51 results about "Probabilistic hough transform" patented technology

A method and apparatus are disclosed for an image preprocessing device that automatically detects chestwall laterality; removes border artifacts; and segments breast tissue and pectoral muscle from digital mammograms. The algorithms in the preprocessing device utilize the computer cache, a vertical Sobel filter and a probabilistic Hough transform to detect curved edges. The preprocessing result, along with a pseudo-modality normalized image, can be used as input to a CAD (computer-aided detection) server or to a mammography image review workstation. In the case of workstation input, the preprocessing results improve the protocol for chestwall-to-chestwall image hanging, and support optimal image contrast display of each segmented region.

The invention discloses a detection method of a lane line. At first, edge information of a road image is extracted by means of a self-adaption threshold value edge extraction algorithm, cumulative probability Hough transformation in which an edge gradient direction constraint is brought is utilized so that straight lines can be quickly detected from an edge image, then the accurate positions of blanking points are estimated between successive frames by means of the detected straight lines through an iterative calculation method, a part of the straight lines not deriving from the edge of the lane line are filtered out by means of blanking point constraints, a tracking door is built so that noise can be further filtered out, candidate straight lines are obtained, and ultimately a pair of optimal straight lines are selected from the candidate straight lines to represent a current lane line through use of color information of candidate straight line neighborhoods and lane line information detected by a previous frame. According to the method, robustness on road environment change is achieved, the calculated quantity is small and real-time performance is achieved.

The invention discloses an expressway pavement detection method based on lane lines. The method includes the steps: continuously acquiring at least one video frame image of a video file and acquiringa target video frame image according to at least one video frame image; detecting the edge of the target video frame image to obtain an edge image comprising road edge pixel points; scanning the edgeimage to obtain a road region, horizontally dividing the road region to obtain sub-regions, detecting each sub-region by probability Hough transformation to obtain road edge line segments, solving vanishing points according to all the edge line segments in the sub-region at the top end of the road region, and determining an intermediate control point of each sub-region not at the bottom end of theroad region and a boundary point of the sub-region at the bottommost end of the road region according to the fact whether a straight line with a maximum slope and a straight line with a minimum slopein each sub-region have an intersection point or not; drawing left and right lane edge lines according to the intermediate control point, the boundary point and the vanishing points. By the method, adaptability to bend scenes is improved.

The invention provides a traffic monitoring image-based automatic detection system for a safety belt non-fastening behavior of a driver. The system comprises a module for extracting a to-be-detected region of the safety belt and a module for judging whether the driver fastens the safety belt or not; the module for extracting the to-be-detected region of the safety belt comprises a license plate detection sub-module, a driver region extraction sub-module, an image enhancement sub-module, a human face detection sub-module and a safety belt detection region extraction sub-module; the module for judging whether the driver fastens the safety belt or not adopts a straight line detection-based safety belt judgment method; and the straight line detection-based safety belt judgment method is used for detecting an edge straight line section of the safety belt by using a Canny edge detection algorithm and a progressive probabilistic Hough transform straight line detection algorithm. The safety belt non-fastening behavior of the driver can be accurately and quickly detected and recorded, so that the disadvantages of high labor cost of manual identification, easy fatigue, easy negligence and the like are effectively overcome, and the safety belt non-fastening driving behavior of the driver is effectively inhibited.

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 highway pavement detection method based on image processing, comprising the following steps: continuously collecting at least one video frame image of a video file, and obtaining a target video frame image according to the at least one video frame image; detecting the edge of the target video frame image to obtain an edge image containing road edge pixels; scanning the edge image to obtain a road region, transversely dividing the road region into sub-regions, and detecting each sub-region through probabilistic Hough transformation to obtain road edge line segments; calculating the vanishing point according to all the edge line segments in the sub-region on the top end of the road region, and determining the middle control points of each non-bottom sub-region and the boundary points of the bottommost sub-region according to whether there is an intersection point between the line with the maximum slope and the line with the minimum slope in each sub-region; anddrawing left and right lane edge lines according to the middle control points, the boundary points and the vanishing point. By applying the embodiment of the invention, the adaptability to a bend scene is improved.

