246 results about "Hough transforms" patented technology

The invention relates to the technical field of goods shelf detection, in particular to a self-detection method and device for goods shelf settlement, and provides the following scheme: obtaining a top view image through an image sensor arranged right above the top of a goods shelf, dividing the image into a plurality of grid units, and positioning a rectangular geometric shape by utilizing Hough transform; and screening a plurality of to-be-detected areas in combination with the edge features. For an area to be measured, homographic registration and ortho-rectification are carried out based on a reference image, a displacement field is obtained by adopting sub-pixel-level dense registration, and a geometric parallax component field corresponding to imaging parameters is obtained through robust estimation. And under the hypothesis of small deformation, inverting the parallax into a pixel normal distance, and carrying out weighted aggregation on the local region to obtain a local distance measurement result. And by iteratively combining adjacent grids, determining a settlement area boundary, and finally outputting a settlement detection result. Millimeter-level settlement quantification can be realized under a single-frame image, hardware transformation is avoided, and the method is suitable for automatic detection and long-term monitoring of multi-specification goods shelves.

The invention discloses a photovoltaic cleaning control method fusing RTK positioning and visual compensation, and belongs to the technical field of photovoltaic module cleaning. According to the invention, through the RTK track points acquired offline, the least square method is combined with the RANSAC algorithm to carry out segmented straight line fitting, and the original points are projected to the corresponding straight lines to realize collinear correction, so that the measurement error is eliminated, and a continuous and collimated navigation path can be generated; edge lines (photovoltaic module grid lines) with geometrical characteristics in the environment are detected in real time through a visual sensor, the line segment direction is extracted through Hough transform, the line segment direction is compared with the target course calculated through RTK, course deviation is estimated, dynamic correction is conducted through a control strategy, and it can be ensured that the photovoltaic cleaning robot operates along the expected path with high precision; and on the basis of a lightweight YOLOv7 model, the photovoltaic panel is subjected to pollution category division and mapped to optimal water pressure and brush speed parameters, cleaning parameters are dynamically modified, and then cleaning strategy control is completed.

The invention discloses a PDF (Portable Document Format) drawing recognition and information structured extraction method. The method comprises the following steps: generating a high-resolution bitmap through image preprocessing; positioning and classifying a text region, a table region and a symbol region in the drawing based on a target detection model of transfer learning; hough transform is combined with SIFT feature matching to identify engineering symbols, and sub-pixel positioning is realized through an RANSAC algorithm; after the oblique text is corrected through affine transformation, the content is extracted through OCR; reconstructing a table structure based on OPTICS clustering and projection analysis; constructing an RDF knowledge graph according to a coordinate association rule; and using U-Net difference to detect and position an omission area and complementing the omission area. According to the method, deep learning and image processing technologies are fused, the problems of low rotating text recognition rate, table structure loss and semantic association deficiency in a traditional method are solved, through lightweight model compression and TensorRT acceleration, the analysis accuracy is remarkably superior to that of the traditional method, and the method can be widely applied to the fields of constructional engineering, petrochemical engineering and the like and has wide application prospects. And the drawing information processing efficiency and the data integrity are improved.

The invention provides an automobile part visual inspection correction method and equipment. The method comprises the following steps: extracting a part contour through edge detection, calculating an inclination angle by adopting Hough transform, executing affine transform on an original image, enabling a rotation angle to be negative to the inclination angle, and ensuring that the parts are horizontally aligned; in the corrected horizontal image, calculating a transverse offset through a feature matching algorithm or template matching, and translating an image area to enable the center of the detection window to be aligned with the actual position of the part; if inclination and offset are detected at the same time, inclination correction is executed preferentially, the offset is calculated based on the image after inclination correction, synchronous correction is achieved through a composite transformation algorithm, and accumulative errors are reduced through iterative optimization; and extracting a size parameter from the corrected image, comparing the size parameter with a predetermined tolerance range, and if the measured size is outside the predetermined tolerance range, marking the image as a defective product, and removing the defective product. The method is used for upgrading the visual inspection system for automobile manufacturing.

