838 results about "Region of background" patented technology

The invention discloses an identification method and system of lateral deviation of a vehicle driving route. According to the method and system, firstly, an image, which is shot by a camera mounted ona vehicle, of a road along a vehicle running direction is acquired; an internal-parameter matrix of the camera is utilized to carry out de-distortion processing on the video frame, and inverse perspective transformation is carried out on a selected to-be-transformed region src; then threshold segmentation is carried out on the video frame to separate lane lines from a background region; then locations where maximum values appear are derived by counting to use the same as intersection points of the lane lines and the bottom of the image; and whether correction of lateral deviation of the driving route needs to be carried out is judged according to separation distances between the locations of the intersection points and a preset lateral mounting location of the camera on the vehicle. In acase where the vehicle deviates from a road center too far, the method and system can know the case in time, and remind a driver or carry out correction; and the method and system can accurately detect all of sudden illumination changes, tree shade shading, road surface stains and many other conditions, and are high in applicability, low in costs, high in precision and better in real-time performance and stability.

A blur classification module may compute the probability that a given pixel in a digital image was blurred using a given two-dimensional blur kernel, and may store the computed probability in a blur classification probability matrix that stores probability values for all combinations of image pixels and the blur kernels in a set of likely blur kernels. Computing these probabilities may include computing a frequency power spectrum for windows into the digital image and/or for the likely blur kernels. The blur classification module may generate a coherent mapping between pixels of the digital image and respective blur states, or may perform a segmentation of the image into blurry and sharp regions, dependent on values stored in the matrix. Input image data may be pre-processed. Blur classification results may be employed in image editing operations to automatically target image subjects or background regions, or to estimate the depth of image elements.

The invention discloses a scene text recognition method based on sparse coding characteristics, and relates to computer vision and pattern recognition. The method includes the steps: inputting a natural scene text image to be recognized; by the aid of a multi-scale sliding window method, detecting and recognizing a window area in the image by a character classifier, for each character class, determining a large output area of the classifier as a candidate character area, determining a small output area as a background area, finding the candidate character area in the image, retaining the area with the largest output value of the classifier and the corresponding character class for the area with a large overlapping ratio by the aid of a non-maximum suppression method, and removing the repetitive and redundant candidate character area to obtain a character detection result; combining detected characters into a word or text line; outputting a scene text recognition result. Structural characteristics of the characters can be more effectively expressed and extracted, so that the recognition rate of a scene text is increased.

The invention discloses a tobacco leaf grading method based on reflection, perspective and microscopic images. The method provided by the invention comprises the following steps: (1) acquiring a high-definition color reflection image, a high-definition perspective image and a high-definition microscopic image of tobacco leaf at real time and transmitting the collected images to an industrial control computer; (2) preprocessing the images to quickly position the edge of tobacco leaf and segmenting tobacco leaf out of a background area to record as a tobacco image; (3) extracting features of the image, wherein the features include shape feature, color feature and texture feature; and (4) analyzing the above features of tobacco leaf by an Adaboost sorting algorithm, and determining the grade of tobacco leaf through part, color and grade classification of tobacco leaf. Through a device for full-automatic machine grading, human factor interference can be avoided, and labor cost is greatly reduced. Thus, huge hidden costs of training, management and the like caused by manual grading are completely avoided.

The present invention provides a technique of accurately extracting areas of characters included in a captured image. A character extracting device of the present invention extracts each character in an image with compensated pixel values. In more detail, the character extracting device integrates pixel values at each coordinate position in the image along a character extracting direction. Then, the character extracting device predicts the background area in the image based on the integrated pixel value. The compensated pixel values are compensated based on integrated pixel values at the predicted background area from integrated pixel values at each coordinate position.

A segmentation of an object depicted in a first visual representation may be determined. The segmentation may include for each image a first respective image portion that includes the object, a second respective image portion that includes a respective ground area located beneath the object, and a third respective image portion that includes a background area located above the second respective portion and behind the object. A second visual representation may be constructed that includes the first respective image portion and a target background image portion that replaces the third respective image portion and that is selected from a target background image based on an area of the third respective image portion relative to the respective image.

The invention relates to a kilowatt-hour meter image automatic identification method which comprises the following steps: 1. image preprocessing: detecting vertical texture of a panel image by using Sobel operator, preliminarily removing the background area by a projection method, extracting the area with abundant vertical texture by an expansion method, and carrying out binarization treatment on the image by a adaptive threshold segmentation method based on an integral projection method; 2. precise positioning of indicating value and bar code: by combining an intelligent judgment method on the basis of indicating value intervals and length-width ratio characteristic of numeric characters under the complex image background, adapting to precise positioning of indicating values of different types of kilowatt-hour meters on the basis of vertical edge detection of the Sobel operator and morphological treatment; carrying out horizontal scanning on the bar code area to extract the bar code characteristic area; 3. bar code identification: identifying different character bar codes by using a similar edge distance normalization method; and 4. indicating value identification: extracting the indicating value by a PCA (principal component analysis) method. By using the PCA character recognition method, various character indicating values can be precisely identified, including identification of half-character.