The present invention relates to a simple and rapid lane line detection method based on the dynamic area of interest. The method comprises: performing preprocessing of an original image to obtain an edge detection result graph; calculating a dynamic area of interest; obtaining the lane line detection result of the previous frame image, and generating a mask graph according to the lane line detection result of the previous frame image; employing the mask graph and the edge detection result graph for operation, and obtaining the local image only having the dynamic area of interest; using the Progressive Probabilistic Hough Transform to detect the line in the local image; traversing all the lines, calculating the length thereof, and if the length is smaller than a threshold value, skipping the line; or else, calculating the distance from the line to the middle point of the image bottom edge; traversing the line set after sorting, and calculating the line slope and color and the deviation degree with the previous frame lane line position; and if the deviation degree is located in each range, taking the line as this frame lane line; and or else, going on searching backwards. The simple and rapid lane line detection method based on the dynamic area of interest can better satisfy the requirement of the vision navigation system on the universality, the timeliness and the robustness.

The present invention discloses a pointer type instrument panel automatic reading method under a long-distance and multi-view environment. The method comprises the steps of dial plate area determination, segmentation processing, pointer extraction fitting and reading estimation, and is characterized by determining a dial plate area in an image by matching the key points of a to-be-identified image and a template dial plate, then distributing a binaryzation dial plate according to a histogram, obtaining the internal information of the dial plate by the binary image morphologic processing, and then inducing and merging the straight lines obtained by the progressive probabilistic Hough transform fitting to obtain a pointer angle, and finally calculating the pointer reading according to a transcendental pointer. The pointer type instrument panel automatic reading method of the present invention solves the problems that the distortion of the dial plate is severe under a long-distance strabismus angle, the scales are fuzzy, and the reading can not be identified completely by the eyes, and has the higher accuracy.

A laser etching industrial detonator code image positioning and correction method is provided. Firstly, the edge information of a preprocessed detonator coded image is detected and binarized by a Sobel operator, and then the rough position of the detonator encoded character is located by the morphological closed operation, then, the binary image after rough positioning is fitted with line segmentsby probabilistic Hough transform, and the tilt angle of detonator coding is estimated, and then the coded image is corrected by rotational transform according to the tilt angle. Finally, the positionof coded characters is located accurately according to the distribution of the fitting line segments in the corrected image, that is to say, precise positioning is realized. This method can solve thetechnical problems of inaccurate positioning of detonator encoding caused by inclination of encoded characters, variation of illumination, random noise and so on, and has high reliability and stability.

The training method and detection method of electric wire based on depth separable convolution neural network comprises that steps of: constructing a neural network using the deeply separable convolution; trainingconvolution kernels with uniformly distributed mini-slices, wherein the trained convolution kernel is used for feature extraction of infrared gray-scale image; binarizing the image according to the threshold value, removing the small area region, and connecting the linear region by the probabilistic hough transform. The invention trains a convolution kernel capable of extracting infrared wire features by a machine learning method, and can effectively extract wire features from infrared gray-scale images. Combined with morphological processing and linear Hough transform, wire detection can be carried out in real time.

The invention relates to a weldment contour detection method. The weldment contour detection method comprises the following steps: 1, image enhancement preprocessing, 2, image histogram equalization processing, 3, image denoising, 4, image background interference removal and 5, image edge improvement. According to the invention, the weldment image is preprocessed in an image enhancement mode, so that the edge detection precision of the weldment image is improved; According to the invention, histogram equalization is introduced to improve the gray level dynamic range of the image, the image contrast is improved, and the image is clearer; According to the invention, the median filtering technology is adopted to carry out denoising processing on the image, operation is simple, implementation is convenient, and edge information of the image is protected while isolated noise points are filtered out; The invention adopts the probabilistic Hough transform method to judge and connect the pixel points disconnected from the edge of the image, and is small in calculation amount, high in practicability and high in efficiency.

