42 results about "Hough space" patented technology

The invention discloses a visual detection method of a road driving line. The invention comprises the following steps: firstly, detecting all of edges of a given original road image through edge detection, and obtaining an edge image; secondly, respectively defining search rules based on direction priority for a left lane line and a right lane line, using the search rules for searching the edge image, and recording each search path in the search so as to transform the edge image to a set of candidate line sections; thirdly, carrying out Hough transformation on each line section of the set of the candidate line sections, counting collinear points of the line sections in the Hough space, and filtering the line sections with the collinear points less then T2; and fourthly, utilizing luminance information of the original image for carrying on a series of transformation processing on the candidate line sections remained after processing in the third step, eliminating cracks of the road surface and regular edge interference of join of the new and old road surfaces, identifying the lane lines, and completing the positioning of the lane lines. The scheme of the invention follows an idea from coarseness to fineness, and has accurate positioning and high real-time performance when being compared with the prior scheme.

A method for depth matching borehole images and / or core section images is disclosed wherein signals from sensors at different levels on a logging tool are converted into an averaged signal representing the average bed signal at the center of the borehole at each of the different levels. A depth matching technique is applied to the averaged signals from the sensors at different levels on the logging tool to determine the optimum depth offset necessary for matching two sets of signals from sensors at the different levels of the logging tool. In an alternative embodiment of the invention a Hough transform is utilized to process the well log images and generate three-dimensional images in Hough space. The three dimensional images are converted into two-dimensional extremum curves. Depth matching is performed on the two dimensional extremum curves to calculate an offset to match the two dimensional extremum curves. The calculated offset is then applied to the well log images to depth match them.

Voting of coordinates of each pixel in a photographic image onto a Hough space corresponding to annuluses is performed, the Hough space being defined by a coordinate system having an X coordinate axis for a circle center point, a Y coordinate axis for a circle center point, and an r coordinate axis for a radial direction. A vote value of each voting position having coordinates (X, Y, r) on the Hough space is acquired as a first integrated vote value from a calculation of a number of votes given to each voting position. A position represented by (X, Y) coordinate values of a voting position associated with the largest first integrated vote value is detected as the position of the center point of the circular pattern.

A Hough transform unit executes Hough transform to HIGH pixels of outline binary image data inputted thereto, and stores the calculation result in a Hough space data storage. A Hough space data calculating / projecting unit sequentially reads out data stored in the Hough space data storage, executes a specific calculation, and thereafter stores the calculation result sequentially in a calculated projection data storage. An angle detector sequentially reads out calculated frequency data stored in the calculated projection data storage, calculates the maximal value of the data read out, and detects an angle that gives the maximal value as the skew angle. The image processing device, being thus configured, allows detecting and correcting the skew angle with high accuracy, even when the input image contains image elements such as photograph images and dot images.

The invention discloses a spliced banknote detection method. The method specifically comprises the steps of acquiring the image of a banknote, rotating and correcting the image of the banknote so as to further recognize the basic information; based on the obtained basic information, positioning a to-be-detected region; figuring out the gradient graph of the positioned region through the sub-sampling process, and measuring a step signal at the spliced part of the banknote based on the sum of first norms at various differential dimensions; selecting an image-binary threshold for the gradient graph and conducting the binarization treatment through selecting a hard threshold; converting the Hough transformation on the binarized image of the gradient graph in the form of polar coordinates; conducting the straight line detection: in the Hough space, searching out all maximum values, wherein the number of maximum values represents the number of linear abnormal signals in the original image space and the values of the maximum values represent the lengths of the linear abnormal signals in the original image space. According to the technical scheme of the invention, various spliced banknotes can be effectively detected, and the anti-counterfeiting means of a money detector is improved. Meanwhile, the detection efficiency of a bill sorting machine is improved. The method can be applied to the detection of various weak straight line signals.