The invention relates to a sun meridian detection and course angle real-time calculation method based on polarized light field vector characteristics, and aims to solve the problems that a traditional sun meridian detection method is easily interfered in low signal-to-noise ratio scenes such as haze and cloud layer shielding, lens distortion causes polarization angle errors, the false detection rate is high, and the calculation accuracy is high. In order to solve the problems that the real-time performance and robustness of course calculation are affected due to the fact that single polarization angle information is difficult to fully reflect the sun direction, polarization parameter calculation is carried out by utilizing a multi-channel polarization image, and the sun meridian is detected by adopting Hough transform through fisheye lens correction, bilateral filtering and candidate region screening, so that the real-time performance and robustness of course calculation are improved. And a multi-point inversion method is combined to accurately obtain a sun vector and a zenith angle and an azimuth angle thereof, and finally a course angle is calculated in real time. The problems of imaging distortion and insufficient real-time processing in the traditional sun positioning technology can be effectively solved, and a powerful navigation positioning technical guarantee is provided for the fields of unmanned aerial vehicles, autonomous navigation, polarized light navigation and the like.

The invention provides an electric power line defect detection method and system based on visual identification, and relates to the technical field of line detection.The method comprises the steps that firstly, a visible light and infrared image dual-light registration and differential operation technology is adopted, and a fusion feature map capable of reflecting component thermal anomaly and material difference at the same time is generated; secondly, a black box type target detection model is abandoned in a part positioning link, but line segment screening and reconstruction are carried out by combining probability Hough transform with specific prior geometric knowledge of a power line, so that dependence on a large amount of labeled data is reduced, interpretability of a positioning process is enhanced, and the positioning accuracy is improved; according to the method, accurate areas of key components such as wire insulators can be extracted in a complex background, the concept of a probability saliency map is introduced in a defect identification core link, so that a defect area is enhanced and highlighted, and then a complete defect contour is determined by adopting an adaptive threshold segmentation and area growing algorithm; accurate mapping from pixel-level features to object-level defects is realized.

The invention discloses an anti-interference visual detection method and system for a welding groove edge position of a double-YOLO model, and the method comprises the steps: collecting n frames of welding region global images when an electric arc is located at different positions of a groove, building a YOLO-I model which outputs the welding region global images containing a plurality of welding feature position frames, the method comprises the following steps: acquiring m feature images which comprise different sizes, shapes and numbers of welding interferences and comprise groove edges, establishing a YOLO-II model which outputs the feature images of the groove edges which comprise interference area position information, generating groove left and right edge ROI images through the YOLO-I model, and identifying the groove left and right edge ROI images through the self-adaptive feature identification capability of the YOLO-II model. Welding features in a global image of a welding area or welding interference information such as welding spatter, welding fume and welding arc light in a groove edge ROI image are effectively recognized, and a groove edge line is reconstructed through a linear principal component analysis method based on Hough transform. And the influence of interference factors in the welding process on the detection precision of the edge position of the groove is obviously reduced.

The invention discloses a tool fracture and fatigue intelligent detection method based on vibration signal analysis, and the method comprises the following steps: S1, installing a vibration sensor, and collecting the vibration signal of a tool in real time; s2, the collected tool vibration signals are preprocessed, and noise in the signals is removed; s3, performing time-frequency analysis on the preprocessed vibration signals, and extracting time-frequency features in the signals; s4, performing deep feature learning on the extracted time-frequency features to form deep features; s5, the depth features are classified and analyzed, and the health state of the cutter is output; s6, according to the health state optimization feature extraction and prediction result of the cutter, generating learning output; s7, evaluating the health state of the cutter in real time according to the learning output, and pushing alarm information; and S8, according to the alarm information, predicting the service life of the cutter and optimizing a cutter replacement and maintenance strategy. According to the method, short-time Fourier transform and Hough transform are combined, and the extreme learning machine is applied, so that intelligent detection on the fracture and fatigue of the cutter is realized.

