100 results about "Image differencing" patented technology

The invention provides a door and window glass defect feedback control method and device, electronic equipment and a storage medium, and relates to the technical field of door and window glass defect detection. The method comprises the steps of collecting a first image after a polarized light imaging detection device is adjusted to suppress reflected light on the surface of glass and enhance scattered light of bubbles in the glass; after the polarized light imaging detection device is adjusted to enhance reflected light on the glass surface, a second image is collected; calculating an adaptive weight group by analyzing local characteristics of the second image; weighting the first image according to the self-adaptive weight group, and subtracting the weighted first image from the second image to obtain a difference image; and extracting bubble defect information in the glass by analyzing the differential image. According to the method, interference caused by non-uniformity of glass surface reflection is overcome, the adaptability of image differential processing to surface reflection background suppression is improved, and therefore image features of internal bubble defects are extracted more effectively and more accurately.

The invention relates to the technical field of image data processing, and discloses an element mounting quality detection method for a chip mounter, which comprises the following steps of: synchronously acquiring images before and after element mounting and a vacuum pressure signal of a mounting suction nozzle, compensating vibration by utilizing an inherent substrate background area in the images, obtaining an optical position deviation through image difference and mass center calculation, and calculating the surface mounting quality of the element according to the optical position deviation. And finally, logically associating the deviation with the change characteristic of the pressure signal, and classifying the physical process of the mounting event. According to the invention, through the collaborative diagnosis of optical detection and physical signals, the physical reason of the mounting event can be clearly distinguished; the inherent limitation caused by vibration interference and unclear fault attribution during high-speed operation of a traditional detection mode is avoided, and the method has good adaptability to incoming material elements with different appearances.

The invention discloses an astronomical positioning method for a space target elongated astronomical image of a large-field-of-view space-based monitoring platform, relates to the technical field of astronomical observation, and solves the problem that the elongated astronomical image is abandoned due to incapability of target association and precise orbit determination caused by inaccurate centering of the elongated astronomical image and low precision of a positioning method of an existing monitoring platform. According to the method, an observation image is divided into 16 sub-view fields, and each sub-view field is an independent view field and is calculated by an independent processor; removing fixed stars in the observation image by adopting an adjacent frame image difference method and extracting a space target star image; secondly, describing the length and angle of the lengthened astronomical image by adopting a minimum enclosing rectangle, and constructing a degradation model; performing motion blurred image restoration on the elongated star image; error points with large errors are eliminated by adopting a mutual comparison method, and track association of the space target is completed; and finally completing precise astronomical positioning of the elongated astronomical image and the like. According to the method, the processing efficiency is greatly improved, and a large number of lengthened astronomical images which are abandoned before are fully utilized again.

The invention relates to a projector automatic obstacle avoidance method and device combining adaptive picture cutting and deep learning, and the method comprises the steps: enabling a projector to project a white background picture, capturing a projection picture through a camera, calculating a difference image based on a projection region background image and a projection picture image, carrying out the adaptive cutting of the difference image, and obtaining a projection region image, the difference image can highlight the difference between the projection area and the background area, the projection area image can be found out more conveniently, then a deep learning model is used for detecting an obstacle in the cut image and recording position information, the maximum projection area is calculated, the position information of an obstacle-free area is obtained and then converted to a projector coordinate system, and the position information of the obstacle-free area is obtained. And finally projecting a final picture. Therefore, the automatic obstacle avoidance of the projector is realized, the clear and complete projection picture is ensured, the problem of poor projection effect in a complex environment due to traditional manual processing and the prior art can be effectively solved, and the projection quality of the projector in a complex scene is improved.

The invention relates to the technical field of defect detection, and particularly discloses a high-precision defect detection system based on image difference and threshold processing, which comprises a registration image acquisition module, a fixed mask conversion module, an image difference detection module, a correction difference graph generation module and a defect set output module, the method comprises the following steps: firstly, completing affine registration by shape features, changing a fixed mask into a dynamic mask, and performing edge robustness in the mask to obtain a preprocessed image; local movement difference is carried out on the preprocessed image and the multiple reference images, fusion of a reference domain and a time domain is carried out to generate a difference response image, and candidate areas are extracted in combination with a dynamic mask; estimating and correcting a displacement field in the candidate range, synchronously mapping to obtain a correction difference graph and a correction mask graph, and determining a peripheral rejection area; according to the method, false alarm and missing detection are effectively reduced, and the method is suitable for online high-precision detection.