The invention discloses a vision-based cutting width measurement method of an intelligent rice and wheat harvester, which comprises the following steps: according to an internal and external parametermatrix obtained by camera calibration, the world coordinate system is matched with a pixel plane coordinate system of a camera image, and an original rice and wheat image obtained by the camera is converted into a top view by using an inverse perspective projection; obtaining the actual target distance from the harvesting boundary line to the separator, obtaining the pixel width of the target area in the top view and determining the target area image by combining the reduction factor; Image preprocessing operation; the harvesting boundary line between harvested rice and wheat and unharvestedrice and wheat was extracted based on morphology, and the binary image of the target region containing the harvesting boundary line was obtained. Using probabilistic HOUGH transform method to detect the harvesting boundary line in the image and obtain coordinate parameters, the pixel distance from the harvester to the harvesting boundary line in the image can be obtained, and the actual cutting amplitude of the harvester can be calculated. The invention improves the reliability and can safely measure the cutting width of the intelligent rice and wheat harvester with high precision in real time.

The invention discloses a variable angle multi-meter simultaneous identification method for patrol robot, which comprises the following steps: obtaining multi-meter image or video data; each frame ora single image of the video data is directly preprocessed; edge detection is performed on the preprocessed image data. An improved Freeman coding method is used to trace the edges. Calculate the chaincode and select the edge of dial of each instrument. Using the selected edge and chain code information of the dial, the dial is positioned and processed by emmetrization. All the obtained dial images are respectively stretched in a circle contrast ratio; Fast Probabilistic Hough Transform for Pointer Recognition in Embodied Dial Image; in the emmetropic dial image, the edge is extracted again, and the radius reduction method is used to calculate and analyze the benchmark scale after the interference of lower sector is filtered out. Finally, the deflection angle is calculated by using the identified pointer and reference scale, and the final reading is obtained by combining the range.

The invention belongs to the field of image processing and discloses a rapid lane line detection method based on morphological transform and adaptive threshold. The problem that the conventional lane line detection algorithm is low in reliability and low in speed is solved. The method comprises the following steps: taking a part, which comprises lane lines, of the lower part of an image as a sensitive area; transforming the colored sensitive area into a gray sensitive area; calculating the threshold based on an adaptive threshold segmentation method of a Gauss distribution model, and performing threshold segmentation on a grayscale image so as to obtain a binary image; performing morphological transform, performing expansive operations on the binary image, performing corrosion operation, and finally making difference between the two images so as to obtain edge information of the images; finally, fitting the lane lines by adopting improved probabilistic Hough transform. The lane line detection method disclosed by the invention is high in reliability and speed, so that the detection time is greatly reduced to 0.016 second, which has a great significance on real-time application.

The invention relates to a method and a system for extracting parking space of underground garage in high-precision map making. The method includes: S1, projecting 3D laser point cloud containing parking space to 2D bird's-eye mode image, projecting 3D lase point cloud containing parking space to 2D bird's-eye mode image, projecting 3D laser point cloud to 2D bird's-eye mode image, projecting 3D laser point cloud containing parking space to 2D bird's-eye mode image, projecting 3D laser point cloud to 2D bird's-eye mode image. S2, calculating an image contrast estimation index, and obtaining abinary image by using different image preprocessing methods according to the contrast estimation; 3, detect a straight line segment of a binary image by using a probabilistic Hough transform, and calculate a rotation angle estimation of a parking line according to that detected straight line segment; 4, rotate that image according to the rotation angle and take the center point of the image as thecenter of the circle; 5, count that number of pixel of each row and column of the rotated image including the parking line to obtain the integral projection of the image in the horizontal and vertical method; S6, searching four inner corner coordinates of the parking space according to the integral projection of the image in the horizontal and vertical directions; 7, that coordinate of the cornerpoint in the parking space is inverse transformed to the point cloud data, thereby extracting the parking space.