The invention relates to a circuit diagram-to-netlist generation method and system based on deep learning and a product. The circuit diagram-to-netlist generation method comprises the following steps: S1, identifying devices in a circuit diagram through a YOLOv8 model to obtain position information and category of each device; identifying a straight line segment in the circuit diagram through a probabilistic Hough transform algorithm to obtain a starting point coordinate, an end point coordinate and an orientation corresponding to the straight line segment; and S2, generating corresponding text description information according to the position information, the category, the starting point coordinate, the end point coordinate and the orientation. According to the method, by effectively integrating multi-modal information and combining image and text data, comprehensive understanding of a circuit diagram is provided, the recognition accuracy and robustness are improved, even if a single component has a recognition error, the influence of noise on the whole model can still be reduced through final information fusion, and the recognition accuracy and robustness are improved. Therefore, misjudgment and missed judgment in recognition are reduced, and the recognition accuracy of the element and the connection relation is remarkably improved.

The present invention relates to an intelligent tree barrier detection method and system based on multimodal perception. The method uses sensors carried by drones to collect data, including RGB images, LIDAR point clouds, multispectral data, etc. First, the data is preprocessed to form a standardized input. Then, a Hough transform is performed on the LIDAR point cloud to extract the position of the power line and generate a three-dimensional line model. Then, a deep learning algorithm is used to locate the tree bounding box from the RGB image, and the high-risk point cloud area is screened in combination with the line model. Then, a three-dimensional tree model is constructed, and the minimum safe clearance distance between the power line and the tree is calculated in real time through a collision detection algorithm to obtain a three-dimensional tree model with a safety distance annotation. Finally, a spatial risk assessment is performed based on this model. The present invention can significantly improve the reliability and efficiency of power line inspections.

The invention relates to the technical field of pupil boundary detection, in particular to a multi-algorithm fusion pupil boundary detection method and system, and the method comprises the following steps: obtaining an eye image through an infrared camera, converting the eye image into a grey-scale map, optimizing the grey-scale map, carrying out the median filtering noise reduction, maintaining the edge information, and determining the pupil boundary through a Sobel operator and Hough transform. And the pupil center displacement is analyzed, the motion consistency is evaluated, the pupil position is corrected, and the pupil positioning accuracy is optimized. According to the method, the image preprocessing process is optimized by applying infrared camera shooting and image processing algorithms, accurate detection, median filtering and local contrast enhancement of pupil boundaries are achieved through multi-algorithm fusion, and particularly noise and edge information in the images are optimized; according to the method, the edge detection precision and the image definition are effectively improved, the pupil boundary is accurately mapped through combined use of the Sobel operator and Hough transform, and the misjudgment rate is greatly reduced.

The invention discloses a camera parameter calibration method and device, a computer readable storage medium and a vehicle, and the method comprises the steps: collecting a checkerboard image shot by at least one camera, and extracting a line contour in the checkerboard image based on an edge detection algorithm; in the line contour, detecting a straight line segment based on Hough transformation or an LSD algorithm, and fitting the straight line segment based on a least square method; determining an internal reference matrix and a distortion coefficient of the camera based on the fitted straight line segment and a nonlinear optimization algorithm; and determining a rotation matrix and a translation vector of the camera based on the straight line segment, the internal reference matrix and the distortion coefficient. According to the method, the robustness of parameter calibration in a calibration field aging, fouling or reflective scene can be improved, the success rate, the automation degree and the efficiency of the calibration process are improved, and the driving safety of a user is improved.