The present invention discloses a security monitoring video optimized transmission method and a video monitoring terminal, which relates to the field of video monitoring technology, and includes the following steps: extracting the current frame and the previous frame from a continuous monitoring video frame sequence, and then extracting two adjacent frames of images from the video stream to provide basic data for subsequent dynamic change analysis; pre-processing the extracted two adjacent frames of images, using image difference technology to calculate the difference between the two adjacent frames of images, and generating a differential image through pixel-level comparison. The present invention extracts features such as motion intensity and change density by pre-processing and differential calculation of two adjacent frames of images, and inputs a deep learning model to analyze dynamic changes, and divides the scene into fast and normal dynamic changes. An initial compression algorithm is used for normal scenes, and the bit rate is dynamically adjusted for fast dynamic scenes to ensure the capture of details and balance video quality and compression efficiency, thereby improving the performance and practicality of the monitoring system.

The invention relates to the technical field of new energy power station monitoring, in particular to a new energy power station anomaly detection method and system based on acoustic imaging, and the method comprises the steps: obtaining the acoustic sequence information of a target new energy power station photovoltaic panel, and arranging a plurality of monitoring units based on the acoustic sequence information; transmitting sound waves with a preset frequency to the photovoltaic panel through a sound wave transmitter, collecting reflected sound wave data through an acoustic imaging sensor, and generating an acoustic image sequence; identifying and identifying anomalous reflection regions of the plurality of acoustic images based on the acoustic image sequence; carrying out image difference processing on the acoustic image sequence, and carrying out region detection on an acoustic difference image to obtain an abnormal region of the photovoltaic panel; and classifying the abnormal areas of the photovoltaic panel, and performing cross verification on the preliminary abnormal detection result by using abnormal reflection distribution to generate abnormal distribution of the photovoltaic panel. By emitting specific sound waves to the photovoltaic panel and analyzing the reflection characteristics of the sound waves, the structural change of the internal material of the photovoltaic panel can be sensed.

The invention relates to the technical field of mold detection, and discloses a special working state inner wall detection device for a neutral borosilicate molded bottle mold, which comprises a workbench, an opening and closing mechanism is arranged on the middle side of the top of the workbench, a clamping mechanism is arranged on the top of the workbench, and a six-degree-of-freedom mechanical arm is mounted on the rear side of the top of the workbench. A ceramic probe is arranged on the outer side of the six-degree-of-freedom mechanical arm, and a base fixing seat is fixedly connected to the outer side of the ceramic probe. A real-time image differential analysis module is arranged, a blowing and powder spraying mechanism is used for carrying out pneumatic excitation on a suspected defect area, and a multispectral endoscope module is controlled to respectively collect optical images before and after excitation; by calculating the normalized difference measure of the two images and comparing the normalized difference measure with the preset physical displacement threshold value, the non-solid-phase attachments displaced due to pneumatic excitation and the solid-phase defects with fixed positions can be distinguished, and the technical problem that two abnormal areas with different properties are difficult to effectively classify in the prior art is solved.

The invention relates to the technical field of image processing, in particular to a weld joint performance visual detection method and system based on image processing, and the method comprises the steps: for any weld joint, obtaining a gray level image of any weld joint in each sampling period, and obtaining a gray level image sequence; carrying out image difference processing on the grayscale image sequence to obtain a difference image sequence, and screening at least two target images with potential damage evolution regions from the difference image sequence; for any target image, acquiring a comprehensive damage coefficient of any target image according to the roughness degree and the texture feature of each potential damage evolution region in any target image and the confusion degree of pixel points in each potential damage evolution region; and the function relationship between the sampling period and the remaining service life of any welding seam is constructed by using the data change trend of the comprehensive damage coefficient of each target image, and the remaining service life of any welding seam is predicted, so that the fatigue damage state of the welding seam is predicted in advance.

A method, system, device and medium for detecting a dark space target and predicting a track under a dense fixed star background comprises the following steps: firstly, obtaining a reconstructed multi-pulse function by using a frequency spectrum phase difference, further estimating motion parameters of a dark star background and a space target, and then designing an image difference fusion coarse processing module to obtain a candidate target image sequence; the number of tracks is effectively reduced; and finally, according to the motion parameters and the candidate target image sequence, directional track association is carried out, an association area is reduced to a fan-shaped range, and redundant false alarms are removed while the target track is accurately and rapidly detected. The system, the equipment and the medium are used for realizing the invention. Based on the distribution characteristic of the space target and the time domain motion characteristic, the method not only achieves the precise suppression of fixed stars with different brightness, but also improves the detection robustness of targets with different star magnitudes, and has a better detection effect on dark space targets with different motion parameters under the background of dense fixed stars.