The embodiment of the invention provides a text image correction method and device and electronic equipment. The method comprises the following steps: carrying out binarization processing on a to-be-corrected text image, and carrying out fuzzification processing on the binarized text image by utilizing a box type filter so as to filter character gaps between adjacent characters in the same row inthe binarized text image and reserve row gaps between different rows; detecting and connecting line edges of each line of characters in the fuzzified text image by adopting a multi-level edge detection algorithm, and obtaining line segments formed by each line of characters in the fuzzified text image through statistical probability Hough transform based on the line edges; obtaining the maximum length of all the line segments, and calculating the inclination angle of the to-be-corrected text image by calculating the slope of the line segment; and based on the inclination angle, rotating the to-be-corrected text image to realize correction. According to the embodiment of the invention, the text image can be effectively and accurately corrected under the condition that the text does not havea frame.

The invention discloses a vehicle yaw angle determination method based on Hough transform. The method comprises converting an image ahead of a vehicle into an 8-bit single-channel image, performing binaryzation by using a global adaptive threshold obtained by OTSU, and obtaining a straight line by using progressive probabilistic Hough transform; obtaining the length and the slope arc tangent value of a lane line to determine whether the lane line is a solid line or a dashed line; obtaining the distance of the solid line, and taking slide line with the smallest distance as a real lane solid line; obtaining a difference value between the dashed line slope arc tangent value and weighted average value thereof, and determining the dashed line with the least difference value as a real lane dashed line; finally, calculating, according to detected left and right lane lines, an intersection angle between the angular bisector of an intersection angle of the left and right lane lines and the axis Y of a coordinate system as the vehicle yaw angle. The vehicle yaw angle determination method is low in calculated amount, high in recognition accuracy and fast in recognition speed, and can quickly calculate the yaw angle of the vehicle, and can filter out various noises on the ground well.

The invention discloses a weld beading size measuring method for a two-dimensional image of a small-diameter tube weld joint root, which comprises the steps of extracting an image weld joint target area according to an actual weld joint position, and establishing an image pixel coordinate system in the target area; extracting a vertical line of the small-diameter pipe wall by utilizing self-adaptive thresholding operation; probabilistic Hough transform is used for detecting and obtaining the position of a vertical line of the inner pipe wall, and an inner welding wall area is segmented; enhancing and highlighting weld beading characteristics by utilizing inverse color, Gamma, morphological operation and Butterworth filtering; segmenting a weld beading from the maximum entropy threshold value; approximately fitting an ellipse of the inner weld joint wall according to the empirical value of the inner weld joint wall; and further fitting weld beading in the ellipse of the inner weld wall,and calculating the size of the weld beading. The reliability is improved, and the weld beading size of the two-dimensional image of the root of the small-diameter pipe weld joint can be accurately and effectively measured.

The invention discloses a power transmission line detection method under a complex background based on image processing. The method comprises the following steps of 1, obtaining a power transmission line image through aerial photography of a camera or a camera carried on an unmanned aerial vehicle; step 2, carrying out edge detection on the power transmission line image by adopting a Canny operator; step 3, removing small connected regions by adopting a region growing method; 4, carrying out linear detection by adopting probability Hough transform; and 5, screening the straight line segments,and removing pseudo straight line segments. The method is high in anti-interference capability, good in stability, good in detection effect and high in robustness, and real power transmission line information can be detected under a complex background.

The invention discloses a pointer type instrument automatic reading method based on scale search, which comprises the following steps: preprocessing an instrument image to obtain an instrument regionbinary image, performing probabilistic Hough transform on the instrument region binary image, and identifying an instrument pointer; performing instrument scale search by utilizing the virtual searchbox; carrying out connected domain detection on the binary image of the instrument area to determine a candidate digital area and identifying scale values in the candidate digital area a based on a convolutional neural network; respectively calculating the distance between the center of each scale value and the starting point of all the scale marks, distributing each scale value to the scale withthe shortest distance, and determining the scale value of the scale without the marked numerical value according to the relation of the scale marks; and determining the instrument reading according tothe distance relationship between the pointer and the two scales closest to the pointer. According to the invention, scale value identification and scale value calculation algorithms are provided, sothat prior information in a database is not needed.