A vehicle travel support device determines presence of a recognition inhibiting factor of a lane mark on a road on which a vehicle is traveling with high accuracy irrespective of an imaging history by a vehicular camera from the same position. The vehicle travel support system generates an edge image by extracting an edge or actualizing an edge in an image obtained through the vehicular camera. When Hough transform of the edge image is performed, votes for a specified vote value of a linear component is evaluated in a ρ-θ space (Hough space). Presence of a recognition inhibiting factor of a lane mark on a road is determined by determining whether or not the votes of a specified vote value in a specified region denoting a standard travel lane of a vehicle in the real space is ≧a threshold in the ρ-θ space.

A recognition system of this invention has feature point detection means (120), Hough transform means (130), and specific pattern output means (140). In the Hough transform means (130), a Hough space is designed so that a magnitude relation of a distance between points in the Hough space is equivalent to a predetermined magnitude relation of an inter-specific-pattern distance indicative of a difference between specific patterns. The recognition system detects the specific patterns using the Hough space. By adopting such a structure to express more similar specific patterns in an image as closer points also in the Hough space, it is possible to achieve an object of this invention.

The present invention discloses a correcting method for the inclination of a vehicle license based on a supporting vector machine. The implementing steps of the method include an images collection, a gray processing, a binary processing, an inclination correction of the vehicle license (comprising that a matrix of an image datum collection is founded, the image matrix is centralized, the matrix is calculated, the inclination vector is calculated and an image is obtained after the inclination correction via transforming the coordinate), an expanding correction of the image and an erosion correction of the image. The correcting method for the inclination of a vehicle license of the present invention supports that a new machine learning tool (the vector machine) is applied. The traditional optimal-requesting problem of a Hough space of a vehicle license image is converted to a solution-requesting problem for a linear matrix. The calculation is simplified and the arithmetic efficiency is improved. The randomicity and instability existing in the searching process of the optimal points of parameters such as the obliquity are avoided. For the images with indistinct frame or noise, a better effect can still be obtained.

The invention belongs to the field of intelligent transportation, in particular to a vanishing point detection method based on cascaded Hough transform, comprising the following steps of step 1 collecting road vehicle video to obtain vehicle target of each frame image; 2 extracting Harris corner point to obtain feature points on vehicle targets of each frame image; 3 obtaining a straight line track of that vehicle; 4 screening that straight track of the vehicle, and recording the set of the selected straight track of the vehicle as L; 5 converting that selected straight line track set L from image space to rhombus Hough space by cascade Hough transform to vote to obtain coordinates of maximum value points aft voting; 6 converting that coordinates of the maximum point into an image space, finally obtaining the coordinates of the vanishing point in the image space, and finishing the detection of the vanishing points. The method is suitable for various weather conditions, thereby avoidingthe false detection of vanishing point in special weather, and greatly improving the accuracy of vanishing point detection.

The invention discloses a power transmission line tracking method and system. The method comprises the steps of: utilizing a ratio operator to carrying out edge detection on an image, and obtaining an edge binary image; carrying out housh conversion on the binary image, outputting a straight line of a power transmission line, and storing hough space parameters of the currently output straight line of the power transmission line; combined with the hough space parameters of a straight line of the power transmission line, applying particle filtering for successively calculating hough space parameters of each subsequent video frame, carrying out line determination on the straight line of the power transmission line by utilizing a Bayes classifier, and obtaining a determination result indicating whether the straight line of the power transmission line is a right power transmission line. According to the invention, collected continuous video frames are used for carrying out power transmission line tracking, the detection efficiency and accuracy of the power transmission line are improved, and the Bayes classifier is utilized for carrying out line determination on the power transmission line obtained by tracking, so that the detection accuracy of the power transmission line is further improved.

