53 results about "Ideal image" patented technology

To suppress an increase in processing time due to a load of inference processing while improving reading accuracy by the inference processing of machine learning. An optical information reading device includes a processor including: an inference processing part that inputs a code image to a neural network and executes inference processing of generating an ideal image corresponding to the code image; and a decoding processing part that executes first decoding processing of decoding the code image and second decoding processing of decoding the ideal image generated by the inference processing part. The processor executes the inference processing and the first decoding processing in parallel, and executes the second decoding processing after completion of the inference processing.

The invention discloses a self-supervision blind image decomposition method for an automatic driving scene, relates to the technical field of automatic driving perception, and solves the problems that in the process that high-quality data is used for supervising and training normal noise, the data is difficult to collect in a real scene, and an efficient noise reduction method easy to train is lacked. A self-supervised blind image decomposition method for an automatic driving scene comprises the following steps: step 1, adding and superposing degraded noise for the automatic driving scene: setting three typical non-ideal images for the automatic driving scene, and carrying out self-supervised blind image decomposition on image blurring caused by sensor noise, rainfall and snowfall; according to the design, three different noises are adopted to simulate a scene, and the noises are added into an image. According to the method, a blind image decomposition mode is adopted, various types of degradation are uniformly removed at a time so as to improve the noise reduction efficiency, and a double-flow decoding mechanism is introduced so as to improve the noise reduction accuracy.

The invention discloses a low-illumination visible light and infrared image fusion method with color recovery capability, and belongs to the technical field of computer vision. The method comprises the following steps: designing a low-light enhancement loss function based on comparative learning, so that the color distribution of an enhanced visible light image is aligned with the color distribution of an ideal image; designing a low-light enhancement neural network based on a Mangbar structure, converting the low-illumination image into a YCbCr color space, decoupling an illumination component and a color component of the low-light image, and respectively implementing enhancement processing; a fusion network based on a Mangban structure is designed, modal complementary information is obtained through a mixed attention mechanism, the enhanced visible light image and the infrared image are fused, color information of the fused image can be effectively recovered, and the visual quality of the fused image is improved.

The invention discloses a refraction distortion automatic correction method and system for an SPIV experiment, and belongs to the technical field of fluid mechanics experiment measurement, and the method comprises the following steps: obtaining a camera shooting posture and medium physical parameters; constructing a multi-layer medium forward refraction model; iteratively adjusting the assumed real angle and repeatedly calling the forward model by taking the measured distortion angle as a target through reverse iterative calculation until the predicted angle of the model is matched with the measured angle, and solving the corrected actual camera angle; constructing a nonlinear mapping relation from a distorted image coordinate to a non-refraction ideal image coordinate according to the difference between an actual angle and a measurement angle; performing reverse geometric transformation and interpolation on the original SPIV image by using the mapping relation to generate a correction image for eliminating the refraction effect; according to the invention, a software algorithm is used for replacing a traditional water prism physical correction device, so that the cost and complexity are reduced, and the adaptability to different experimental conditions and the correction precision are improved.

An information processing device (110) comprises: an optical calculation unit (112) that calculates each of the optical path of light from an imaging field of view in an imaging device that performs imaging using an optical element and the optical path of light from outside the imaging field of view that is outside the imaging field of view, calculates stray light produced by the optical element and a point spread function (PSF) of the optical element, and generates a degraded image in which the stray light and the PSF are applied to an ideal image that is handled as correct answer data in teacher data; a noise image generation unit (113) that generates a noise image including noise predicted to occur in an imaging element of the imaging device during imaging; an image calculation unit (114) that superimposes the noise image on the degraded image to generate a learning image; and a teacher data generation unit (115) that generates teacher data in which the learning image is employed as training data and an image corresponding to the imaging field of view of the imaging device within the ideal image is employed as correct answer data.

The invention discloses an image distortion correction method and device, equipment, a storage medium and a product. The image distortion correction method comprises the following steps: projecting an ideal image to obtain a distorted image, and obtaining a distorted point coordinate in the distorted image and an ideal point coordinate corresponding to the distorted point coordinate in the ideal image; calculating a distortion coefficient according to the distortion point coordinate and the ideal point coordinate; pre-processing the ideal image by using the distortion coefficient to obtain a pre-distorted image; and projecting the pre-distortion image to obtain a corrected image. According to the method, the image correction cost is reduced, and the problem that the imaging system is too heavy is avoided.