The invention discloses an OCR-based medical document intelligent identification method and system, and relates to the technical field of document identification, and the method comprises the steps: collecting a medical document image through a mobile terminal, generating a binary image based on the collected image through U-Net in combination with multi-scale feature fusion and an attention mechanism, and cutting the binary image; based on the cut binary image, text information is extracted through an OCR model, and a structured field is extracted according to the typesetting rule and geometric distribution of the medical document; and performing deterministic rule judgment and risk assessment on the structured data. According to the method, the image calculation complexity is reduced through standard graying processing, the text region feature extraction precision is improved by fusing a U-Net structure of a CBAM attention mechanism, effective fusion and noise suppression of multi-scale features are realized in combination with Attention Gate, text direction correction is realized in combination with Hough transform, and the text detection robustness and recognition accuracy are improved.

The invention relates to the technical field of image data processing, in particular to an induction furnace coil temperature field visual monitoring method based on thermal imaging, and the method comprises the steps: obtaining the three-dimensional temperature field data of an induction furnace coil, and extracting high-temperature voxels; an adaptive weight is constructed for each high-temperature voxel, and the weight is composed of two parts: 1, the local linearity of the voxel is determined by calculating the projection density of the voxel in multiple directions; secondly, obtaining a colinearity score by calculating the consistency of the candidate voxel and other candidate voxels in the neighborhood of the candidate voxel in the local main direction; and performing weighted Hough transform on the candidate voxels to vote by taking the product of the two as an adaptive weight. Through the intelligent weighting mechanism, real linear fault signals can be effectively amplified, noise interference is suppressed, and the detection accuracy and robustness of early discontinuous linear high-temperature faults are improved.

The invention discloses a paper stack counting method and device, and relates to the technical field of electronic sensing, and the method comprises the steps: carrying out the optimization through a three-stage exception handling mechanism according to the initial paper number of first laser ranging and second laser ranging, and generating the optimization result of the paper number of the first laser ranging and second laser ranging; and performing absolute value calculation on the paper quantity optimization results of the first laser ranging and the second laser ranging to generate a paper difference value, and performing cross validation on the paper difference value and an error threshold to obtain the final paper quantity. Paper stack graph data are collected and preprocessed, a paper stack inclination angle is obtained by combining Hough transform, and the paper stack inclination angle is compared with a set paper stack inclination angle threshold interval, so that accurate identification of a measurement area is realized, the rationality of a subsequent laser scanning path and the accuracy of data collection are ensured, and the adaptability of the device to the posture change of the paper stack is improved.

A display video correction device includes: a feature extractor that extracts a feature from a camera image obtained from an in-vehicle camera; a Hough transformer that performs Hough transform on the feature extracted; a straight line detector that detects a plurality of straight lines in the camera image, based on a transform result of the Hough transform performed; a vanishing point calculator that calculates, based on the plurality of straight lines detected, first coordinates indicating coordinates of a vanishing point in the camera image; a difference calculator that calculates a difference between the first coordinates calculated and predetermined second coordinates; and a position corrector that corrects, based on the difference calculated, a position of a display area in the camera image to be displayed on a display device.

The invention provides an intelligent inspection system based on a robot and a camera, and relates to the field of intelligent manufacturing, the system adopts a four-layer architecture including a sensing layer, a transmission layer, a platform layer and an application layer, the sensing layer is composed of a fixed camera and an intelligent inspection robot; the transmission layer adopts optical fibers and 5G hybrid networking; the platform layer is integrated with a distributed server cluster, a data platform and an AI inference engine; the application layer provides a Web and mobile terminal management interface. The core of the system adopts a plurality of AI algorithms, including an instrument reading recognition algorithm based on Hough transform and CRNN, a leakage detection algorithm combined with background difference and an optical flow method, and a multi-device collaborative scheduling algorithm based on reinforcement learning, and constructs a digital twinborn inspection scene of virtual-real mapping. According to the invention, automatic and intelligent inspection and visual monitoring of power plant instruments, pipelines, power distribution rooms and other areas are realized, the inspection efficiency and accuracy are effectively improved, and the labor cost and the safety risk are reduced.

