20 results about "Adaptive histogram equalization" patented technology

This invention discloses a small target tracking method based on image enhancement and attention mechanisms, comprising: performing adaptive histogram equalization on shallow feature maps to enhance local contrast, and fusing with deep semantic features to construct a multi-scale feature pyramid; generating attention weight maps in the horizontal and vertical directions through a coordinate-decoupled attention module to highlight the salient regions of small targets; introducing a dynamic anchor box generation algorithm based on Gaussian distribution, dynamically sampling and generating anchor boxes based on a Gaussian distribution model constructed from the target position in the previous frame; and using the deep cross-correlation layer of a Siamese network to calculate the similarity response map between the template and the search region to determine the center coordinates and bounding box of the small target. This invention effectively solves the problems of insufficient feature extraction, low spatial positioning accuracy, and inflexible anchor box generation in existing methods, significantly improving the accuracy and robustness of small target tracking.

The application discloses a consumer complaint response method and system based on AI assistance and language understanding, which splits the complaint response process into two core sub-problems of multi-modal consumer complaint data processing and feature fusion and demand attribution and response generation. In multi-modal consumer complaint data processing and feature fusion, first, regular expressions are used to denoise text, spectral subtraction is used to denoise voice, and Gaussian filtering and adaptive histogram equalization are used to denoise images; then, modal features are extracted, text is used as the core of cross-modal fusion, and entity and relationship are extracted to construct a multi-modal semantic knowledge graph. In demand attribution and response generation, Graph Transformer is used in combination with the graph and domain prior knowledge to output primary and secondary demands; a static complaint graph is constructed, an attribution path is mined through BFS and is verified through multi-modal verification; and an "emotion-demand-attribution-prevention" structure is used to optimize text and adjust the format by using LLM, and an individualized complaint response is output.

A tail water pipe air supply system sealing state recognition method based on computer vision, the method comprises collecting unit large shaft images under multiple working conditions, and constructing an image data set; then image preprocessing is completed through adaptive histogram equalization, median filtering, bilateral filtering and RGB- HSV space conversion; then label three types of areas of normal state, trace leakage and obvious water leakage, and divide training set, validation set and test set; then a YOLOv5 model embedded with an improved channel attention mechanism is constructed, the backbone network adopts CSPDarknet structure to extract multi-scale features, the neck network combines FPN and PAN to realize bidirectional fusion of features, and the detection head adopts a decoupling structure to improve positioning and classification accuracy; finally, the pre-training weight fine-tuning model is loaded, the training and verification are completed by using a multi-objective loss function, the real-time and accurate recognition of the sealing state of the tail water pipe air supply system is realized, the equipment fault risk of the unit is effectively reduced, and the method is suitable for the safety monitoring scene of the hydroelectric generating set.

The present application relates to the technical field of cell image analysis and medical artificial intelligence, and discloses a cell segmentation and adaptive cascade reasoning method and system based on prior frame guidance. The method comprises: acquiring a multi-modal cell image with a detection frame; performing adaptive histogram equalization preprocessing on the image and executing global preliminary screening segmentation; calculating a multi-dimensional weighted score of the mask and the detection frame, if the score is lower than a first matching threshold, extracting a local image block to dynamically adjust a flow field threshold and an estimated diameter for cascade retry; if the score after retrying is lower than a second threshold, triggering a label consistency protection mechanism to discard or back up poor samples; finally, performing connected domain purification on the retained mask, and calculating a topological solidity, and performing convex hull reconstruction on the mask with low solidity. Through two-stage reasoning, double-track quality filtering and topological constraint, the present application effectively improves the segmentation effect of dense and irregular cells and improves the generalization performance of downstream analysis.

This application relates to the field of computer vision technology, and in particular to a method for wildlife identification and tracing using infrared image enhancement and multi-scale fusion. The method includes: acquiring raw images from an infrared camera and enhancing the raw images using Retinex and contrast-limited adaptive histogram equalization; performing animal target detection, feature extraction, and feature fusion by optimizing the YOLOv8 model and using a multi-modal fusion Transformer to obtain multi-scale fused features; identifying species categories using a pre-defined primary species classification model and extracting and matching individual biometric features of key species using a pre-defined secondary individual identification model to generate individual identifiers for corresponding animals; and automatically associating the species identification results, individual identifiers, and multi-source forestry data to generate an intelligent decision-making report for wildlife population analysis and conservation management. This application contributes to achieving high-precision, automated wildlife identification and tracing.