The invention discloses an image splicing method and device, mobile equipment, a storage medium and a product. The method comprises the following steps: acquiring pose information of a preset number of fisheye cameras in a world coordinate system and fisheye images shot by the preset number of fisheye cameras; converting a first pixel point on the target image into a world coordinate system according to preset target image view information and target image size information, and obtaining a first space point of the first pixel point in the world coordinate system; converting the first spatial point into a camera coordinate system according to the pose information to obtain a second spatial point; converting the second spatial point into the fisheye image by using the internal reference of the ideal camera and the corresponding relationship between the ideal image and the fisheye image to obtain a second pixel point; and according to a mapping relation between the first pixel points and the second pixel points, splicing the fisheye images to obtain a target image. According to the image splicing method provided by the invention, the calculation amount is saved, the splicing process is simplified, and the splicing effect is improved.

The present invention relates to an image enhancement method for constructing a spatial visual neural network model. The present invention first normalizes a binary noisy two-dimensional image to satisfy the small parameter restriction imposed by the adiabatic approximation condition for activating FHN neurons; secondly, convolves the spatial signal with an FHN neuron template to obtain a two-dimensional image encoding signal of a spatial neuron pulse potential sequence; constructs an anisotropic diffusion function based on improved bidirectional adaptive selection as a visual receptive field model with a directional selection mechanism, convolves the neuron pulse encoding signal with the constructed receptive field model; and finally finds the optimal binary image result. The invention effectively combines the directional selectivity of the biological visual system and the neuron pulse emission mechanism, making it conform to the physiological characteristics of real visual neuron pulse triggering and signal selectivity, thereby achieving a relatively ideal image denoising effect.

The application discloses a kind of three-dimensional face reconstruction and emotion analysis method, device and equipment based on text guide, comprising the following steps: obtaining the facial image information of human body and pre-processing and marking, construct training set;Three-dimensional face reconstruction model is trained based on text guide using training set, the face three-dimensional grid is obtained by inputing the face image to be analyzed into the three-dimensional shape model trained, and the face three-dimensional grid is input into emotion analysis model to judge face emotion category.The attribute matrix generated by natural language is jointly coded with image features and input into the three-dimensional reconstruction framework, which can significantly improve the geometric reconstruction accuracy in complex scenarios such as blurred face, local occlusion or dramatic / fine changes in facial expression, thereby effectively alleviating the problem of reconstruction failure or feature loss under non-ideal image conditions in traditional methods.

The invention discloses a phase difference method wavefront detection method combined with image sensor noise suppression, and the method comprises the steps: building an ideal light spot image of an image surface and a defocus position based on a phase difference method phase recovery system diffraction model; calculating a noise suppression factor of each defocus position according to the average noisy image of each defocus position and the image sensor parameters, and applying the noise suppression factor to the corresponding average noisy image to obtain a corrected noisy image of each defocus position; and finally, substituting the ideal light spot image at each out-of-focus position and the corrected noise-containing image into a target function, optimizing through a nonlinear algorithm, repeatedly updating the Zernike coefficient and recalculating the ideal image, and finally obtaining a restored wavefront phase. According to the invention, the wavefront detection precision can be significantly improved.

An object is to obtain training data through an ideal image generated from an actually-shot image.A data generating method of generating data used in learning a model using a processor, the data generating method including: generating an ideal image corresponding to output data of the model from a captured image obtained by a predetermined device; and generating a degraded image corresponding to input data of the model from the ideal image.

The utility model relates to the technical field of optical detection, in particular to an adjusting device and optical detection equipment, in the adjusting device, a bearing part fixedly provided with an optical element can be installed in an installation hole of an installation base; the flexible arm in the mounting base has a first state for locking the bearing part and a second state for enabling the bearing part to rotate relative to the mounting base, the mounting base is provided with the avoiding hole communicated with the mounting hole, and the positioning structure capable of being exposed out of the avoiding hole is arranged on the peripheral surface of the bearing part, so that when the flexible arm is in the second state, the bearing part can be positioned in the avoiding hole. In other words, before the bearing part is locked on the mounting base, an auxiliary tool penetrates through the avoiding hole and is matched with the positioning structure on the bearing part so as to drive the bearing part to rotate in the circumferential direction of the mounting hole, and the angle deviation between the optical axis of the optical element and the optical axis of the system can be reduced by changing the mounting angle of the bearing part in the circumferential direction of the mounting hole. Therefore, the aberration between the acquired image and the ideal image is reduced, and the imaging effect is improved.