The invention discloses a routing inspection image feature detection and disease recognition method, which comprises the following steps: firstly, matching a routing inspection image acquired by an unmanned aerial vehicle with a template image in a pre-constructed image sample library by adopting a template matching algorithm, and determining an original template image corresponding to the routing inspection image; secondly, performing feature extraction on the inspection image and the matched template image by adopting an image feature extraction algorithm to obtain image features of the inspection image and the template image; and then, according to the extracted image features, adopting an image difference algorithm to obtain difference features between the inspection image and the template image, and obtaining a change feature image of the inspection image, thereby realizing intelligent identification and marking of a disease area in the inspection image, and finally, inputting the change feature image into a trained disease classification model, thereby realizing disease classification. And classifying and identifying diseases in the inspection image through the disease classification model.

The invention discloses an electric power material warehouse-in and warehouse-out management method and system based on image recognition, a medium and a processor, and relates to the technical field of warehouse management, and the method comprises the steps: obtaining the basic information of electric power materials, and dividing the electric power materials into large-class materials and subdivided materials according to the basic information; acquiring image data of electric power materials to obtain an electric power material image; processing the electric power material image through an image difference method, extracting the contour area of each electric power material, and calculating an area threshold according to the contour area; extracting text information in the to-be-detected electric power material image through an optical character recognition model, and matching the text information with a database to obtain a recognition result of a large class of materials; and performing category identification on the to-be-detected electric power material image, and outputting an electric power material identification result after contour area threshold judgment. Different warehouse-in and warehouse-out management processes can be adopted for different types of electric power materials, the management efficiency is improved, and the warehouse-in and warehouse-out management data accuracy is enhanced.

The invention provides a wafer defect detection method and system based on image segmentation and image difference, and relates to the technical field of defect detection, and the method comprises the following steps: firstly, obtaining a difference image based on a detection image and a template image; then carrying out gray value extraction on the detection image and the template image based on the difference image, and calculating the penetration rate of the defect area according to a gray value extraction result; performing line segmentation processing on the detection image, and performing intersection positioning on a line segmentation image based on the difference image to obtain an intersection area image; performing geometric operation on the difference image and the intersection area image to obtain a defect width and a line width, and calculating a coverage ratio of a defect area based on the defect width and the line width; and finally, comprehensively judging the severity of the defect according to the penetration rate and the coverage rate of the defect area. The problems that the influence degree of the defect cannot be measured in the existing wafer appearance defect detection and over-detection is easy to occur are solved.

The application discloses a method for intelligent identification of changes in river and lake water areas by a UAV and early warning of abnormal problems in rivers and lakes, comprising the following steps: constructing a digital twin bottom plate data of rivers and lakes; manually editing a basic flight route of the rivers and lakes, intelligently generating a shooting flight route based on the basic flight route and the digital twin bottom plate data of the rivers and lakes, splicing the basic flight route and the shooting flight route, and obtaining a dynamic flight route of the UAV; performing dynamic flight route flight according to the dynamic flight route, collecting data information during flight, and returning the data information to a UAV base station through a network and transmitting the data information to an edge computing device; generating river and lake orthographic images and river and lake spatial terrain data by combining the data information collected by the UAV with water area management lines; and identifying and comparing the orthographic images before and after the two times by using a water area surface identification model and an image difference method, so as to obtain a water area surface change area and abnormal problems in the rivers and lakes. The method solves the problems of low identification accuracy, low image spatial resolution and unintelligent flight route scheduling in the existing UAV patrol of the rivers and lakes.

The invention provides an inspection method and an inspection device, which can restrain workload and cost for judging whether an inspection object is normal or abnormal. The inspection apparatus includes: an image generator storage unit in which an image generator constructed by machine learning is stored, the image generator generating an image complemented with a defect region from an image having a defect region in a part; an image storage unit that stores an image to be inspected; a defect image generation unit that generates, from the image to be examined, a defect image in which a region to which a window of a pre-specified shape and size is applied is damaged; a complemented image generation unit that generates a complemented image in which the region is complemented by inputting the defect image to the image generator; a difference acquisition unit that obtains a difference between the image to be inspected and the complemented image; and a determination unit that determines whether the region is normal or abnormal by comparing the difference with a predetermined reference.