The invention discloses an automatic instrument reading method and system based on embedded machine vision. The method comprises the following steps: deploying an embedded image acquisition unit with wide dynamic imaging capability; performing multi-mode preprocessing of white balance correction, contrast enhancement and noise suppression on the image; self-adaptively identifying the instrument type and positioning a key area by using a lightweight convolutional neural network; hough transform and sub-pixel edge detection are adopted for a pointer type instrument to achieve high-precision angle analysis, and optical character recognition is completed for a digital instrument through a bidirectional long-short-term memory network; and finally, fusing multi-frame results through Kalman filtering, evaluating credibility, and outputting stable readings with timestamps. According to the technical scheme, high-robustness, high-precision and low-delay automatic reading can be realized at the embedded end, and the operation stability and the long-term adaptability of the system in a weak network environment are remarkably improved.

The invention discloses a multi-information fusion ghost point elimination and target tracking method. The target tracking precision is improved by eliminating false associated point locations by using the scattering characteristic of target angle cooperative positioning. Aiming at the problem that the tracking precision is reduced due to a large number of false associated ghost points, the invention provides a two-stage ghost point elimination and target tracking algorithm based on angle measurement and target motion characteristic fusion. According to the algorithm, a cooperative positioning strategy of first correlation and then estimation is adopted, and a vision field grid map and an energy accumulation matrix are constructed by establishing a mapping relation between angle measurement noise and positioning errors; the method comprises the following steps: analyzing spatial geometric distribution characteristics between a real target and false correlation ghost points in a vision field, designing a novel elimination criterion based on the idea of Hough transformation, and realizing primary coarse elimination of the ghost points; by researching the distribution characteristics and motion characteristics of a target positioning fuzzy region, a prediction tracking gate is constructed by using motion parameter identification, and secondary fine elimination of ghost points is realized from the kinematics level.

The invention belongs to the field of submarine pipeline risk early warning, and relates to a submarine oil and gas pipeline burying state identification method based on unsupervised machine learning and 3D Hough transformation. The method comprises three parts of pipeline data acquisition, pipeline point cloud data clustering, and pipeline state detection and identification, wherein the pipeline point cloud data clustering part comprises the following steps: firstly, removing most noise points in point cloud data by using cloth filtering and statistical filtering, and then clustering the point cloud data by using a DBSCAN algorithm of a self-adaptive threshold value; the higher the density is, the higher the class interestingness is. According to the pipeline state detection and identification part, 3D Hough transformation is used for carrying out cylinder and circular truncated cone detection on interest classes, and the exposed state of the pipeline is determined according to different detected results.

The invention relates to the technical field of image processing, in particular to an automobile part cutting and positioning method based on machine vision, and the method comprises the steps: obtaining an image of a to-be-positioned blank of an automobile part, extracting a sub-contour, for each pixel point of the sub-contour, calculating the local form continuity according to the pixel point and adjacent pixel points at the two sides, and obtaining the local form continuity of the to-be-positioned blank of the automobile part; determining local gradient divergence according to gradient amplitudes and gradient directions of all pixel points of the pixel point in a local range, fusing the local form continuity and the local gradient divergence, determining the reliability degree of the pixel point, executing generalized Hough transform based on the reliability degree of all pixel points of all sub-contours, and obtaining the reliability degree of all pixel points of all sub-contours. And performing registration based on the positioning center and a pre-obtained standard component, and guiding cutting equipment to execute cutting operation. According to the method, the cutting accuracy of the to-be-positioned blank of the automobile part is improved.

The application belongs to the technical field of bridge intelligent construction, and specifically discloses a high-pier large-span continuous rigid-frame bridge pier settlement displacement monitoring method and system, which comprises the following steps: collecting image sequences at a main target of a pier and at a reference point sub-target under a main girder construction condition; processing the images, including low-illumination image enhancement and high-precision segmentation of optical marker points under a complex background; identifying the optical marker point center coordinates at the main target and the sub-target based on an adaptive threshold gradient Hough transform (ATGHT), extracting the displacement values at the main target of the pier and the reference point sub-target, and then compensating for the camera self-motion measurement error, so as to obtain the pier settlement value of the main girder in the cantilever construction process. The application utilizes a non-contact device to monitor the pier settlement displacement in the main girder construction stage, realizes non-contact, multi-measurement-point, high-precision and high-efficiency monitoring of the pier settlement displacement of the super-high-pier large-span continuous rigid-frame bridge under a complex background, and has the advantages of low cost, convenient use and flexible measurement point selection.

