102 results about "Maximally stable extremal regions" patented technology

The invention belongs to the technical field of the pattern identification and image processing, and relates to a license plate detection and identification method based on the maximum stable extremal region, which comprises the following steps: extracting the maximum stable extremal region (MSER) to obtain text regions of candidate license plates; adopting one effective feature for description of each extremal region, classifying the extremal regions into 'text' regions and 'non-text' regions with a classifier obtained in prior training, and extracting a license plate from original images based on the characteristics of the structure of the license plate; and describing with the characteristics of the shape context, and completing the character identification through template matching. Since the maximum stable extremal region has the affine invariance, high stability and multi-scale features, and the regions are determined only according to the gray value and are not light-sensitive,the license plate detection and identification method with the maximum stable extremal region as the base is applicable to the complex background, and has the advantages of good stability and high identification rate.

There is provided an efficient, fast image processing apparatus with low error probability for rapidly scrutinizing a digitized video image frame and processing said image frame to detect and characterize features of interest while ignoring other features of said image frame. There is further provided an efficient fast image processing method with low error probability for rapidly scrutinizing a digitized video image frame and processing said image frame to detect and characterize features of interest while ignoring other features of said image frame. In a first embodiment of the invention an image processing apparatus comprises an imaging device coupled to a digital electronic image processor. Video data from the imaging device is linked to a location data source. Objects of interest in a scene are identified by comparing computed Maximally Stable Extremal Regions (MSERs) of captured images with MSERs of images of objects contained in a object template database.

The invention belongs to the technical field of image processing, and specifically relates to an English text detection method with a text direction correction function. The method comprises the following steps of: respectively carrying out maximally stable extremal region detection on each channel of an English text image so as to obtain candidate text regions; establishing a convolutional neuralnetwork model-based classifier, and filtering wrong candidate text regions so as to obtain preliminary text regions; grouping the preliminary text regions by utilizing a double-layer text grouping algorithm; and carrying out direction correction on the grouped preliminary text regions so as to obtain a corrected text. According to the method, an enhanced multichannel MSER model is utilized to obtain more refined text regions, a parallel SPP-CNN classifier is imported to better distinguish text regions and non-text regions, images with any sizes can be processed, and pool features can be extracted under multiple scales, so that more features can be known through multilayer space information of source images; and the method is capable of processing slightly inclined scene texts.

The invention relates to the technical field of mode recognition and computer vision, and in particular relates to a document key area detecting and positioning method using edges and text areas. Thegrayscale image of a document image is acquired by preprocessing the document image, and a document area image is acquired. A maximum stable extremum area algorithm is used to extract candidate areasfrom the document area image, and the candidate areas are filtered. Text areas are retained. The filtered text areas are combined into text blocks. The relative positions of the text blocks are compared. Field contents represented by the text blocks are determined, so that information of all kinds of documents can be effectively extracted. The method has good versatility and practicality, can be widely used for image recognition of different occasions and different carriers, has the advantages of high recognition degree and fast efficiency, and is fast and robust.

The invention discloses an extremum connected domain based character detection method in a natural scene image. The method comprises the steps of acquiring the natural scene image firstly, and extracting a maximally stable extremal region separated from the natural scene image, wherein the output of the separated maximally stable extremal region is a series of non-overlapped regions, and each region is a connected component; after obtaining the connected components, extracting various features of the connected components, wherein the feature combination can well express the connected components; from the character structure, performing intra-character combination firstly, and then performing inter-character combination, wherein the intra-character combination method is for detecting a single character, and the inter-character combination method is for detecting text lines; and finally, analyzing angular point distribution of the text lines, screening to obtain a text region, and demarcating the text region by a rectangular frame. The method starts with character edge features, and has good pertinence in character detection, thereby having higher initiative and accuracy.