The invention relates to the field of image processing, in particular to a textile fabric surface stain detection method and system based on image processing, and the method comprises the steps: carrying out the image difference according to the possibility of a pixel point in a fabric wrinkle shadow region, the illumination intensity of each pixel point and the gray value of the pixel point in a fabric image transmission process, calculating a difference value of each pixel point on the fabric images of the adjacent frames and a difference value mean value; calculating the number of the pixel points of which the difference values are greater than the mean value of the difference values, and taking the ratio of the number of the pixel points to the total number of the pixel points on the fabric image as a difference result; and judging that stains exist on the surface of the textile fabric when the difference result is greater than a preset threshold value. The method and the device have the effect of improving textile stain identification accuracy.

The invention relates to the technical field of industrial robots, and discloses a robot flexible plug-in teaching method and system based on double-vision collaboration.The method comprises the steps that S1, a double-vision system and a robot coordinate system are calibrated, and a feeder area is defined; s2, in a teaching mode, a human hand track is tracked through a global positioning camera, and a'material taking 'action at a feeder is automatically identified by using a sliding window clustering algorithm; s3, triggering a high-precision Mark point camera to shoot a local image of the PCB after the human hand leaves the plug-in, and accurately calculating a plug-in position coordinate of the component through an image difference technology; and S4, automatically synthesizing a robot executable program according to the identified action sequence and the calculated accurate coordinates. According to the method, traditional manual programming is replaced with visual teaching, the operation technical threshold is remarkably reduced, the industrial-grade operation precision is guaranteed through a double-layer cooperation mechanism of hand guiding and visual fine calibration, the multi-variety production line changing time is greatly shortened, and flexible automatic plug-in is achieved.

The invention relates to the technical field of ocean engineering, in particular to a remote-sensing-image-based automatic monitoring and early warning method for deformation of an offshore guide dike, which comprises the following steps: acquiring remote sensing images and standard images of the offshore guide dike, and forming a remote sensing image set by the remote sensing images according to acquisition time and position; dividing the remote sensing image set into a plurality of image pairs, and performing image registration according to the positions of the plurality of image pairs and the standard image on the sea guide dike to obtain a target remote sensing image; carrying out image difference on the target remote sensing images at different time points to obtain deformation features, and determining deformation regions related to distribution of the deformation features; combining the deformation areas according to the distribution of the deformation characteristics, determining deformation trend directions corresponding to the deformation areas, and performing influence analysis on deformation quantities and deformation rates in the deformation trend directions to obtain influence paths corresponding to the deformation areas; based on the obtained influence path, determining an alarm type of the corresponding offshore guide embankment; and the accuracy and efficiency of offshore guide dike deformation identification are improved.

The present invention relates to a method and device for automatic obstacle avoidance for a projector that combines adaptive cropping of images with deep learning. First, the projector projects a white background image, then uses a camera to capture the projected image. A differential image is calculated based on the background image of the projection area and the image of the projection screen. The differential image is adaptively cropped to obtain the projection area image. The differential image can highlight the difference between the projection area and the background area, making it easier to find the projection area image. A deep learning model is then used to detect obstacles in the cropped image and record their position information. The maximum projected area is calculated, and the position information of the obstacle-free area is obtained. This information is then converted to the projector coordinate system, and the final image is projected. This method achieves automatic obstacle avoidance for the projector, ensuring a clear and complete projection image. It can effectively solve the problem of poor projection effects in complex environments caused by traditional manual processing and existing technologies, and improve the projection quality of the projector in complex scenes.

The present invention discloses a method for detecting sea surface targets based on wake wave characteristics and a computer-readable medium. The present invention uses image differential processing to obtain differential detection images from multiple consecutive frames of images; performs target identification on each differential detection image frame and obtains a target determination result; repeatedly determines whether a target exists on all differential detection images, and if a target exists, performs target background cancellation to obtain a target background cancellation image; performs primary target identification on the target background cancellation image, and based on this, performs clustering, optimization, and pairing to obtain a secondary target identification image; estimates the target's heading based on the secondary target identification image; and obtains target heading information based on multiple frames of the secondary target identification image. The present invention can achieve effective detection of sea surface targets with a low false alarm rate, and is based on image domain processing, with the advantages of small computational data volume and high operational efficiency.