The invention provides a self-adaptive vacuum defoaming process optimization method based on an intelligent control system, and aims to solve the problems of micro-bubble residue, too high energy consumption and the like caused by the fact that a traditional defoaming technology cannot adapt to liquids with different viscosities, and accurate defoaming is realized through a multi-dimensional sensing technology and strategy generation and regulation. Presetting a high-precision vacuum parameter and a viscosity grading strategy; establishing a liquid characteristic vector model by adopting a YOLOv7 and Hough transformation algorithm, and matching a historical optimal strategy in combination with weighted cosine similarity; generating a defoaming strategy based on the pre-training network and a multi-objective optimization algorithm, and verifying the defoaming strategy in the virtual environment; defoaming is carried out in four stages, and efficient foam breaking is achieved through sectional PID control, resonance vibration and cooperation of a two-stage pump; a Kalman filtering algorithm is integrated, pressure, temperature and vibration data are fused, and vacuum pump power and a heating strategy are adjusted; and constructing a multi-index evaluation system, and driving strategy iterative optimization through a reward signal.

The invention provides a visual celestial navigation method based on sun centroid extraction and sea-sky-line fitting. Belongs to the application of image processing technology in the field of celestial navigation. The method mainly comprises the steps of extracting a high-precision sea-sky-line by using a textural feature segmentation and Canny-Hough transform combined algorithm, and extracting a sky region based on a sea-sky-line segmentation result. A solar facula image candidate region and a sun candidate contour are constructed through a sky region, a dual optimization strategy of region similarity screening and facula feature scoring is proposed, and sub-pixel-level sun centroid coordinate calculation is realized by combining an intensity weighted center method and nonlinear least square two-dimensional Gaussian model fitting. And further constructing a camera coordinate system space vector through the orthogonal projection point of the sea-sky-line and the sun centroid, calculating a sun elevation angle based on a vector included angle, and performing angle compensation by adopting an atmospheric refraction correction model. And finally, fusing multi-moment observation data through a sun shift line positioning method, completing autonomous calculation of the ship position, and improving the positioning precision.

The invention discloses a method and a system for automatically testing a virtual cabin of a whole vehicle based on digital twinning, and relates to the technical field of vehicle testing, a three-dimensional working condition space containing an eye point coordinate set and a polarization angle set is constructed by simulating the height and head micro motion of a driver, and a head-up display image in the three-dimensional working condition space is acquired through a camera; constructing a correction coordinate system and a gray brightness mapping table by combining a standard calibration board, and converting the image into a physical brightness matrix; marking elements in the physical brightness matrix through a brightness threshold value, extracting two parallel straight lines of a main track and a ghosting track through Hough transform and ODR sub-pixel refitting, and calculating a visual angle difference and a ghosting brightness ratio; and generating a test result table containing the three-dimensional working condition index based on a double-threshold joint judgment rule. The system comprises an image acquisition module, a feature extraction module, a data calculation module and a test judgment module, realizes automatic and accurate quantitative automatic test based on the system, and improves the virtual cabin test efficiency and reliability.

The present invention discloses a reflective portable imaging and image processing method for metal mesh defect detection. The method comprises the following steps: constructing a reflective portable imaging system and combining it with a carrying motion device to capture images of the metal mesh in situ; extracting prior information about block defects and broken wire defects from the metal mesh image using a method based on spectral filtering fusion and Hough transform; inputting the captured image and the prior defect information into a low-rank decomposition model to obtain the low-rank portion, sparse portion, and noise portion of the image; and performing threshold segmentation on the sparse portion to extract defects. The method can achieve in-situ imaging measurement of the metal mesh, identify and detect various defects in the metal mesh image, and provide accurate and efficient detection results.