The present invention discloses a rapid image text detection method based on a multi-channel and multi-dimensional cascade filter. The problem is mainly solved that the recall ratio is low and the speed is slow in the prior art. The method comprises: 1) extracting a maximum stable extremum region in the different channels and scales of an input image as a character candidate region; 2) removing the background region in the character candidate region by employing a cascade filter from coarse to fine, namely setting a threshold value for the morphological features of the character candidate region, and performing the first grade coarse filtration; setting thresholds for the stroke width and the stroke width variable coefficient of the character candidate region, performing the second grade coarse filtration, then removing the overlapping regions, and employing a convolution neural network binary classifier perform fine filtration; and 3) aggregating the region into the character string according to the geometry and the position feature of the character candidate region after cascade filter through a graph model. The rapid image text detection method based on the multi-channel and multi-dimensional cascade filter has high recall ratio, high accuracy and fast speed, and can be used for detection of image text at various interference surroundings.

The invention discloses a crown information extraction method and system based on the high spatial resolution remote sense image. The method comprises the following steps: obtaining a forest land remote sense image; preprocessing the remote sense image, obtaining the preprocessed remote sense image; adding a forest form map on to the preprocessed remote sense image, using the subcompartment boundary of the forest form map as the clipping region to clip the image information corresponding to each subcompartment in the forest form map, extracting single forest stand image information; utilizing the maximally stable extremal region method to divide the crown and background areas of the single forest stand image information; and extracting the crown image of single tree in the multi-tree crown area, and calculating the number of trees and the crown factor of each tree in the crown distribution map. By using the method and system of the invention, the connected crowns are easy to separate under the condition that the crown image contains fewer pixels, the operation efficiency is high; and the efficiency and degree of automation of the forest resource investigation can be efficiently increased, the forest resource information can be accurately obtained, and manpower and material resources are saved.

The invention belongs to the technical field of video retrieval, relates to related image processing knowledge, and in particular relates to a video text detection method. The video text detection method is characterized by extracting video subtitles from a video to be detected and being used for recognition and video retrieval. The video text detection method comprises the steps of: firstly enhancing the text boundary of an input image by utilizing a gradient amplitude map (GAM); secondly, filtering partial background interference by using morphological filtering in two directions and enhancing the contrast of text and background; thirdly, detecting the saliency map of video text by using a maximally stable extremal region (MSER) detector, and acquiring the optimal segmentation of the text by utilizing Graph Cuts; and finally, connecting the texts to a text row by utilizing the geometrical distribution feature of the texts, and removing non-text regions by using multiframe confirmation and a certain starting education method. The detection method has the effects and benefits that the sensitive technical difficulties of blur text boundary, low contrast and complicated background and the like in text detection are overcome, and the detection results can be directly used for character recognition.

The invention relates to a large viewing angle image matching method capable of combining region matching and point matching. The method comprises the steps of 1, inputting two images having large viewing angle changes; 2, carrying out region detection on the images by a maximally stable extremal region (MSER), and fitting an elliptic region by the mean value and the variance of the region; 3, normalizing the elliptic region into a circular region, and describing the circular region by a scale invariant feature transform (SIFT) descriptor; 4, adopting the nearest-neighbor than the next-nearest neighbor strategy, and selecting the initial region matching pair; 5, in the region matching pair, detecting feature points by an SIFT method; 6, describing the feature points to obtain an MSER-based 128-dimensional descriptor and a 2-dimensional space descriptor; 7, adopting a similarity strategy combined with the distance, and selecting an accurate matching point pair in the two images. The large viewing angle image matching method overcomes the defect that in the prior art, the description of the feature points does not have affine invariant and leaves out of consideration of space information, and can extract the matching point pair with higher accuracy so as to enable the matching point pair to be better used for image registration.

The invention provides a text region positioning method and apparatus for an image. The image is an RGB image. The method comprises the steps of graying the image to obtain a gray map of the image; converting the image into an HSV space and obtaining an H channel map and an S channel map of the image; calculating an image gradient to obtain a gradient map of the image; obtaining all maximally stable extremal regions of the gray map, the H channel map, the S channel map and the gradient map; combining all the maximally stable extremal regions of the gray map, the H channel map, the S channel map and the gradient map as a candidate text region; judging whether the candidate text region is a text or not by using a neural network and deleting a non text region; and determining a text region of the image according to the position of the candidate text region after removal of the non text region. According to the method and apparatus, the text region of the image can be positioned according to brightness information and color information of the image.

