Patents
Literature
Hiro is an intelligent assistant for R&D personnel, combined with Patent DNA, to facilitate innovative research.
Hiro

89 results about "Perspective distortion" patented technology

In photography and cinematography, perspective distortion is a warping or transformation of an object and its surrounding area that differs significantly from what the object would look like with a normal focal length, due to the relative scale of nearby and distant features. Perspective distortion is determined by the relative distances at which the image is captured and viewed, and is due to the angle of view of the image (as captured) being either wider or narrower than the angle of view at which the image is viewed, hence the apparent relative distances differing from what is expected. Related to this concept is axial magnification -- the perceived depth of objects at a given magnification.

Method and apparatus for resolving perspective distortion in a document image and for calculating line sums in images

Perspective distortion is estimated in a digital document image by detecting perspective pencils in two directions, one being parallel to text lines, and the other being parallel to the boundaries of formatted text columns. The pencils are detected by analyzing directional statistical characteristics of the image. To detect a pencil, a first statistical line transform is applied to transform the image into line space, and a second statistical score transform is applied to transform the image into pencil space. A dominant peak in pencil space identifies the perspective pencil. In addition, a computationally efficient line summing technique is used for effecting sums of pixels along inclined target lines (or curves) through an image. The technique includes pre-generating partial sums, and summing along step segments of a target line using the partial sums.
Owner:XEROX CORP

Water level gauge positioning and water level measuring method based on image processing

The invention discloses a water level gauge positioning and water level measuring method based on image processing, wherein the water level gauge positioning method comprises the following steps: calculating the longitudinal inclination angle of a mark rod main body of a pre-extracted water level gauge image; obtaining a water level gauge main body image after inclination correction by adopting arotation correction technology according to the longitudinal inclination angle of the mark rod main body; correcting perspective distortion by using a calibration point technology for the main image of the water level gauge after inclination correction; using a projection method to carry out adjacent pixel repair on a missing value part in the distortion-corrected image; carrying out accurate target clipping on the repaired image according to the coordinate equation of the left and right dividing lines; the starting position of the water level gauge and the position of the boundary line with the water surface are identified and the main body position of the water level gauge is cut according to the starting position and the position of the boundary line with the water surface. According tothe water level gauge positioning and water level measuring method based on image processing, the distortion correction is carried out on the water level gauge image after the inclination correction,the distortion problem caused by the non-parallel lateral boundary of the target caused by the viewpoint selection angle of the image is solved, and the precision of target extraction is improved.
Owner:NANJING UNIV OF POSTS & TELECOMM

A metal plate strip product label information identification method based on computer vision

The invention discloses a metal plate strip product label information identification method based on computer vision. the position of a product label area is obtained through segmentation of a lightweight network; the coordinate information of the product label is obtained through an image processing means; correction to enabline see-through transformer is realized, the VGG16 is used for identifying the rotating text; character rotation small-angle registration is carried out by using a variance method; the text position detection precision and the text recognition precision are effectively improved; YOLOv3 and ENet are adopted, so that text correction and position acquisition are faster and more accurate; the loss of the computer and the requirement on the performance of the computer areeffectively reduced; the detection of the text with the uncertain length is realized by utilizing the characteristics of the LSTM in the CRNN; The detection performance is effectively improved, the good recognition performance is achieved in natural scenes such as non-uniform illumination, complex backgrounds, multi-language mixing, text complex formats, product label picture rotation, affine distortion and perspective distortion, and convenience is provided for inputting of label information of metal plate strips.
Owner:NORTHEASTERN UNIV

Tool setting method of mechanical arm feeding type laser etching system

The invention discloses a tool setting method of a mechanical arm feeding type laser etching system, belongs to the field of non-traditional machining, and relates to a tool setting method of a mechanical arm feeding type laser etching system based on a machine vision technology. The laser etching system is optimized and improved; a vision sensor and an active projection indication laser are added; and an interactive information obtaining function between heterogeneous space non-cooperative electromechanical equipment in the system is added. The problems of non-front-view perspective distortion and vision sensor imaging distortion introduced by an inclined vision measurement configuration are considered; and an inverse perspective transformation geometric correction and camera distortion compensation technology is adopted to correct image information. The precise alignment between a laser etching focusing plane and a part initial characteristic point to be machined in a working space is realized by combining multi-frame sequence image dynamic transformation and multi-axis driving feeding device space shape and position characteristic information. The method solves the precise positioning tool setting problem in the laser etching machining initial phase, and improves the machining precision of target workpieces and the finished product quality.
Owner:DALIAN UNIV OF TECH

Self-adaptive receptive field crowd density estimation method based on cavity convolution

The invention discloses a self-adaptive receptive field crowd density estimation method based on cavity convolution, which belongs to the field of computer vision, and comprises the following steps: segmenting an original data set image and a crowd density map to obtain image blocks and crowd density map blocks; constructing and training an adaptive receptive field population density estimation network, wherein the model comprises a cavity convolution module and a classification module, the classification module is used for classifying the segmented image blocks, the cavity convolution moduleadaptively selects a cavity convolution sub-network corresponding to the receptive field according to the image block category output by the classification module, and performs feature extraction on the segmented image blocks to obtain a crowd density map; and inputting the picture to be predicted into the trained adaptive receptive field crowd density estimation model to obtain a crowd density estimation result. According to the method, the cavity convolution sub-network corresponding to the receptive field can be adaptively selected for crowd density estimation, and the problem of perspective distortion is solved, so that the accuracy of crowd density estimation is improved.
Owner:HUAZHONG UNIV OF SCI & TECH