An object of the system according to the embodiment is to allow a user to specifically imagine an ideal user himself / herself and to match the user with another person on the basis of the image.SOLUTION: A system includes an image generation part, a matching part, an appointment setting part, and a consulting part. The image generation unit generates an ideal image of the user based on the photograph of the user. The matching unit performs matching between the ideal own images. The promise setting unit sets a promise to actually meet when the matching is established. The consulting part provides consulting for approaching an ideal self for an appointment.SELECTED DRAWING: Figure 1

The present application provides an adaptive neural network non-uniformity correction method, a terminal device, and a storage medium. The adaptive neural network non-uniformity correction method includes the following steps: collecting a plurality of images in real time; selecting valid images from the plurality of images; the valid images include a plurality of pixels distributed in a matrix; obtaining the pixels with neighborhoods in the valid images, that is, valid pixels; the neighborhood is a set of pixels distributed along the circumferential direction of the valid pixel and at a 45-degree angle to the valid pixel; obtaining the input irradiance X of the valid pixel i,j ; obtaining the output irradiance Y of the valid pixel through a formula i,j , and the parameters in the formula include the variance of the pixel neighborhood, the gain correction coefficient, and the bias correction coefficient. This method can obtain an ideal image without ghosting and without aliasing effects.

The application discloses an unsupervised pedestrian re-identification method based on camera perception graph learning. Firstly, pedestrian image features are extracted, pseudo labels are obtained by using a clustering algorithm, and an in-memory dictionary is constructed. Each camera is regarded as an independent image domain, and samples are extracted from each image domain to form small batch training samples. Then, all image sample features are used to construct a camera perception graph, image samples are fused into a unified feature space by using a graph convolution network, and ideal image features are generated by feature aggregation. A non-parametric contrast loss based on the in-memory is constructed to train the feature extractor and the graph convolution network, and a camera adversarial loss is constructed to generate more robust and effective image features. The foregoing steps are repeatedly executed until the model converges. Finally, the pedestrian image is mapped into a feature space by using the feature extractor, a feature vector of a to-be-queried pedestrian image is compared with feature vectors of candidate pedestrian images in terms of Euclidean distance, and the pedestrian re-identification result is obtained.

The invention belongs to the technical field of state monitoring, and particularly relates to a mining belt conveyor start-stop state monitoring method, which comprises the following steps: carrying out linear fitting on representative gray values of sub-regions at the same position in no-load images with the same serial number acquired by two adjacent cameras mounted on a mining belt conveyor, and obtaining a linear fitting result; obtaining an ambient light influence relation and ambient light influence degree of two adjacent cameras in each sub-region, processing a monitoring image collected by the previous camera according to the ambient light influence relation to obtain an ideal image of the next camera, and weighting a difference result of the ideal image and the monitoring image of the next camera according to the ambient light influence degree to obtain a difference result of the monitoring image and the ideal image of the next camera. And the coal flow landslide probability of the monitoring image of the next camera is obtained and is used for judging whether the camera has a fault hidden danger in a corresponding monitoring area on the mining belt conveyor or not, so that the start-stop state of the mining belt conveyor is controlled. According to the invention, the start-stop control accuracy of the mining belt conveyor is improved.

The invention relates to the technical field of digital image processing, and discloses a digital image detector and method, and the detector comprises an image collection module which is used for obtaining original digital image data; the calculation processing module is used for processing the image, and the processing comprises the steps of correcting macroscopic geometric deformation of the image; analyzing the image by using a pre-trained generative model to identify microscopic content defects, and generating an ideal image and a structure residual image; the ideal image is used as semantic guidance, and geometric regularity and material uniformity of the image are evaluated; finally, various analysis results are fused, a comprehensive quality score is calculated, and a comprehensive defect positioning map is generated. And the output feedback module is used for outputting the comprehensive quality score and the comprehensive defect positioning map. According to the method, a physical and structural analysis two-way collaborative semantic guidance detection mechanism is constructed, so that accurate, comprehensive and objective automatic comprehensive evaluation on the digital file image quality is realized.

The present invention aims to provide a camera for a projection device and a projection device, which are applied to the field of projection technology. The camera includes a lens, a lens mount assembled with the camera lens, a circuit board, and an image sensor connected to the circuit board. With a horizontal plane constructed through the center line of the camera lens, the relative positions of the image sensor and the camera lens are set such that one of the upper and lower regions on the side of the lens imaging plane in the horizontal plane is the image acquisition region, and the other is the non-image acquisition region. By changing the relative positions of the image sensor and the camera lens, the image acquisition region of the camera is changed, so that the effective pixels of the projected image captured by the camera are more, facilitating the subsequent processing of the image. Its structure is simple, and it can achieve the most ideal image acquisition effect with the smallest modification method, facilitating production and use, and having strong practicability.