The application discloses a kind of visual feedback drive's water tank cleaning robot intelligent control method and system, belong to robot control technical field.It includes: obtaining water tank inner wall original image, generates feature enhancement image by median filter and contrast limited adaptive histogram equalization;Input semantic segmentation model identifies stain area, calculates stain abundance index;According to stain abundance index, determine pollution grade, dynamically adjust cleaning motor speed and electric pressure push rod displacement, to change brush head positive pressure;In the process of advancing, collect cleaned area image and calculate residual stain abundance index, if greater than threshold value, then record non-compliance area three-dimensional coordinates, complete global scanning and plan compensation path to compensate cleaning.The application effectively improves cleaning efficiency and compliance rate, reduces energy consumption and mechanical loss.

The present application relates to a kind of metal computer support material surface defect identification method, belong to image processing technical field.The method is first in the segmentation of metal computer support gray scale chart, with the maximum gray consistency in segmentation area and the maximum gray difference between segmentation area as target to determine the best segmentation size, then according to the gray difference between the defect pixel point in the obtained segmentation block and neighborhood and the gray difference between normal pixel point and neighborhood Determine suspected defect pixel point, then according to the distribution characteristics of suspected defect pixel point and the similarity of the distribution characteristics of the real defect of metal computer support surface Determine the defect feature degree of segmentation block, so as to carry out the contrast of restriction according to the defect feature degree of each segmentation block Adaptive histogram equalization processing is targeted to highlight and enhance the small defects of metal computer support from the overall image.The present application realizes more accurate identification of the surface defects of metal computer support material by combining region segmentation with region targeted enhancement.

This application discloses a method and apparatus for panoramic infrared image stitching. The method includes: acquiring multiple initial infrared images with overlapping regions to construct an infrared image input set; performing Gaussian filtering and size normalization on each initial infrared image to obtain a normalized infrared image; inputting the normalized infrared image into a feature extraction model trained for infrared scene adaptation to extract key points and descriptors to obtain a feature set; matching the feature sets of adjacent images and eliminating mismatched pairs based on homography matrix constraints using a random sampling consensus algorithm to obtain a precise matching set; calculating the optimal homography matrix based on the precise matching set and performing perspective transformation to obtain a spatially aligned infrared image; adjusting brightness consistency using adaptive histogram equalization; calculating a fusion weight map based on the distance from pixels to image boundaries and performing weighted fusion; cropping invalid black border areas to obtain the final infrared panoramic image.

The application relates to the technical field of medical image processing, and provides a target lesion identification method and system based on medical images, which solves the problems of limited and unstable lesion identification precision. The method comprises the following steps: acquiring a cone beam CT image of a mandible before oral implantation and reference gray data; performing metal artifact reduction and adaptive histogram equalization algorithm enhancement processing with limited contrast to obtain an enhanced image; extracting a three-dimensional gray histogram of an implant target area to calculate gray skewness and gray kurtosis, and extracting a texture feature set; performing principal component analysis on the texture feature set to obtain principal component features, and splicing the principal component features with the gray skewness and the gray kurtosis into a fusion feature vector; analyzing the vector by using a density peak value clustering algorithm, and mapping the result to an anatomical partition, and performing Z-score testing in combination with the reference gray data to identify a bone lesion area. The application can realize accurate and automatic positioning of a lesion.

The present disclosure relates to the technical field of panel defect detection, and provides a method and system for implementing panel defect detection on the basis of image grayscale equalization. The method comprises: collecting statistics about grayscale values of a panel image by means of image grayscale value statistics collection, so as to obtain a grayscale histogram of the panel image; on the basis of the grayscale histogram of the panel image, using an adaptive histogram equalization algorithm to perform grayscale value equalization processing on the panel image, so as to obtain an equalized image; performing feature labeling on the equalized image, and performing neural network training on the basis of the feature-labeled equalized image, so as to obtain a target detection model; and performing panel defect detection on the equalized image on the basis of the trained target detection model, so as to obtain a defect detection result. By performing grayscale value statistics collection and grayscale value equalization processing on the panel image, defect imaging features can be enhanced, so that the target detection model can accurately perform defect detection and locating, thereby solving the problem that existing Mura defect detection is prone to missed detection.