The invention discloses a cross target detection method and system and a storage medium. The detection method of the cross target comprises the following steps: firstly, carrying out convolution and other processing on a grayscale image containing the cross target to obtain an edge image; mapping all pixel points in the edge image into a Hough space; obtaining the angle and position information ofall the line segments in the edge image according to the Hough transform matrix and the edge image, and finally determining the center position of the cross target according to the angle and positioninformation of all the line segments in the edge image. A template does not need to be calibrated in advance, and the applicability is good; and since the angle and position information of the line segment is acquired in the Hough space, the method is not affected by image rotation and zooming, the purpose of correct identification can be achieved even if the cross target in the grayscale image is partially shielded, and the robustness of the detection method of the cross target is improved.

The invention provides a lane line detection method. The method comprises the following steps of firstly setting a possible lane region at the lower part of a lane line image as an ROI; performing gray processing and binary processing on the ROI and obtaining an edge image by adopting a Canny operator; performing Hough transform on the edge image, and converting an edge line segment to a Hough space from an image space; performing screening check on Hough points in sequence by using parallel constraints, length constraints, pairwise constraints and equal-width constraints in sequence; and searching for an optimal lane combination in a probability matrix by adopting dynamic planning. The method has the beneficial effects that for the defects of complex process and relatively low detection speed of an existing Hough transform-based lane line detection method, a lane line equation is directly screened out by using the length constraints, the parallel constraints, the pairwise constraintsand equal-interval constraints in the Hough space according to road design specifications, so that the process is simplified and the detection speed is increased.

The invention discloses a lane line fast detection method based on Hough transform. The method comprises the following steps: the image preprocessing step and the lane line extraction step. In the image preprocessing step, at first, image noise is eliminated through improved median filtering; and then, edge detection is performed through 45-angle and 135-angle Sobel operators; and finally, the lane line is segmented through the OTSU method. In the lane line extraction method, the lane line angle range is detected at the first frame to reduce the Hough space search range; in the subsequent frame detection, search is performed near the previous frame of frame lane line image space and the parameter space at first, and if no lane line is detected or the parameter difference is large, then the search range is expanded. According to the method, the lane line recognition speed can be quickened, and the system anti-interference capability can be improved.

The invention relates to a polling map graph part target identification method based on generalized Hough transformation, and the method comprises the steps: A, carrying out the selection of referencepoints in the inspection tour, and carrying out the positioning of each reference point after the selection; B, after positioning the reference points, establishing R-table matrix according to the number of the reference points and the distance; Step C: according to the R-table matrix, establishing a Hough space; D, selecting a maximum peak point as an optimal reference point in the Hough space to form a reference image; And step E, searching reference coordinates in the inspection tour map by using a generalized Hough transformation method, extracting an image from the inspection tour map according to the reference coordinates, and performing mutual information calculation by using the extracted image and the reference image so as to judge whether the part is lost or not. According to the invention, the generalized Hough transformation is used to identify the parts in the inspection tour map, and whether the parts in the inspection tour map are missed or not can be quickly and accurately judged, so that real-time detection of the unmanned aerial vehicle on the equipment in the inspection tour process is realized.

The invention discloses an object identification method based on shape fragments. The method comprises the steps of extracting edge shape fragments of a training image and edge shape fragments of a test image; constructing an edge shape fragment model of an object according to the edge shape fragments of the training image; selecting specific edge shape fragments from the edge shape fragments of the test image as candidate edge shape fragments, wherein the similarity between the specific edge shape fragments and an edge shape fragment in the edge shape fragment model of the object must be larger than a first threshold; calculating the maximum probability value of Hough space by means of Hough transformation to determine the reference point position of an object to be detected; screening the candidate edge shape fragments according to the reference point position of the object to be detected to obtain an actual object profile fragment. The object identification method based on the shape fragments effectively solves the problem that a traditional method can not be used for identifying flexible objects or objects with unstable appearance features.

The invention provides a method for quickly detecting a human head, and the method comprises the steps: calculating a resolution image and the minimum and maximum radius images of a human head according to a calibration line; calculating edge point images and edge gradient vector images of the images; obtaining a Hough space image according to Hough voting; extracting candidate points of the human head; calculating the radius and confidence of candidate human heads, and obtaining candidate human head regions; carrying out the candidate human head regions; enabling the residual candidate human head regions to serve as human head detection regions, and outputting the residual candidate human head regions. The method achieves the quick detection of the human head, and can be used in pedestrian counting.