Table correction and recognition method based on combination of image processing and deep learning

The invention relates to the technical field of image processing and image recognition, in particular to a table correction and recognition method based on combination of image processing and deep learning, which comprises the following steps: step 110, acquiring original image data of a table; step 120, image preprocessing; step 130, positioning a character area; 140, reconstructing table information; by designing and improving the existing form recognition method, the accuracy of form recognition is improved by performing text direction judgment, inclination correction and perspective distortion processing when the form image is recognized, and the problem that the form recognition accuracy is low after the form image is obtained by using equipment in the existing method for recognizingthe form in the image is solved. The table row and column frame line positions are detected by analyzing the optical characteristics of the whole-page digital image so as to detect the format structure of the table, and the method is generally only suitable for the conditions that the input image quality is good, the table positions and formats are fixed, and the table frame lines are obvious, andthe problems of character direction overturning, inclination, perspective distortion and the like exist in the image.
Owner:晶璞(上海)人工智能科技有限公司

Lane guide arrow identification method in intersection monitoring environment

The invention discloses a lane guide arrow identification method in an intersection monitoring environment, belongs to the field of computer vision, and relates to image processing related knowledge.The method comprises the steps: acquiring a traffic video image sequence based on an intersection monitoring camera; establishing a lane background model by using a background modeling technology; detecting a lane line by utilizing Hough transform on the basis of the established lane background model, taking an end point of the detected lane line as an original coordinate point of image perspective transformation, and finishing perspective transformation in combination with a least square method to obtain a lane top view, so as to correct perspective deformation generated by shooting of a camera; then, respectively projecting the lane top view in the vertical direction and the horizontal direction, and carrying out screening segmentation according to the ratio of the length of a standard guide arrow to the width of the lane to obtain all target area images only containing a single guide arrow; finally, performing similarity matching on the guide arrows obtained through segmentation anda standard guide arrow template one by one, and recognizing the types of the guide arrows of all lanes.
Owner:南京慧视领航信息技术有限公司

Method for detecting and counting distribution of dense crowds in video

The invention provides a method for detecting and counting dense crowd distribution in a video. Firstly, acquiring a large number of videos containing crowds with different densities to construct a data set; then constructing a deep neural network of multi-scale feature fusion and an attention mechanism, inputting the training set into the network, outputting prediction results of a corresponding crowd density map and an attention map, constructing a loss function model in combination with the real density map and the attention map for training, and generating an optimized network; obtaining a density map of a crowd video image through optimized multi-scale feature fusion and deep neural network prediction of an attention mechanism, furthering performing point clustering on the estimated density map by using a grid-based hierarchical density space clustering method to identify a group, and obtaining the number of people and position information of the group quickly. According to the invention, the problems of perspective distortion, scale change and background noise influence of the camera can be solved, and the counting precision and stability are improved; and meanwhile, the crowd is divided into groups, so that the distribution condition of the crowd can be visually displayed.
Owner:WUHAN UNIV

Intelligent and accurate inspection method and system for power tower by unmanned aerial vehicle

The invention discloses an intelligent and accurate inspection method and system for a power tower by an unmanned aerial vehicle. The method comprises the steps: designing and achieving the inclined frame detection of a target based on a target detection algorithm, and extracting an inclined frame target object; designing a target three-dimensional detection network, namely fusing a target inclined frame detection task and a depth estimation network, running on embedded equipment in a multi-thread mode, and obtaining position information and depth information of a target; calculating a pose vector group of the target by using the position information and the depth information of the target; and calculating a distance correction amount and a holder angle correction amount of the unmanned aerial vehicle according to the pose vector group of the target, so that the unmanned aerial vehicle performs inspection photographing after correction. The system comprises an inclined frame target detection unit, a position and depth information acquisition unit, a pose vector group acquisition unit and a distance and holder angle correction unit. According to the invention, the problem of perspective distortion of an image during inspection shooting of the unmanned aerial vehicle can be solved, and the quality of the inspection shooting image is improved.
Owner:SHANGHAI UNIV

Camera calibration device and method for microscopic three-dimensional shape measurement system

The invention relates to the field of three-dimensional morphology measurement, and particularly discloses a camera calibration method and device for a 3D microscopic shape measurement system, and the method comprises the steps: generating a phase target, carrying out the snapshot of a camera, carrying out the phase analysis of a snapshot calibration picture, generating a phase diagram, extracting the feature point of the circle center of an integer pixel from the phase diagram, performing equiphase line extraction in a D * D neighborhood of a pixel size such as 35 * 35 by taking the whole image circle center feature point as a center, performing ellipse fitting to obtain a sub-pixel precision circle center feature point, and calculating a camera parameter initial value according to a telecentric camera calibration algorithm according to a pixel coordinate of the sub-pixel precision circle center feature point and a corresponding world coordinate; and obtaining high-precision camera parameters by using a repeated iteration method. The device and method have very high robustness to image blurring and noise, the target is more accurate, and the extraction of the circle center feature point is not influenced by lens distortion and perspective distortion.
Owner:SICHUAN UNIV
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Try Eureka
PatSnap group products