9 results about "Noise index" patented technology

The invention belongs to the technical field of image anomaly detection, and particularly relates to a combustion chamber inner wall carbon deposition detection method and system based on machine vision so as to solve the technical problem that an existing method is prone to interference and misjudgment due to neglect of texture directivity and lack of adaptivity. The detection method comprises the following steps: S1, calculating a kurtosis coefficient of a global brightness histogram of an RGB image to obtain a first chrominance component and a second chrominance component; s2, acquiring respective high-frequency sub-bands in horizontal, vertical and diagonal directions under at least two scales; calculating a directivity index representing the anisotropy degree of the chrominance noise; s3, determining a judgment threshold value of the directivity index; and S4, performing weighted summation on the chrominance noise energy obtained under each scale in combination with the center frequency corresponding to each scale to obtain a comprehensive noise index representing the carbon deposition degree. The comprehensive noise index obtained through multi-scale information fusion can accurately represent the carbon deposition degree.

1. Name of the product in this design: Graphical User Interface for Rural Environmental Monitoring in Electronic Equipment. 2. Intended use of this design: for use in an electronic device. 3. The key design features of this product are: the layout of the graphical user interface elements. 4. Images or photos that best illustrate the design key points: Interface state change diagrams. 5. Purpose of the graphical user interface: Enter the username and password in the main view to enter the interface change state diagram. The interface change state diagram is the main user interface for rural environmental monitoring and management, including the air quality index monitoring section to display the air quality index, the noise index monitoring section to display the specific noise value, the environmental status indicator light section for alarms and indicator light flashing when the environment is abnormal, the environmental parameter value section to display the real-time changes of specific environmental parameters by clicking, the water quality monitoring to display water quality statistics, and the real-time data display of the real-time changes and historical curves of different environmental indicators.

The application discloses a multi-scale semi-supervised target segmentation method based on noise index guidance, belongs to the field of computer vision, and aims at the problems that the robustness of an existing single traditional segmentation method is not strong, the feature capturing capacity is limited, a deep learning method U-Net pays insufficient attention to context information, and a full supervision method is complex in pixel-level labeling. The application relates to information fusion at two levels: since the feature capturing capacity of a single traditional segmentation method is limited, a plurality of traditional methods can be fused to obtain more accurate pseudo labels; since the deep learning network U-Net pays insufficient attention to context information, the down-sampling feature maps of different scales can be fused in the skip connection of each stage, so that the fused feature maps will carry multi-scale background information and save fine-grained target position information, and then the network can pay attention to more important regions.

This invention discloses a training and application method for a Cherenkov video denoising model for radiotherapy, relating to the field of video image processing technology. Addressing the technical challenge of extremely low signal-to-noise ratios due to extreme physical limitations in Cherenkov imaging and the difficulty in obtaining noise-free reference ground truth, this invention obtains consecutive noisy video frames and the number of pulse accumulations from the linear accelerator, converting this number into a noise index characterizing the effective noise level. Using this noise index as a conditional variable, adaptive instance normalization is applied to feature modulation of the video denoising model. A pseudo-noisy image is generated by combining a pre-trained noise prediction model, and the network is updated solely based on the cycle consistency loss calculated between the noisy frames and the pseudo-noisy image. In the application phase, the denoised image is time-series accumulated according to control nodes and compared with the reference dose distribution. This invention achieves high-fidelity removal of extremely low photon-physical mixed noise under ground truth conditions, significantly improving the accuracy of in vivo radiotherapy applications.

The present application belongs to the technical field of image anomaly detection, and particularly relates to a combustion chamber inner wall carbon deposition detection method and system based on machine vision, to solve the technical problem that the existing method is prone to misjudgment due to interference because it ignores the texture directionality and lacks adaptability. The detection method comprises: S1, calculating the kurtosis coefficient of the global brightness histogram of the RGB image to obtain a first chroma component and a second chroma component; S2, obtaining high-frequency subbands of horizontal, vertical and diagonal directions under at least two scales; calculating a directionality index representing the anisotropy degree of chroma noise; S3, determining a determination threshold of the directionality index; S4, weighting and summing the chroma noise energy obtained under each scale in combination with the corresponding center frequency of each scale to obtain a comprehensive noise index representing the carbon deposition degree. The comprehensive noise index obtained by multi-scale information fusion can accurately represent the carbon deposition degree.