An inspection and a method for the automated identification of an undesired deformation of an edge of a crossbeam of a shelf are disclosed, the method comprising the steps of: projecting, with a laser, a structured light pattern onto the inspection area, wherein the light pattern is aligned parallel to the inspection area and positioned in close proximity to the inspection area; capturing, with a camera, a raw image of the supporting element, which includes at least part of the inspection area and a reflection of the projected light pattern; generating an inspection image from the raw image by computer, wherein lines are extracted from the reflected light pattern of the raw image and converted into a line pattern and stored as the inspection image, and wherein straight lines are selected from the line pattern, in particular by means of a Hough transform;Computer-implemented verification, in the test image, of whether the selected lines include a broken straight line and / or a straight line with a different slope; and computer-implemented determination that the test area exhibits damage caused by deformation if the broken straight line and / or the straight line with the different slope is present.

Provided in the present invention is an artificial-afforestation area extraction method based on a high-resolution remote-sensing image. The method comprises: acquiring a remote-sensing image, and preprocessing the remote-sensing image; inputting the preprocessed remote-sensing image into an intelligent forest extraction model, so as to obtain a distribution map for all forests; performing edge detection and Hough transform line detection on the distribution map for all the forests, so as to obtain artificial afforestation regions and natural forest regions; and calculating the areas of the artificial afforestation regions and the areas of the natural forest regions, so as to obtain statistical results for artificial afforestation and natural forests. The present invention can solve the shortcomings of conventional monitoring means in terms of efficiency, cost and accuracy, and provides technical supports for large-scale forestry investigation and ecological monitoring.

The invention relates to the technical field of optical fiber device precision machining, and discloses an aperture positioning method and system for optical fiber capillary chambering based on image recognition, and the method comprises the steps: obtaining multi-angle image data and process parameter data of an optical fiber capillary, and executing time synchronization and noise preprocessing; performing real-time extraction and enhancement on capillary aperture edge features by using a deep convolutional neural network and an edge enhancement algorithm; constructing a layered aperture recognition architecture, performing morphological optimization, and generating an accurate aperture contour; on the basis of a multi-parameter optimization algorithm of Hough transform and ellipse fitting, parameters of the aperture of the capillary tube are obtained through calculation; and an instruction is generated according to the capillary aperture parameter, and a laser reaming system is driven to execute precise reaming operation. The defects of irregular shape, large size fluctuation and the like of a traditional chambering process can be overcome, and the automation level and the production efficiency of chambering of the optical fiber capillary tube are greatly improved.

The invention relates to the technical field of remote sensing image processing, computer vision and three-dimensional modeling, discloses a monocular remote sensing image building three-dimensional automatic reconstruction method based on an improved UNet network and a multi-parameter coupling model, and aims to solve the problems that a traditional reconstruction method depends on DSM or laser radar data, and is poor in complex terrain adaptability and low in automation degree. According to the method, a single high-resolution remote sensing image serves as input, building and roof pixel-level segmentation is achieved through an improved UNet network fusing a channel-space-semantic three-mode CBAM attention mechanism and a resolution adaptive SMU activation function, and footprint regularization extraction is completed in combination with contour curvature constraint and Hough transform. And constructing an elevation inversion model based on the roof and footprint offset, the imaging geometric parameters and the terrain correction coefficient, and finally generating a three-dimensional white model through a CGA parametric modeling rule. Experimental verification shows that the average absolute error of the method is about 1.06 m (the mountain scene error is smaller than or equal to 1.2 m), the height error of more than 93% of buildings is smaller than 2 m, the method does not need to depend on additional auxiliary data, the process automation degree is high, the method is adaptive to complex terrain scenes, and the method can be directly connected with a smart city and a digital twin system.