caused by embedding trigger full-frequency information into an original image in the existing method, the attack success rate of the backdoor sample is effectively improved by using trigger high-frequency characteristics and original image spatial domain characteristics as supervisory information training models.

The present invention discloses a tissue-mimicking phantom for omnidirectional slice thickness detection of ultrasonic tomography equipment. The tissue-mimicking phantom is in a truncated cone shape and comprises: a phantom shell, an acoustic window (3), a background tissue-mimicking material (9) and a target; the phantom shell is fixed by an upper panel (2), a lower panel (6) and a truncated cone target seat (8) therebetween; the phantom shell and the acoustic window (3) pasted on the side of the truncated cone together form a closed space, the closed space is filled with background tissue-mimicking material (9), the target is a diffuse scattering layer (7) closely attached to and covering the side of the truncated cone target seat (8); the upper panel (2) is provided with an inlet for filling the background tissue-mimicking material (9), and a sealing rubber is pasted on the upper panel. The tissue-mimicking phantom of the present invention can be used for omnidirectional slice thickness detection and evaluation of ultrasonic tomography equipment.

This application provides a picture encoding method and apparatus, and a picture decoding method and apparatus. The picture encoding method in this application includes: determining a reference frame number of a current picture based on channel feedback information, where the channel feedback information indicates information about a picture frame received by a decoder side; obtaining a first reference layer number set of a first picture frame corresponding to the reference frame number, where the first reference layer number set includes layer numbers of N1 layers, 1≤N1<L1, and L1 indicates a total quantity of layers of the first picture frame; determining a reference layer number of the current picture based on the channel feedback information and the first reference layer number set; and performing video encoding on the current picture based on the reference frame number and the reference layer number to obtain a bitstream. This application can fully consider a change of a channel, and ensure that reference pictures used by an encoder side and a decoder side are consistent, to improve encoding efficiency and avoid artifacts.

The application discloses a video processing method, device and equipment and a computer readable storage medium, and applies to the field of video processing, and comprises the following steps: receiving a code stream output by an encoder; if a saturation enhancement enabling flag in the code stream is valid, determining whether a pixel belongs to a skin color protection area according to a preset condition in a linear light domain; if the pixel belongs to the skin color protection area, performing smoothing interpolation processing on a saturation scaling factor in metadata parameters by using a sliding window according to a relative position of a hue angle of the pixel in a skin color interval to obtain an effective enhancement coefficient; if the pixel does not belong to the skin color protection area, taking the saturation scaling factor as the effective enhancement coefficient; and obtaining an enhanced color component according to a color difference component of the pixel and the effective enhancement coefficient and a luminance value of the pixel. The application solves the problems of insufficient saturation, over-saturation of the skin color area, skin color distortion and other video quality problems of an HDR video after dynamic range conversion.

The application discloses a video stabilization method and system based on video inter-frame homography transformation, belongs to the technical field of video supervision, and effectively distinguifies foreground and background in a video through a foreground mask module; after the foreground part is covered by using the mask, effective shallow semantic features of a background area are acquired through a credible shallow feature extraction module; for the effective shallow features of the background area, a motion estimation module associated with local features and global features is used to estimate the homography relationship between continuous adjacent frames; through a video stabilization smoothing module, the homography between the transformation frame and the reference frame obtained by using the motion relationship between adjacent frames is used to realize the stabilization of the whole video by using STN transformation. The effective combination of the foreground mask and the credible shallow feature avoids the artifact or distortion phenomenon caused by large disparity, the local and global feature relationship is established based on the credible shallow feature, accurate camera motion is obtained, and finally, the video stabilization is realized.

The invention provides an image processing method and device and a computer readable storage medium, and a terminal device comprises the steps that a first video and a second video are acquired, and the second video is a video obtained after the first video is enhanced; determining a first overlapping region with the maximum structural similarity in two adjacent frames in the first video, wherein the two adjacent frames comprise a first current frame and a previous frame thereof; calculating a first pixel difference between the pixel variation of the first video and the second video in the current frame and the pixel variation of the first video and the second video in the previous frame; and calculating a correction degree according to the first pixel difference, and correcting each pixel point of a second current frame in the second video in the first overlapping region according to the correction degree and the first pixel difference. According to the technical scheme, the flicker correction effect can be improved, and the video quality is improved.

The embodiment of the invention provides an infrared image processing method and device, equipment and a medium, and relates to the technical field of image processing, and the method comprises the steps: obtaining an initial infrared image of a first resolution and a visible light image of a second resolution, the first resolution being lower than the second resolution; a first weight of the initial infrared image and a second weight of the visible light image are determined by using the similarity between the initial infrared image and the visible light image, the similarity is in positive correlation with the first weight, and the sum value of the first weight and the second weight is a preset value; and fusing the initial infrared image and the visible light image by using the first weight and the second weight to obtain a first target infrared image with a second resolution. By applying the technical scheme provided by the embodiment of the invention, the resolution of the infrared image can be improved.

The utility model relates to medical anti -counterfeit image neck cover technical field, and disclose a kind of adjustable pressure type inflation anti-counterfeit image neck cover device, including pneumatic support assembly, the quantity of pneumatic support assembly is two groups, pneumatic support assembly includes telescopic air bag and two groups of plastic supporting block;Inflation and deflation component, the inflation and deflation component includes air pipe and micro dual-purpose air pump;Neck wear, the both ends of neck wear are respectively connected with corresponding plastic supporting block fixedly;Guiding component, the guiding component is fixedly installed between two groups of plastic supporting block, and guiding component includes support sleeve and cylindrical rod, and compression spring is arranged in support sleeve cavity. By setting telescopic air bag between two groups of plastic supporting block, by micro dual-purpose air pump cooperation air pipe to telescopic air bag inflation and deflation, telescopic air bag inflation when bulging will two groups of plastic supporting block be mutually far away, to support between the head of patient and shoulder, effectively reduce the neck shaking amplitude caused by breathing, to avoid the appearance of CT artifact.