This invention belongs to the field of engineering geological mapping technology. To address the current problems of low efficiency, large errors, and insufficient automation in processing engineering information, it provides a deep learning-based method for digitizing railway engineering drawings and extracting quantities. This method optimizes input by establishing a railway professional icon library and employing various data augmentation techniques such as adaptive histogram equalization. It utilizes an improved YOLOv8 model (incorporating SCConv, EMA, and RFA mechanisms) to achieve high-precision detection of graphic targets, and combines this with a PaddleOCR model to extract text information. Finally, the recognition results are digitized and automatically integrated with the engineering bill of quantities for quantity calculation, outputting standardized results conforming to an enterprise-level structured coding system. This method significantly improves the accuracy and efficiency of drawing recognition, automates and standardizes quantity calculation, and provides reliable technical support for the digital management of railway engineering projects.

The application discloses a kind of motion estimation methods for fusing deep learning feature optical flow and binocular vision, comprising: based on controllable adaptive histogram equalization preprocessing to driving image dataset;Construct the optical flow feature extraction model based on deep learning, and the moving target is identified training;Distance measurement is carried out by binocular camera, and the target position is obtained;Vehicle motion speed is acquired.Compared with traditional optical flow speed measurement, this method is based on deep learning optical flow and binocular imaging principle, and the motion parameter estimation of carrier displacement and speed can be realized according to video data, which solves the problem that traditional optical flow estimation method is too sensitive and cannot be stably estimated when driving at night or in weak light environment, and further improves reliability.Meanwhile, this method avoids the cumulative error of traditional inertia sensor and the shortcomings of poor anti-interference ability and low update frequency of GPS positioning speed measurement.

The application discloses a kind of turbulent flame geometry three-dimensional tomography method and system based on multi-probe CGI, comprising: N single-pixel detector is arranged around flame to form multi-probe detection array, dynamic space illumination modulator is used to generate structured light field to illuminate flame, each detector synchronously acquires barrel detection signal;Through compressed sensing reconstruction algorithm combined with convex optimization solving method, flame two-dimensional projection image under each view angle is recovered;Using adaptive histogram equalization, combined with filtering method, the calculation ghost imaging noise is inhibited, and two-dimensional projection image preprocessing is realized;Based on multi-view geometric projection relationship, construct tomographic projection matrix, use algebraic reconstruction technique combined with full variation regularization iterative algorithm to solve three-dimensional voxel distribution, and the flame front geometry three-dimensional structure is reconstructed.The application uses single-pixel detector to replace traditional area array camera, with the advantages of low cost, strong ability to resist harsh environment, high time resolution, suitable for high space-time resolution three-dimensional measurement.

Provided are an electronic device for improving local contrast of an image, and an operation method therefor. The electronic device may: partition the entire area of an original image into a plurality of first tiles; acquire a first image by applying a contrast limited adaptive histogram equalization (CLAHE) algorithm to the plurality of first tiles; partition the entire area of the original image into a plurality of second tiles by using straight lines connecting center points of sides of each of the plurality of first tiles; interpolate pixel values of pixels included in second tiles commonly overlapping four adjacent first tiles by using pixel conversion values by means of a histogram cumulative distribution function of each of the four adjacent first tiles; acquire a second image by converting pixel values of pixels located outside the second tiles among pixels in any one of the four adjacent first tiles by using pixel conversion values by means of the histogram cumulative distribution function of each of the four adjacent first tiles; and acquire a final image by combining the first image and the second image.