The invention relates to the field of image recognition, specifically discloses a traffic sign recognition method, a storage medium and processing equipment, and aims at completing the detection and classification of traffic signs in a same module. The traffic sign recognition method comprises the following steps of: extracting maximally stable extremal regions from a normalized RGB channel and anormalized grayscale channel as candidate regions of a traffic sign; carrying out screening on the basis of a preset traffic sign feature condition, zooming the obtained regions to fixed sizes, and decreasing an average value of the preset RGB channel to obtain a fourth group of traffic sign candidate regions; inputting the fourth group of traffic sign candidate regions in a traffic sign recognition network to extract a detection result and a classification result; and removing overlapped regions through a non-maximal suppression algorithm so as to obtain a final recognition result, wherein the final recognition result comprises position, size, specific strategy and confidence coefficient information of the traffic sign. Through the method, the detection and classification of traffic signscan be completed in a same module.

The embodiment of the invention discloses an image detection method and device. The image detection method comprises the steps that an image to be detected is obtained, and a maximally stable extremal region (MSER) is extracted from the image to be detected, wherein the MSER is a communication region, and a text region in the image to be detected is obtained by filtering the MSER. The MSER is extracted to serve as a candidate region by extracting the MSER from the image to be detected in a communication region dividing mode, then the extracted MSER is filtered and screened, and finally the text region in the image to be detected is obtained. Region division is conductive to decrease of calculation amount and improvement of the detection efficiency, meanwhile the interference of an image background can be decreased by extracting the MSER, and the accuracy rate during detection of images having complicated backgrounds can be improved.

The invention discloses a space debris star extraction and positioning method, and aims to solve the technical problem of difficulty in segmenting adjacent space debris targets in a current method. According to the technical scheme, a data structure tree is constructed in a candidate space debris region through an MSER (Maximally Stable Extremal Region) algorithm to extract a space debris pixel set in the region and determine the number of space debris stars. In a pixel point tree structure obtained through the MSER algorithm, a training sample is generated for subsequent fitting of a GMM (Gaussian Mixture Model) model by detecting a maximum MSER in the candidate space debris region. In a GMM-based positioning process, the mean value and covariance of Gaussian components in the mixture model is solved iteratively through an EM (Expectation and Maximization) algorithm, and the corresponding mean value is the mass center of each space debris star light spot. According to the method, the characteristics of the MSER and the GMM model are combined, and the mass centers of near-distance dispersed light spots are solved together through the GMM model, so that the technical problem of difficulty in segmenting the adjacent space debris targets is solved.

The invention provides a free scene text detection method based on affine transformation and relates to the field of image processing. According to the free scene text detection method provided by theinvention, MSER (Maximally Stable Extremal Regions) and SWT (Stroke Width Transform) are combined to realize text detection; then a vertex of a peripheral quadrilateral of each character is utilizedand is combined with an inertia principal axis to find out a quadrilateral for reflecting text distortion characteristics, so as to find out an affine parameter to carry out affine transformation; finally, detection of an image text and elimination of distortion are realized. According to the free scene text detection method provided by the invention, the accuracy of the text is remarkably improved; compared with a single text detection method based on a communication region, the recalling rate and the detection efficiency can be improved and automatic affine transformation is carried out; finally, the detection and the elimination of the image text are realized; compared with other manual affine transformation, the working efficiency is greatly improved and OCR (Optical Character Recognition) in the future is more accurate.

The invention, which belongs to the technical field of computer vision and particularly relates to text localization study in a complex scene, discloses a railway scene text localization method based on combination of a maximum stable extreme value region and a stroke width. On the basis of an improved histogram equalization algorithm, an original image is preprocessed, so that the contrast of an image can be improved effectively; and with an MSER algorithm, a weak target area in a railway scene is detected and a non-text area is removed based on a stroke width feature of a character, so that the false drop rate is reduced and thus problems of difficult text detection in the railway scene and difficult realization of accurate text localization can be solved. The method has the following advantages: a block sliding window search strategy is employed based on the spatial structure characteristic of the text line, so that the computational complexity can be reduced. The method can be applied to a complex railway character localization scene.