The invention discloses a deep learning image denoising device and method based on wavelet guidance. A deep learning image denoising device based on wavelet guidance comprises an image sequence and noise index input module used for receiving noise indexes N and K continuous original image frames and preprocessing the original image frames; the wavelet guiding module comprises K-2 wavelet U-shaped networks (WBU-Net), and the wavelet guiding module is used for carrying out grouping processing on the K preprocessed original image frames to obtain K-2 pieces of spliced image data; and the U-shaped network module is used for carrying out fusion processing on the K-2 spliced image data to obtain a denoised image frame of the original image frame. The technical problem of exposure overlapping or tail end frame estimation deviation caused by sacrifice of time resolution due to traditional time domain median filtering and increase of on-chip accumulation times to improve the signal-to-noise ratio is solved, and the technical effect of quickly realizing high-quality image denoising under the condition of high time resolution is achieved.

The application discloses an image noise evaluation method and system based on image blocking, relates to the technical field of image processing, collects image data with noise, and is divided into multiple small blocks, separates the brightness channel and the chroma channel; based on the brightness channel, the brightness noise index is calculated, based on the chroma channel, the chroma noise index is calculated, and the block noise label is formed; a super-lightweight convolutional neural network combined with dimension permutation is constructed; the noise image is input into the trained super-lightweight convolutional neural network, and the brightness noise and chroma noise evaluation matrix of each block are output; the noise image is evaluated according to the brightness noise and chroma noise evaluation matrix, and the evaluation result is obtained. The application can carry out noise evaluation on the image in the region, provide accurate guidance for noise reduction, output the brightness and chroma noise evaluation results at the same time, and the fused results are stable and comprehensive and are not affected by local small changes. The application has low data requirement, strong universality, and only needs a lightweight convolutional neural network, and has high execution efficiency.

The invention relates to the technical field of image recognition, and discloses an intelligent crack detection and early warning system for bridge safety operation and maintenance, and the system comprises a crack signal-to-noise index field calculation module which obtains a crack signal-to-noise index field through calculation; the enhanced graph calculation module is used for calculating to obtain an enhanced graph; the segmentation mask calculation module is used for calculating and obtaining segmentation masks; the feature extraction module is used for extracting a skeleton and Euclidean distance transformation; the risk prediction module obtains a node risk and a global maximum risk through a risk prediction model; and the risk grade dividing module is used for dividing risk grades. According to the method, the crack visibility is quantified by constructing a crack signal-to-noise index field, the crack definition of a low-visibility area is improved based on self-adaptive enhancement of the index, and image distortion caused by global enhancement is avoided; the self-adaptive threshold segmentation reduces the boundary shrinkage of the segmentation mask, compensates the crack width by combining the width correction driven by the signal-to-noise index, captures the inter-segment space correlation through the risk prediction model, and guarantees the final risk and grade judgment to be accurate.

The invention belongs to the technical field of light amplification, and discloses a low-noise wide-dynamic-range adaptive light amplification optimization system, which is characterized in that a balance control module adopts a master-slave agent deep reinforcement learning architecture, and takes the lowest noise index, the adaptive dynamic range and the minimum response delay as core optimization targets; adjusting rear-end amplification gain, pumping power and mode equalization parameters in advance in combination with a wavelet transform signal pre-judgment mechanism; the dual-mode amplification module can automatically switch a single-mode working mode or a multi-mode working mode according to a mode crosstalk coefficient and is matched with a noise source and pumping parameter dynamic matching algorithm to accurately suppress interference aiming at different noise types; the expansion module adopts a pre-stage and post-stage grading gain design, the pre-stage is adaptive to a strong signal and a burst signal, and the post-stage is adaptive to a weak signal and a wide signal, so that the gain is continuously adjustable; according to the method, the noise index can be maintained at a low level, the dynamic range coverage range is obviously widened, the response delay is greatly shortened, and the method is successfully adapted to a burst wide-width signal scene.