The present invention comprises using error propagation for building feature spaces with variable uncertainty and using variable-bandwidth mean shift for the analysis of such spaces, to provide peak detection and space partitioning. The invention applies these techniques to construct and analyze Hough spaces for line and geometrical shape detection, as well as to detect objects that are represented by peaks in the Hough space. This invention can be further used for background modeling by taking into account the uncertainty of the transformed image color and uncertainty of the motion flow. Furthermore, the invention can be used to segment video data in invariant spaces, by propagating the uncertainty from the original space and using the variable-bandwidth mean shift to detect peaks. The invention can be used in a variety of applications such as medical, surveillance, monitoring, automotive, augmented reality, and inspection.

The invention discloses a microseismic data assisted tight oil and gas reservoir fracture distribution inversion method. According to the method, Hough transform is adopted to transform microseismic data into Hough space, so that a Hough space function can be obtained, and an initial mean value of a Hough space random field can be formed; the sampling of crack distribution in a physical space is generated by adopting Hough inverse transformation; production prediction data corresponding to sampling of crack distribution in a physical space is predicted through numerical simulation; a correlation matrix of the Hough space random field and the production prediction data is obtained by adopting an algebraic statistical method; the posterior probability of the Hough space random field is maximized by using the correlation matrix and adopting a sample set-based iterative algorithm to form a crack distribution inversion method. According to the method, the precision of tight oil and gas reservoir fracture distribution inversion and development well production prediction can be improved.

The invention provides a lane line detection method, and the method comprises the following steps: firstly setting a lower part of a lane line image and a possible lane region as an ROI; carrying outthe graying and binarization of the ROI, and then obtaining an edge image through a Canny operator; carrying out the Hough transformation of the edge image, and converting an edge line segment into aHough space from an image space; sequentially carrying out the screening and inspection of a Hough point through parallel constraint, length constraint, pairing constraint and width constraint; and searching an optimal lane combination in a probability matrix through dynamic planning. The beneficial effects of the invention are that the method irons out the defects that a conventional lane line detection method based on Hough transformation is tedious in flow and is slower in detection speed; the method achieves the direct scanning of a lane line equation through the length constraint, parallel constraint, pairing constraint and equal-width constraint in the Hough space according to a road design standard, thereby saving the process, and improving the detection speed.

The invention provides a density-based geometrical calibration method in image matching. The method comprises the steps of: (1) generating a candidate feature matching pair set of two images; (2) estimating a probability density of each matching pair in the candidate feature matching pair set in a Hough space, and taking the probability density of each matching pair as a weighting factor of a matching score of the matching pair; and (3) accumulating the matching scores of all feature matching pairs in a feature matching pair set to obtain the matching score between the images, as a similarity between the two images. According to the method, a large weight is endowed to the matching pair located at an area with a high density in the Hough space, a small weight is endowed to the matching pair located at the area with a low density in the Hough space so as to reflect the correct probability of the matching pairs, the multi-model matching can be processed, and great flexibility is brought while maintaining the advantage of high efficiency of a Hough voting method.

The present invention comprises using error propagation for building feature spaces with variable uncertainty and using variable-bandwidth mean shift for the analysis of such spaces, to provide peak detection and space partitioning. The invention applies these techniques to construct and analyze Hough spaces for line and geometrical shape detection, as well as to detect objects that are represented by peaks in the Hough space. This invention can be further used for background modeling by taking into account the uncertainty of the transformed image color and uncertainty of the motion flow. Furthermore, the invention can be used to segment video data in invariant spaces, by propagating the uncertainty from the original space and using the variable-bandwidth mean shift to detect peaks. The invention can be used in a variety of applications such as medical, surveillance, monitoring, automotive, augmented reality, and inspection.