The invention relates to a meteorite crater matching method based on area ratio, which solves the problems of relatively large rotation of images, mistake matching, mismatching and the like under the conditions of scaling and deformation existing in the existing meteorite crater matching method. The meteorite crater matching method based on area ratio relates to the technical field of image processing, comprising the following steps: selecting out local images in a global image on the surface of a planet as a reference image; then carrying out extraction of meteorite craters on an image shot in landing process based on maximally stable extremal region method; carrying out ellipse fitting on the extracted meteorite craters in the shot image, and calculating the area of fit ellipse; respectively calculating the area ratio between different meteorite craters in the shot image and the area ratio between different meteorite craters in the reference image; and finally utilizing Hausdroff distance as similarity measurement, and matching the meteorite craters detected in the shot image with the meteorite craters in the reference image. The method can efficiently prevent the disadvantages of conventional matching methods, and is an ideal method for carrying out meteorite crater matching in landing process.

The present invention relates to a night vehicle license plate positioning method based on MESR and SWT combination, and provides a novel license plate positioning method based on the MESR and SWT combination aiming at the problem that a conventional license plate positioning method can not position the license plates effectively on a night poor illumination condition or due to the vehicle speed influence. The method comprises the steps of after the contrast ratio of an original image is enhanced, carrying out the Canny edge detection and the MSER extraction, segmenting the MSER by the edge expansion and screening the segmented MSER according to the license plate character geometrical characteristics; then carrying out the SWT based on the morphological processing in the screened region; finally, aggregating a candidate region, and combining the license plate geometrical characteristics to finish the fine positioning of the license plates. By the test verification, the method is high in positioning accuracy, can position the license plates accurately on different scenes and different illumination conditions in the daytime, at the same time, can position the night low-resolution license plates effectively, and has much higher positioning accuracy and robustness than other conventional license plate positioning methods on a low-resolution condition.

The invention discloses an edge-reinforced color space maximally stable extremal region detection method which comprises the steps of using a multiscale probability of boundary (mPb) edge detection method to detect edges of a color image to obtain edge information and obtain distance a weighting coefficient according to the edge information and a distance conversion formula, calculating the surface characteristics and the distance value of difference between adjacent pixels in a color space, using the distance weighting coefficient to weight the distance value to obtain a final distance set, obtaining a distance threshold set according to the distribution situation of the final distance set, combining the adjacent pixels with the distance value smaller than the threshold to the same region by continuously increasing a distance threshold, and extracting regions with the size change rate reaching to the local minimum along with the change of the threshold as the maximally stable region characteristics. The edge-reinforced color space maximally stable extremal region detection method comprehensively uses edge information and color information of the image to further improve fuzzy mapping invariance in keeping the self invariance of the maximally stable extremal region to the greatest degree.

The invention discloses a road traffic sign detecting method based on phase symmetry. The flows of the method comprise red-blue standardization pretreatment, phase symmetry detection, morphological filtering, and maximally stable extremal region (MSER) feature detection. Specifically, the method comprises: performing color pre-segmentation on an image by a red-blue standardization method according to the characteristic red or blue of a traffic sign; computing a highlight prominent traffic sign candidate region by using phase symmetry according to the symmetry of the external shape of the traffic sign; and extracting a sign candidate region from a phase symmetry energy diagram by using MSER features. The method is suitable for detection of road traffic signs in complex natural environment and has better detection effects. Further, compared with a conventional detection method just based on color or shape information, the method has better robustness in complex road scenarios by using the color and shape characteristics of the traffic signs.

The invention discloses a meteor crater detecting method based on bright and dark area pairing, relates to the technical field of image processing, and aims to solve the problems that the conventional meteor crater extracting method based on a landmark navigation task in a planet landing section is high in mis-extraction rate and hard to detect irregular meteor craters, and the like. The method comprises the following steps of: carrying out primary detection on an image obtained by an optical camera during the planet landing process based on a maximum stable extreme value area method; extracting shadow areas and bright areas in the image; deleting overlarge or oversmall areas; taking the moment center of each detected shadow area of the meteor crater as the center; searching the bright areas within a circle with a R radius, wherein the difference of gray level average values of each bright area and each shadow area is larger than a given threshold; calculating a vector from the moment center of the shadow area to the moment center of each bright area; and calculating an included angle between the vector and a projection vector of a sunshine vector in a camera image plane, wherein if the value of the inclined angle is less than a given threshold, a meteor crater is formed by the shadow area and the bright area.