The invention relates to the technical field of visual SLAM under autonomous navigation, and discloses a dynamic scene RGB-D SLAM method based on image difference and particle filtering, and the method comprises the steps: firstly introducing a PROSAC algorithm for matching optimization based on visual SLAM matching feature points of FAST features and BRIEF descriptors; secondly, calculating a basic matrix based on the optimized matching point set, re-projecting the previous frame of RGB image to the current image, and roughly detecting object motion by differentiating the two images; thirdly, based on particle filtering tracking, object motion detection is enhanced, specifically, differential image information is used as measurement information of particle filtering, the weight of particles is updated, and therefore moving object position estimation is optimized. In each iteration, a MAP estimate of the moving object position is calculated. Furthermore, the detection result is further optimized through vector quantization of the depth map, so that the moving object is accurately segmented. And finally, by removing the features in the segmented regions, a basis is provided for realizing accurate self-motion estimation of the camera and effective mapping.

The present application relates to the technical field of visual detection device, especially to a mirror shift type double-sided visual detection device. Its technical scheme comprises a rack, a conveying component and a visual detection component. The conveying component adopts double-sided conveying belt support to make the middle part of the workpiece suspended, and the visual detection component realizes synchronous imaging of the top surface, bottom surface and side surface of the workpiece through upper and lower industrial cameras and mirror reflection. In view of the problem that the mirror surface is easy to be contaminated, leading to misjudgment, the device is provided with a movable protective film covering the mirror surface, and when the initial detection is unqualified, the protective film is driven to move and secondary shooting is performed. Through image difference intelligent discrimination of defects, real defects and stains on the film are effectively distinguished. The present application is also integrated with a spacing adjustment and automatic sorting mechanism, which can complete multi-angle and high-precision detection and sorting in a single process, and has the advantages of high detection efficiency, strong anti-interference ability, good adaptability and the like.

The invention discloses a printing defect detection method and detection device for a cylindrical bottle body, and relates to the technical field of image matching, and the method comprises the following steps: collecting a bottle body image in a rotating state; extracting positioning points by using a normalized cross correlation algorithm; abnormal points are screened out through linear regression, and image cutting is completed through affine transformation; selecting an optimal period image by adopting a perceptual hash algorithm; correcting the stretching distortion in the vertical direction through an improved thin-plate spline transformation model; and finally, carrying out image difference on a corrected image and a template image and identifying printing defects. The detection device using the detection method can automatically complete bottle body conveying, image acquisition and defect identification and rejection, and high-precision and automatic detection of printing defects of cylindrical bottle bodies of different specifications is realized.

The present invention discloses a printing defect detection method and system based on single-sample learning and skeleton reconstruction. A data augmentation scheme suitable for printing images is used to augment a single printing image, avoiding overfitting of the model to training samples. The image is reconstructed by the skeleton through the network, and the reconstruction task can effectively control the generalization ability of the model, enabling the model to accurately reconstruct the normal area while generating a large reconstruction error in the abnormal area. At the network design level, an improved denoising autoencoder structure based on dilated convolution and randomly copied and pasted noise is adopted to capture both global information and local details. An image comparison algorithm suitable for the printing field is used. Through image differencing and combined with opening and closing operations, the defect position is finally determined. Through experiments and engineering verification, the present invention can effectively detect various printing defects and has wide engineering application value.

The application is suitable for the field of machine vision and industrial detection technology, and provides a metal workpiece defect detection method and system based on image difference, which comprises the following steps: collecting a to-be-detected image and a normal reference image, and pre-processing the two images, then performing pixel-level absolute difference operation to generate a difference image; the to-be-detected image, the normal reference image and the difference image are spliced to form a multi-channel input tensor, which is then input into a lightweight convolutional neural network to output a defect probability heat map and determine a coarse positioning mask; the coarse positioning mask is subjected to connected domain analysis to filter noise and obtain a candidate defect region; the candidate defect region is subjected to fine segmentation to obtain an accurate defect mask; and the candidate defect region and the accurate defect mask are superimposed and displayed on the original to-be-detected image to obtain a detection result. The application combines the difference information of the normal reference image and the to-be-detected image, and introduces a fine segmentation network, thereby effectively improving the detection accuracy and robustness.