The present application provides a calibration method and a mask detection method for an extreme ultraviolet mask detection system. The method includes: after the extreme ultraviolet mask detection system performs mask detection on each group of target patterns, determining the mask detection image of each group of target patterns, the target pattern is a standard pattern selected based on spatial frequency, each group of standard patterns includes at least two elbow-shaped patterns, and the directions of the elbow-shaped patterns are different; obtaining an ideal image of each group of target patterns; determining the point spread function from the mask detection image to the ideal image of each group of target patterns; and determining the target point spread function of the extreme ultraviolet mask detection system based on each point spread function, the target point spread function being used to adjust the mask detection image of the sample to be detected. The present application solves the optimal point spread function of the extreme ultraviolet mask detection system by including a target pattern with at least two elbow-shaped patterns and the corresponding mask detection image and ideal image, thereby improving the resolution of the detection system and achieving high-resolution mask detection.

The invention relates to the technical field of local dimming, and discloses a local dimming method, which comprises the following steps: image acquisition and data processing: a PC host controls a PG tester to output a standard test picture to a display panel, photographs and captures an actual display image of the display panel, and a system compares an ideal image with the actual picture to generate an original Mura data matrix; compression compensation data generation: fitting a Gamma curve of the display panel, based on the fitted Gamma curve and the original Mura data matrix, calculating to generate a precise compensation coefficient per pixel or partition, and compressing the compensation data; and burning the data and verifying the effect. Compared with a traditional module, a layer of black and white panel and a control panel are additionally arranged on hardware, a local dimming program is integrated on the control panel, and local dimming of more dimming partitions can be achieved by applying a smaller number of LEDs.

The present invention provides an infrared remote sensing image restoration method, device, electronic device, and storage medium, relating to the field of image processing technology. The method comprises: obtaining a target observation image; determining a constraint term for removing stripe noise based on the differential curvature value of each pixel in an ideal image, wherein the differential curvature value of the stripe noise contaminated area in the ideal image is greater than the differential curvature value of the non-strip noise contaminated area; constructing a denoising energy functional model based on the constraint term; inputting the target observation image into the denoising energy functional model for iterative processing, and determining the resulting ideal image as the restored image with stripe noise removed, wherein: during each iteration, the ideal image output from the previous iteration is input into the denoising energy functional model until the iteration constraint condition is satisfied. The present invention can effectively remove stripe noise caused by solar pollution, and during the stripe noise removal process, the original image radiation information remains unchanged.

The present invention discloses a projection lithography moiré alignment device, wherein the device includes: an illumination system, a mask mark, a lithography projection objective lens, an alignment mark, a DMD spatial modulator, a CCD camera, and a sensor motion stage; the illumination system emits light, which passes through the mask mark and then forms an aerial image through the lithography projection objective lens; the position of the aerial image is found by controlling the movement of the sensor motion stage, so that the light passing through the alignment mark is diffracted to form moiré fringes; the moiré fringes reach the DMD spatial modulator and are encoded; the encoded moiré fringes reach the CCD camera, and the moiré fringes are reconstructed based on the collected light intensity information; the position deviation between the ideal image and the actual image is obtained based on the reconstructed moiré fringes, and the position deviation is fed back to the sensor motion stage for correction to achieve alignment. The present invention does not require a spatial filter, simplifies the optical system, reduces hardware costs, and can be widely applied in the field of optical precision detection technology.

The present invention relates to the field of robotic arm control technology, specifically providing a robotic arm visual servo tracking method based on adaptive neural network compensation. Based on the image feature error between real-time image features and ideal image features, adaptive neural network compensation is performed on the traditional robotic arm control rate to obtain a compensated robotic arm controller. The controller is used to obtain the control joint velocity of the robotic arm, and the control joint velocity is used to control the robotic arm to track a dynamic target. The method of the present invention can compensate for system tracking errors using an adaptive neural network without increasing the control rate of the robotic arm servo system, thereby greatly improving the robotic arm's tracking accuracy, reducing human intervention, and simplifying the complexity of robotic arm control.

The invention discloses a three-dimensional face reconstruction and emotion analysis method, device and equipment based on text guidance, and the method comprises the steps: obtaining the face image information of a human body, carrying out the preprocessing and labeling, and constructing a training set; and training the three-dimensional face reconstruction model by adopting the training set based on text guidance, inputting a to-be-analyzed face image into the trained three-dimensional shape model to obtain a face three-dimensional grid, and inputting the face three-dimensional grid into the emotion analysis model to judge the face emotion category. According to the method, the attribute matrix generated by the natural language and the image features are jointly coded and input to the three-dimensional reconstruction framework, so that the geometric reconstruction precision in complex scenes such as face blurring, local shielding or violent / subtle facial expression change can be remarkably improved; therefore, the problem of reconstruction failure or feature loss of a traditional method under a non-ideal image condition is effectively relieved.