The invention discloses a method for extracting a maximally stable extremal region with scale invariance. The method includes the steps that firstly, an initial maximally stable extremal region is detected in an original image through a maximally stable extremal region algorithm; then a scale pyramid of the initial maximally stable extremal region is built through M-scale wavelet transform, characteristic points with the scale invariance are determined in the scale pyramid according to energy operators of an M-scale wavelet transform coefficient, extremal regions corresponding to the characteristic points are obtained from all layers of images of the scale pyramid of the maximally stable extremal region, and the maximally stable extremal region with the scale invariance is extracted through the stability indexes of the extremal region in a multi-scale space; finally, the maximally stable extremal region with the scale invariance is adjusted to be in an oval shape, and the final maximally stable extremal region with the scale invariance is obtained. According to the method for extracting the maximally stable extremal region with the scale invariance, the scale invariance and the maximally stable extremal region are combined, the maximally stable extremal region is extracted, and full affine invariance is achieved.

The invention discloses a detection method and system of a black smoke vehicle. The detection method of a black smoke vehicle includes the steps: through the image information acquired by a crossing monitoring device, based on the improved black smoke detection algorithm of the maximum stable extremum region, performing color enhancement transformation on the black smoke region to highlight the black smoke region in the acquired image after transformation; and detecting the maximum stable extremum region on the image which is obtained through transformation, segmenting the black smoke region,thus avoiding the defect that a traditional method based on color information cannot accurately segment black smoke, accelerating the speed of detecting the black smoke vehicle, and greatly reducing the complexity of detecting the black smoke vehicle. The detection system of a black smoke vehicle is convenient to operate, and has bigger market application space.

The invention discloses a maximally stable extremal region and SVM based traffic sign recognition method, which is characterized in that a maximally stable extremal region (MSER) algorithm is adopted to detect a traffic sign part in an RGB image, graying treatment is performed on the image, and the stability of the MSER algorithm is brought into play. The method applies an HOG eigenvector to act as a method for edge detection and image segmentation for a region to be identified, thereby being capable of suppressing influences brought about by translation and rotation to a certain extent. The method is insensitive to changes of illumination, so that the interference brought by changes in illumination intensity to the image can also be reduced. An SVM classifier is used in the stage of classification and recognition to avoid easy error occurrence performance of manual marking and great time consumption of machine training. The method well balances the requirements of accuracy and timeliness, and realizes automatic detection and recognition for traffic signs. The method performs recognition on test images in a German traffic sign detection benchmark database and acquires good effects.

The invention discloses an image-text area recognition method, comprising the steps of processing an image to be recognized to obtain a first binary image to be processed which has multiple maximum stable extreme value areas; aggregating the multiple maximum stable extreme value areas to obtain multiple alternative boxes, and using the first binary image to be processed as a mask for the image tobe recognized so as to obtain a second binary image to be processed; deleting non-text areas from maximum stable extreme value areas of the second binary image to be processed; adding the alternativeboxes to the second binary image to be processed which is subjected to deleting operation, and using the alternative boxes containing the maximum stable extreme value areas as text areas. The invention also discloses a television and a readable storage medium. The image-text area recognition method, the television and the readable storage medium provide improved recognition precision for image-text areas.

The invention relates to a region-matching-based fast electronic image stabilization method, which specifically comprises the following steps: (1) reading video image frames, and selecting a first frame of an image as a reference frame; (2) sequentially performing region matching on current frames and the reference frame to obtain matching scores by adopting an MSER (maximally stable extremal region) algorithm, performing wavelet transformation, sequencing the obtained matching scores from high to low scores, and selecting regions corresponding to the first n matching scores; (3) performing region labeling on the regions corresponding to the selected first n matching scores, performing FAST characteristic point detection on the labeled regions, and extracting characteristic points; (4) performing characteristic point matching, and estimating a motion parameter; (5) performing reverse motion compensation on the image. According to the method, an image sequence is stabilized by virtue of an image stabilization algorithm, so that the method has the advantages of fastness, high accuracy, low power consumption and the like.