This invention provides a method for multi-scale feature extraction and automatic identification of defects in weld X-ray inspection images, belonging to the field of non-destructive testing and image processing technology. The method includes: acquiring the original X-ray image of the weld and welding process parameters; constructing a process condition encoding vector and calculating an adaptive filtering kernel size for non-local mean filtering and noise reduction; calculating a histogram clipping and limiting coefficient based on material thickness parameters for block-based contrast-limited adaptive histogram equalization enhancement; extracting multi-scale feature maps through a backbone network and spatial pyramid pooling; fusing the process condition embedding vector with the multi-scale feature maps; employing a channel-space dual attention mechanism to enhance the feature response of the defect region; and outputting the defect location, type, and size through a detection head network, with feedback adjustment of enhancement parameters when the confidence level is below a threshold. This invention solves the problems of low sensitivity in detecting small-sized defects and poor generalization ability under different process conditions.

The invention relates to the technical field of water ecological environment monitoring, and particularly discloses a blue-green algae water area intelligent identification method and system based on unmanned aerial vehicle vision, and the method comprises the steps: obtaining an original monitoring image; converting the image into a YCbCr color space, sequentially carrying out wavelet transform noise reduction, contrast-limited adaptive histogram equalization, bilateral filtering and median filtering processing on a brightness component, and reconstructing an enhanced image with a chrominance component; annotating the enhanced image and introducing brightness disturbance of a weather feature vector to carry out data augmentation; calculating a weighted fusion amount of local gray range and neighborhood gray co-occurrence matrix entropy, defining the weighted fusion amount as a blue-green algae texture condensation index, constructing an iterative morphological driving rule taking the index as an independent variable, and alternately executing structural element expansion and edge contraction until an overlapping rate converges to obtain a stable space mapping rule; segmenting the blue-green algae area by using the rule and converting the actual coverage area; the method accurately fits the amorphous boundary of the blue-green algae, adapts to complex illumination and weather changes, and improves the area measurement and calculation precision.

The application provides an improved CLAHE short-wave infrared image real-time enhancement method based on FPGA. In view of the problems of low contrast, unobvious detail information and easy amplification of noise of InGaAs short-wave infrared imaging under low illumination, an improved limited contrast adaptive histogram equalization image enhancement algorithm is provided, and the optimal clipping threshold is automatically determined through an adaptive search mechanism, so that the local contrast is enhanced and the noise amplification is inhibited. The application can effectively improve the contrast and detail distinguishability of the short-wave infrared image, reduce the influence of noise, and is suitable for fields of short-wave infrared imaging systems, target identification systems and weak light vision systems.

This invention discloses a deep learning-based method and apparatus for segmenting the microstructures of high-temperature alloys, comprising: collecting images of the microstructures of high-temperature alloy materials from scanning electron microscopes; labeling and annotating the microstructures in the images; constructing a dataset through image preprocessing and image enhancement; constructing an improved Mask R-CNN model, employing adaptive histogram equalization to expand the dynamic range of image gray levels and enhance local contrast; fusing image edge information into the network model using the Sobel operator and introducing the Huber loss function to improve model performance; training the model through transfer learning; and using the trained instance segmentation model to identify and segment the SEM images to be analyzed, obtaining the segmentation mask and labeling rectangle for each microstructure in the image. This invention achieves accurate identification of microstructure objects in high-temperature alloy materials based on the improved Mask R-CNN model.

The present application relates to the fields of medical image analysis, artificial intelligence, machine vision, etc., and provides a multi-disease classification method for color fundus images, aiming to solve the problems of low classification accuracy and weak generalization ability caused by class imbalance, coexistence of multiple lesion characteristics and image scale diversity. The method includes data preprocessing (sample balancing, black border removal, contrast limit adaptive histogram equalization enhancement, uniform scaling), data enhancement to generate A / B branch channels; three branch design: enhanced branch A / B parallel input Resnet101 main network, deep interaction is realized through the multi-head bidirectional cross attention mechanism of the feature interaction fusion module at the predetermined level, the interaction features are used as the input of the next level, and the fusion branch splices the interaction features for dimension reduction output. The three branches fuse the features through the multi-scale fusion module, each set a classifier to output logits, and the whole network parameters are collaboratively trained with classification loss, consistency loss and contrast learning loss. The present application effectively improves the classification accuracy and generalization ability.