The invention relates to the technical field of image processing and defect detection, in particular to an image enhancement method for surface defects of a mining sputtering sensor, which comprises the following steps: an image mapping step: acquiring a single-exposure two-dimensional digital image; mapping a pixel gray value into a node temperature, and constructing a virtual enthalpy field matrix; a matrix evolution step: calculating a local structure tensor, extracting a texture main direction, and generating a heat conduction coefficient matrix; virtual time iterative evolution is started, and a local second derivative is calculated; a barrier construction step: constructing a heat insulation wall if a preset sudden change condition is met; otherwise, keeping unchanged; a decoding difference step: generating a steady-state thermal field and reversely decoding the steady-state thermal field into a substrate background image; performing differential operation to extract a heat dissipation residual image; a fusion output step: superposing the residual image to a non-highlight area to output an enhanced image; extracting or filtering a thermal gradient output defect topological coordinate system; according to the method, self-adaptive smoothing of highlight and forced sharpening of defect edges are realized.

The invention relates to the technical field of microlens detection, in particular to an automatic microlens detection method and system, and the method comprises the steps: obtaining a stripe image of a microlens under the illumination of a stripe structure; preprocessing the stripe image to obtain a preprocessed image with a block area as a foreground; determining a rotation angle according to the boundary of the reference block region, and rotating the corrected stripe image to align the stripe with the row direction; segmenting the rotating image, identifying a connected block, solving an external frame to generate a block mask and a complementary set interval mask thereof, and obtaining a block image and an interval image; performing segmentation on the block image, taking a median value according to rows to replace and generate a block reference image, performing differential binarization to obtain a block defect candidate, and performing independent binarization and merging on an abnormal region; overlapping the interval image and the block mask, performing closed operation to suppress line noise, and then performing binarization inversion to obtain an interval defect candidate; and superposing the two candidate output defect mark images. According to the invention, the defect can be automatically identified and marked and output by the microlens under the stripe background.

The invention provides an automatic checking method and system for an IDC (Internet Data Center) based on an image processing algorithm. The method comprises the following steps: step 1, deploying a plurality of inspection robots in an IDC data center, and enabling the inspection robots to autonomously move in a channel of the data center; the inspection robot is deployed to autonomously collect images, the tedious process of traditional manual checking is avoided, the checking efficiency is greatly improved, a large amount of time and labor cost are saved, meanwhile, a path planning algorithm, an automatic focusing algorithm and the like are utilized, it is ensured that the robot efficiently moves and collects clear images, and the overall working efficiency is further improved; feature extraction and asset information comparison are carried out based on an image processing algorithm, errors caused by human factors are reduced, the accuracy of asset identification is improved, meanwhile, the illegal moving-in and moving-out condition of equipment is detected by adopting an image difference algorithm, abnormity can be found timely and accurately, disorder of asset management is avoided, and intelligent management is realized.

The present application provides a seat positioning estimation device, a seat positioning estimation method, and a seat positioning estimation computer program. The seat positioning estimation device has: a masking image generation section (31) that generates a masking image by masking a region representing a driver in an image representing the driver in a vehicle cabin generated by an imaging section (2) during a period in which the driver is seated in the vehicle (10); a difference image generation section (32) that generates one or more difference images by differencing each of one or more reference images that differ from each other in positioning of a driver's seat (11) and the masking image; and an estimation section (33) that inputs each of the one or more difference images to an identifier, obtains an individual estimation value of the positioning of the driver's seat for each difference image, and calculates a statistical representative value of the individual estimation values obtained for each difference image as an estimation value of the positioning of the driver's seat.

The present invention discloses a method, system, and computer-readable storage medium for detecting and extracting linear defects in LCD screens. The method comprises: S1. acquiring an image to be inspected and performing ROI image cropping on the image to extract the actual area to be inspected in the image; S2. Gaussian filtering the obtained ROI image using filter kernels of different sizes to obtain a foreground image and a background image of the ROI image; S3. performing image difference on the foreground image and the background image to obtain a foreground-background separation image; S4. performing Canny edge detection on the foreground-background separation image to obtain a Canny detection image with the defects outlined; S5. performing a progressive probabilistic Hough line transform on the Canny detection image to obtain linear defect information; and S6. re-clustering the extracted defect information to obtain accurate defect information. This invention improves the accuracy and efficiency of linear defect detection in screens.