When one viewing condition parameter is applied to all pixels, a visual effect between a single color and a background expressed on a rasterized image cannot be reflected on a color matching result. To prevent this, when color matching using a human color appearance model is to be performed, the distance between an image and a viewer and the resolution of the image are input. On the basis of the input distance and resolution, a stimulus region, adjacent region, and background region based on a plurality of field angles (e.g., 2°, 4°, and 10°) for a pixel of interest on the image are defined, thereby determining a viewing condition parameter and performing color matching.

The invention provides an image coding method which is applied to the field of computer image processing. The method comprises the following steps of: firstly, marking the region of interest in an image; performing wavelet transform of the original image; finding the wavelet coefficient corresponding to the pixel of the region of interest after the transform in a wavelet coefficient matrix; then,quantizing the background region and the region of interest according to different quantizing standards so that the coefficient of the region of interest is located in a higher plane; performing threshold quantization of the wavelet coefficient matrix nearby; and coding by a method based on SPIHT (set partitioning in hierarchical trees). The method provided by the invention guarantees the high compression ratio as well as the display quality of the region of interest of a user; and when transmitting or browsing an image in a wireless network with a relatively low rate, the method can enable the user to get the most important information in the image very soon.

The invention discloses a cutting-resistant JPEG image compression history and synthetic tamper detection method. The method comprises the following steps: (1) calculating compression history trace and cutting position of an image to be detected; (2) determining whether the image to be detected has a tampered operation or not; (3) carrying out tamper area positioning on the image determined in the step (2), respectively figuring out difference images to RGB image three channels according to the mode of the step (2), carrying out binary processing on the difference images, carrying out logic and operation on the obtained three channel binary images to obtain a new binary image, carrying out morphological corrosion expansion operation on the image, and mapping back to the image to be measured to obtain a tampered area positioning image. According to the invention, problem is solved that block position of a JPEG image may be destroyed after cutting, which leads to sectional or complete failure of the prior art. Also effective detection and tampered area positioning on an image with cutting background area can be carried out by the method provided in the invention.

A self-contained system uses light reflectivity to examine intensity of a dyed spot on a device membrane surrounded by background area to discern information about the specimen that produced the spot. In a preferred embodiment, a master clock alternatively drives one LED focussed upon the spot center, and then drives two LEDS focused on the background area. Light reflected from the spot and background is detected by preferably two photodetectors (“PDs”) spaced-apart a multiple of 90° azimuthal.

The invention discloses a printed matter defect detection method and device based on artifact elimination. The method comprises steps of image positioning and registration, carrying out the positioning and registration of a standard template image and a target image of a to-be-detected printed matter based on a line mod feature point positioning and registration algorithm; target image artifact elimination, respectively dividing the standard image and the target image after positioning and registration into a plurality of sub-blocks with the same size, and eliminating artifacts caused by localdeformation of the target image by using a sub-block neighborhood sliding artifact elimination method; extracting background region defects of the final difference image; extracting a contour of thestandard image and performing mathematical morphology expansion operation to obtain a contour mask, segmenting the final differential image into a contour area and a non-contour area by using a contour mask covering method, extracting and judging defects in the non-contour area and the contour area of the final differential image, and finally integrating, outputting and displaying the defects. Themethod can be used for successfully detecting the defects such as smudginess, incompleteness, ghosting, shifting, scratching and skip printing.

A fast splitting method for image includes the following steps: a) splitting the image into three parts of foreground area, unknown area and background area; b) calculating colour component F and B for foreground and background by setting any point C in unknown area as colour of according to the split areas; and c) adjusting colour value for F and B according to the estimated colour component of foreground and back ground, and finally to estimate alpha value of c point.

The invention discloses a stock ground material volume measurement method and system, and the method comprises the steps: S1, obtaining the point cloud data of a stock ground material pile through a laser scanner; S2, extracting color features of the point cloud data, and performing coarse segmentation on a stock ground background region and a stock pile target region according to the color features to obtain point cloud data of the stock pile target region; S3, performing fitting registration on the point cloud data of the material pile target area by adopting a coarse registration method anda fine registration method; and S4, according to the point cloud data obtained through fitting registration, calculating the volume of the material pile by using a mode of combining a Monte Carlo method and an integral method, and completing volume measurement. According to the invention, the volume of the material pile can be rapidly and accurately obtained, automatic measurement of the materialvolume of the material pile is realized, the labor cost is reduced, and the accuracy is high.

The invention relates to a text detection method and device and a text information extraction method and system, and provides a method for detecting a text region in an image which comprises at least one connected component (CC). The method comprises a text background region (TBR) detection step for detecting at least one TBR from an image; a CC filtering step for filtering the at least one CC based on the detected TBR in order to keep at least one candidate text CC; and a CC combining step for combining the at least one candidate text CC to obtain the text region. According to the method disclosed by the invention, a text position in the image can be pointed out rapidly and/or accurately, so that the performance of text detection is improved.