The present application relates to the technical field of rectal cancer CTV segmentation, in particular to a rectal cancer CTV segmentation method based on CViL-UNet, comprising the following steps: preprocessing rectal cancer CT image data, converting a 3D CT sequence into a multi-channel 2D input format, removing image noise and intensity non-uniformity, and standardizing the pixel value range to adapt to the CViL-UNet model input. The present application realizes accurate identification of the key areas of CTV primary tumor, subclinical spread area and micro-infiltration focus through local-global feature cooperation and high-frequency detail enhancement, solves the pain points of missing micro-infiltration and misjudging healthy tissue in traditional methods, can process conventional CT images without intestinal preparation, breaks the dependence of traditional methods on ideal image conditions, and the automatic and real-time segmentation process in clinical value eliminates the subjectivity and inefficiency caused by manual intervention, while meeting the dual requirements of millimeter-level precision and rapid deployment for radiotherapy planning.

A three-dimensional (3D) scanning system includes a projector that generates an emitted light projected on an object, a reflected light being reflected from the object; a sensor that generates image data according to the reflected light; and a depth processor that generates depth data according to the image data and at least one modified factor representing corresponding deviation amount between the image data and an ideal image data due to a circle of confusion caused by the reflected light passing a lens of the sensor and then irradiating a sense plane of the sensor out of focus.

An information processing device (110) includes an optical computation unit (112) that computes a PSF of an optical element in each of a plurality of regions obtained by dividing an image-capturing surface of an image-capturing element that performs image capturing using the optical element, and generates an aberration image that has influence of aberration by the optical element on an ideal image by applying a corresponding PSF to each of a plurality of region images obtained by dividing the ideal image in accordance with the plurality of regions, a noise image generation unit (113) that generates a noise image representing noise at the time of image capturing, an image computation unit (114) that generates a learning image by superimposing the noise image on the aberration image, and a labeled training data generation unit (115) that generates labeled training data with the learning image being used as training data and the ideal image being used as correct answer data.

The invention relates to the technical field of optical data processing, in particular to an optical image simulation and data set automatic generation method, which comprises the following steps of: firstly, shooting an original image; taking a zero-aberration point spread function as an Airy spot convolution kernel, and performing convolution operation on the shot original image by using the Airy spot convolution kernel to obtain an ideal image reflecting a diffraction limit condition; taking the point spread function of the specified aberration as an aberration convolution kernel, and performing convolution operation on the shot original image by using the aberration convolution kernel to obtain a simulated aberration image; and synchronously cutting the obtained ideal image and the simulation aberration image to obtain a simulation data set. According to the method, a zero-aberration point spread function is used as an Airy spot convolution kernel, a point spread function of a specified aberration is used as an aberration convolution kernel, parameterized convolution is carried out on an original image, and an ideal-aberration strict paired image is efficiently obtained; through consistent cutting and unified file storage mapping, standardized organization and traceable management are realized.

The application discloses a kind of for the correction method and device of cone beam CT system geometry parameter error, method includes: acquisition projection image projected on flat panel detector;Obtain the geometry parameter error of cone beam CT system;By the translation error of transverse and longitudinal of the plane where flat panel detector is, correct projection image, obtain first correction image;The normal rotation error of the plane where flat panel detector is and the distance error between flat panel detector and ray source are corrected to first correction image, obtain second correction image;For second correction image, respectively correct along the rotation error of transverse and longitudinal of the plane where flat panel detector is, obtain third correction image;The distance error of third correction image's rotating table and ray source is corrected, obtain the ideal image of geometry parameter error correction.By geometry parameter error correction projection image, without adjusting the position of cone beam CT system device, it is easy to calculate, effectively eliminate the influence of system error on image three-dimensional reconstruction.

The application relates to a method and device for eliminating crosstalk between image sensor pixels. The method comprises: establishing a crosstalk convolution mathematical model between an actual image function and an ideal image function of the image sensor pixels, and obtaining regular filter coefficients based on a convolution kernel in advance to obtain crosstalk coefficients and store the same; obtaining an actual image function collected by the image sensor; and performing convolution calculation on the crosstalk convolution mathematical model according to the actual image function and the crosstalk coefficients to obtain an ideal image function in which crosstalk between pixels is eliminated and output the same. The application converts the crosstalk problem between pixels into a convolution problem by establishing the crosstalk convolution mathematical model. For the image sensor under the same process condition, the same convolution kernel can be used to eliminate the crosstalk between pixels and restore the ideal image by the regular filter without changing the pixel structure.