The invention discloses a composite material ultrasonic defect burial depth detection and imaging method, which belongs to the technical field of nondestructive testing, and comprises the following steps: collecting an A scanning signal of a composite material test piece through a water immersion type focusing probe; performing continuous wavelet transform and dynamic threshold noise reduction on the signal; constructing a particle swarm optimization wavelet transform modulus maximum (PSO-WTMM) model, and searching an optimal parameter by using a fitness function of envelope entropy fusion peak value penalty; extracting the echo time difference of the upper and lower surfaces based on the optimal parameters, and calculating the defect burial depth in combination with the sound velocity; and constructing a time difference and amplitude ratio matrix through point-by-point scanning, and outputting a C-scan image after interpolation enhancement. By adopting the composite material ultrasonic defect burial depth detection and imaging method, the defects that parameters depend on experience and the robustness is poor in a traditional method are overcome, the defect depth positioning precision and the imaging quality are improved, and the method is suitable for nondestructive detection and health monitoring of aviation composite material structures.

In order to ensure a sufficient amount of light while avoiding the generation of distortion and artifacts caused by a rolling shutter, and unevenness in brightness and horizontal stripes are not obvious even if a change in a pulsed light-emitting organ relates to the rolling shutter, the present disclosure proposes a light source device that generates illumination light irradiated to a subject, and that provides a light source device that is capable of generating illumination light that is irradiated to the subject. The semiconductor light-emitting device includes: a semiconductor light-emitting element that emits light; and a light source control unit that generates a light-emitting profile of the semiconductor light-emitting element and drives the semiconductor light-emitting element on the basis of the light-emitting profile. The light source control unit performs: a process of dividing one frame period in an image signal acquired by the imaging element into N (N = 2m) integers; m is an integer of 2 or more); generating a light emission profile in which the light emission intensity is controlled in units of 2k (k = an integer from 0 to m) dimming sections in accordance with the target light amount during one frame; and performing dimming control of the next frame on the basis of the light emission profile (referring to Figure 7B).

The present invention discloses a network structure and method for multi-scale super-resolution reconstruction of a single image. The network structure of the present invention comprises: a feature extraction layer, the input of which is the original image, and the output of which is connected to the input layer of a dilated convolutional neural network; a cascade expansion layer, the input of which is connected to the output layer of a dilated convolutional neural network; an aggregation layer, which is used to fuse the low-level coding information of different scales of the feature extraction layer and the decoding information of the cascade expansion layer; a reconstruction module, the input of which is the output of the cascade expansion layer, and the output of which is the reconstructed image. The present invention adopts a back-end upsampling network to perform image mapping transformations in a low-resolution space, thereby reducing the computational complexity and spatial complexity.

The application discloses a kind of based on degradation perception's fundus image quality enhancement method, system, equipment and medium, wherein method includes: to original low-quality fundus image is standardized pretreatment;To the fundus image after pretreatment is divided into area, feature extraction and degradation degree evaluation, and introduce the self-attention mechanism with degradation perception penalty term, realize differentiation feature enhancement;Build generation network, including encoder and decoder, finally output enhanced fundus image;Build discriminant network, including multiple parallel multi-scale convolution branches, the output of multiple branches is fused, and through double output head respectively outputs true probability value and authenticity confidence heat map;The generation network and discriminant network are trained against;For the fundus image to be enhanced, after image pretreatment and local feature self-adapting enhancement, input trained generation network, output quality enhanced fundus image.The application can realize high-precision, high-fidelity enhancement to low-quality fundus image.

The invention provides an underground cable three-dimensional imaging method based on X-ray data, and the method comprises the steps: obtaining the X-ray data used for detecting an underground cable, and determining the compressed projection data and an initial attenuation coefficient field of the underground cable according to the X-ray data; based on the compressed projection data, the initial attenuation coefficient field and a preset objective function, the deformation field and the attenuation coefficient field of the underground cable are iteratively updated until the attenuation coefficient field of the underground cable converges, a target attenuation coefficient field is obtained, the objective function is used for realizing sparse reconstruction through cable geometric priori constraint, and in each iteration, the deformation field of the underground cable and the attenuation coefficient field of the underground cable are subjected to sparse reconstruction. Performing parameter optimization by using the updated attenuation coefficient field in the current iteration round; and based on the target attenuation coefficient field and the updated deformation field in each iteration round, performing three-dimensional model fusion to obtain a three-dimensional image of the underground cable. Therefore, efficient, accurate and nondestructive detection of the underground cable can be realized.

The invention relates to a compressed domain weak light image enhancement method and system based on 2D Gaussian splashing, and belongs to the technical field of image processing and computer vision. According to the method, an input weak light image is converted into discrete geometric primitive representation by using a 2D Gaussian splash compression reconstruction strategy, and explicit parameterization of the image is realized by optimizing position, covariance, color and opacity parameters; the method comprises the following steps: constructing a semantic-guided hybrid expert enhancement network model, extracting multi-scale semantic features by using a lightweight encoder, generating a spatial hybrid weight, and guiding a group of learnable color transformation operators to perform residual color modulation on Gaussian primitives; and then constructing a multi-target unsupervised loss function, obtaining an enhanced Gaussian parameter through iterative optimization, and rendering a normal illumination image. According to the method, the strategy of'compression, namely enhancement 'is put forward for the first time, calculation redundancy and secondary distortion caused by pixel domain processing are effectively avoided, and high-quality weak light image enhancement is achieved while the high compression rate is kept.