The invention relates to the technical field of network security, in particular to a network security encryption method and system based on artificial intelligence. The method comprises steps of converting data to be encrypted to a binary image; obtaining a voting value sequence corresponding to each pixel point in the Hough space, determining a first importance degree of each pixel point by comparing the voting value sequence with a voting threshold value, and screening out base point data; obtaining the second importance degree of each angle according to the difference between the voting value information under each angle in the Hough space and the reference angle, screening out perspiration important angles, and obtaining space angle data according to the voting value mean square error in the Hough space corresponding to the important angles; the number of pixel points in each column in the binary image forming initial phase data, and the base point data, the space angle data and the initial phase data serving as encrypted data. According to the invention, the to-be-encrypted data is converted into the binary image, and the image is subjected to data splitting encryption, so that the security in the transmission process is ensured.

The invention discloses an improved probability Hough transform curve detection method based on partitioning minimum samples fixing. The method comprises the steps that the edge or a framework of an image is extracted, so that a binary feature image is obtained; according to a given partitioning size, the feature image is partitioned; a feature point set of each partition is obtained; according to a given random sampling rate, the partitioning minimum sampling number, the partition area and the size of each feature point set, the actual sampling number of each partition is obtained; through unrepeated random sampling, feature points with the number equal to the actual sampling number are extracted from the feature point set of each partition; a union set of feature point sample sets extracted from the partitions serves as a total sample set; standard Hough transform is conducted on the total sample set, so that curve detection is finished. The method can combine the advantage that standard Hough transform is simple and easy to achieve and the advantage that probability Hough transform is high in speed, false curves caused by complex texture regions in the image can be restrained, and the detection rate of a highest peak value point set in Hough space is improved.

The invention relates to the technical field of fabric defect detection, in particular to a fire hose defect detection method and system based on computer vision. The method determines the abnormal region through the optical flow information. And converting the texture edge in the abnormal region into a Hough space to obtain the form information of the texture edge. Clustering is carried out according to the Hough curve form in the Hough space, and a first cluster is obtained; and obtaining a first defect area according to the number of the abnormal samples in the first cluster. And determining a second defect area according to the abnormal distance between the parallel texture edges corresponding to the first cluster. According to the method, the morphological information of the texture edges is obtained through Hough space conversion, and defect detection is accurately carried out according to the state between the edges.

Voting of coordinates of each pixel in a photographic image onto a Hough space corresponding to annuluses is performed, the Hough space being defined by a coordinate system having an X coordinate axis for a circle center point, a Y coordinate axis for a circle center point, and an r coordinate axis for a radial direction. A vote value of each voting position having coordinates (X, Y, r) on the Hough space is acquired as a first integrated vote value from a calculation of a number of votes given to each voting position. A position represented by (X, Y) coordinate values of a voting position associated with the largest first integrated vote value is detected as the position of the center point of the circular pattern.

An inshore vessel detection method based on contour refinement and improved generalized Hough transform relates to an optical remote sensing image inshore vessel detection method. The method of the invention solves the problem of low ship detection accuracy when the training sample is insufficient and the bow contour is damaged in the detection. The method comprises the steps of obtaining the length and width information of the ship to be measured and the binary edge detection result image of the ship; acquiring an equal-frequency quantized Hough space voting gradient angle reference table, utilizing remote sensing images of scenes to be detected to acquire a plurality of suspected parking areas of ships near the coastline; voting each point of the edge contour of the suspected parking area in Hough space, wherein the point whose voting cumulative value is greater than the threshold value A is the center reference point of the bow in the suspected parking area; using a minimum rectangular frame to mark a suspected ship in a remote sensing image of a scene to be detected; using the minimum rectangular frame to mark a suspected ship in a remote sensing image of the scene to be detected; carring ou the removed false alarms and fusion overlap processing on all suspected vessels to complete the detection of inshore vessels. The method of the invention is applicable to the detectionand use of offshore ships.