The invention provides a macroblock layer bit allocation optimization method based on visual attention. The macroblock layer bit allocation optimization method based on visual attention comprises a step 1 of determining whether a macroblock belongs to a foreground region or a background region and detecting a motion region, a step 2 of detecting a structural texture region by use of a gradient, and a step 3 of allocating the target bits of the macroblock. The method is characterized in that based on the visual attention characteristics of an HVS (Human Visual System), firstly, the motion region attracting visual attention in an image is extracted by use of a motion vector of the macroblock in a reference frame in combination of an inter-frame difference method and the position of the current macroblock, and then the texture region is detected by use of an average gradient, and finally, the target bit allocation of the macroblock is optimized according to the region of the macroblock in the image. The algorithm involved in the macroblock layer bit allocation optimization method based on visual attention has the same coded image quality as a JVT-G012 algorithm; the fluctuation of a video sequence PSNR (Peak Signal to Noise Ratio) is reduced, the entire coded image is kept stable in objective quality, and meanwhile, excellent subjective quality of the coded image is also achieved.

The invention relates to the technical field of industrial surface defect detection, and provides a plate strip steel surface defect detection method based on a significance label information propagation model. The method comprises the following steps of firstly, acquiring a plate strip steel surface image I; then, extracting a bounding box from the image I, and executing a bounding box selectionstrategy; then, performing super-pixel segmentation on the image I, and extracting a feature vector from each super-pixel; then, constructing a significance label information propagation model, constructing a training set based on a multi-example learning framework to train a classification model based on a KISVM, classifying a test set by using the trained model to obtain a category label matrix,calculating a smooth constraint item and a high-level prior constraint item, and optimizing and solving a diffusion function; and finally, calculating a single-scale saliency map under multiple scales, and obtaining a final defect saliency map through multi-scale fusion. The surface defects of the strip steel can be efficiently, accurately and adaptively detected, a complete defect target can beuniformly highlighted, and a non-significant background area can be effectively inhibited.

A method to detect a given pattern in an image, as for obtaining, by a computer, transformation parameters that facilitate transformation of a template as to be overlapped with the object, comprising: inputting the template having contour points and region distinguishing information for each points; inputting contour candidate points and region distinguishing information for each points; transforming a position or a shape of the template by using each of transformation parameter set candidates; obtaining an evaluation value, or similarity of a distribution of the object and background regions, on each overlapping pair between one of the contour candidate points in the image and one of the contour points of the template; obtaining sum of the evaluation values only for those not smaller than a threshold evaluation value, for the each of the transformation parameter set candidates; and outputting a transformation parameter set having largest sum of the evaluation values.

The invention relates to a quick rectangle detection method of images with high resolution and high order of magnitude. The method comprises the following steps: firstly, pretreating each boundary pixel point set to obtain a cross point set taking a boundary pixel as a center; using a PPHT (progressive probabilistic Hough transform) algorithm and comparing with surrounding Hough parameters to obtain a straight line; and finally searching towards two sides of the straight line and deleting the whole cross point set to which the boundary pixels on the straight line belong. The process is circulated until a small quantity of pixels residues. On the basis of the extracted straight line, the established rectangle rule in which a pair of accurately matched parallel lines are detected firstly and then right-angle sides not completely matched are detected is utilized to detect the rectangles in the images. The method has good robustness for interferences caused by false boundaries and lens distortion due to dusts of a background area and is not limited by the different rectangle sizes; and thus a reliable rectangle detection method is applied for modernized industrial production of large-size rectangle boards.

The invention discloses a self-heuristic type strategy-based natural scene character detection method and system. According to the method and system, a two-level cascaded filtering mechanism is established to judge character areas and background areas; provided that each character area has a dark-background light-character mode or a light-background dark-character mode in a gray scale natural scene, the contrast types of the character areas are judged, so that the contrast of the character areas in the gray scale natural scene is unified to be the dark-background light-character mode; and results which are judged as characters after two-level cascaded filtering processing is performed are adopted as seed characters; self-heuristic type search strategies are constructed at regions near the seed characters for each seed character; a trained deep neural network is adopted to detect undetected charters through sliding a detection window so as to improve the recall rate of character detection; and adjacent characters are separated from each other with words adopted as units. The self-heuristic type strategy-based natural scene character detection method and system of the invention is of great practical value in the semantic automatic understanding of characters in the natural scene.

The invention relates to a weather identifying method based on dynamic scene perception. The weather identifying method comprises the following steps of: analyzing imaging difference of a dynamic sky scene object and background under different weather conditions, and determining positions of an object region and a background region by adopting a corner point detection algorithm; then respectively extracting texture and power spectrum slope characteristic of the background region and comparison and acutance characteristic of the object region; and fusing the background with the object region characteristics by utilizing a multicore learning method and identifying the weather conditions. The weather identifying method based on the dynamic scene perception has the advantages that influence of scale variation of an object on weather identification can be eliminated, useful characteristic information is increased by virtue of the object region characteristics for improving weather identification rate, and especially the weather identification rate with uniform background and similar characteristics is improved.