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 discloses an outdoor facility inspection method. According to the outdoor facility inspection method, a color image is read, and the image is converted into a grey-scale image; line segments in the image are calculated by employing probability Hof transform, and the line segments with length exceeding a set threshold are selected; the K near neighbor clustering algorithm is utilized, and the clustering result is traversed till clustered line segments can form a profile structure of equipment; a detection-needing area is selected; and a localized binary image in the area is calculated; a white connected component of the binary image is simplified through a binary image refinement algorithm to form a curve line segment; the binary image containing the curve line segments is traversed, points at end points of the curve line segments are extracted, and the points are marked at end point queues; the end point queues are classified and processed by utilizing a gradient histogram and a support vector machine, and an inspection result is acquired. The method can adapt to a real illumination situation and further has advantages of high detection accuracy, faster processing speed, high stability and relatively high noise resistance performance and inclination resistance performance.

The invention relates to a method for automatically identifying the position of a guide pin in an X-ray perspective image. The method includes acquiring an image contour of the X-ray perspective image, detecting line segments in the image contour through a probability Hough transform algorithm, subsequently identifying a pair of line segments, which are parallel or approximately parallel, at the edges of two sides of the guide pin or a sleeve placed in a bone surface among the detected line segments, and drawing an extension line of the sleeve or the guide pin in the X-ray perspective image according to a center line of the pair of line segments so as to achieve automatic identification of the position of the guide pin in the X-ray perspective image. The method for automatically identifying the position of the guide pin in the X-ray perspective image is capable of obviously reducing exposure time of X-rays, remarkably shortening operation time, reducing secondary damages to patients, improving positioning accuracy of the operation guide pin, and improving operation efficiency.

The invention discloses a circular pointer type meter reading method of a switch cabinet detection device. The prior knowledge of the instrument to be identified is fully utilized to determine the optimal image processing parameters and improve the recognition accuracy of the device. According to the positioning function of the switch cabinet detection device, it is not necessary to stop at the designated position to collect images, so as to improve the detection efficiency of the device. After the circle boundary is recognized, the effective area is determined, and the elements that affect the recognition pointer, such as meter frame and dial, are deleted, so as to improve the precision of pointer recognition. The effective region is expanded to avoid too few pixels at the pointer boundary, which makes the probabilistic Hough transform unable to recognize the pointer boundary and improves the accuracy of pointer recognition.

The invention discloses a monocular vision-based gravitational field method accelerometer calibration method, which mainly comprises the following steps of: proposing a plumb bob method to determine the direction of a gravitational field, and then using a plumb line as an initial reference of a rotation angle of a turntable; using a monocular vision system to shoot a rotary table image, and usinga template matching method to determine a rectangular target feature area in the rotary table image; realizing sub-pixel extraction of the characteristic edge of the rectangular target based on a Zernike moment method; extracting a long edge of the rectangular target by using a cumulative probability Hough Transform (PPHT) algorithm, wherein an included angle between the edge and a plumb line is the rotation angle of the rotary table at the current position; the sensitivity of the accelerometer is calculated by using the measured rotation angle, and the gravitational field method accelerometercalibration at any rotation angle can be realized. According to the method, the defects of limited calibration precision and installation error in multi-axis acceleration timing calibration in the traditional gravitational field method accelerometer calibration are overcome.

The invention provides a diamond saw wire fracture prediction method based on image processing, and the method specifically comprises the steps: reading a photographing content, and reading a to-be-fractured and fractured image of a saw wire; performing gray processing on the picture; carrying out edge detection by adopting a Canny operator, processing the grayscale image, and extracting the edgeof the grayscale image; extracting local features of the image by adopting statistical probability Hough transform to obtain a straight line segment in the image; extracting polar coordinates of segment selection endpoints; establishing a vector a of a reference coordinate line segment, and performing included angle calculation on the segment selection vector b obtained in the previous step and the reference vector a; determining a limit angle record alpha before the saw wire is broken; detecting the machining states at different moments, and measuring the saw wire deflection angle cut by thesaw wire at any time and recorded as [beta]; and comparing [beta] with alpha, wherein if [beta] is infinitely close to [alpha], feeding or machining of the workpiece is stopped. The method has the beneficial effects that compared with a shot picture drawing angle measurement method, the time cost is saved, and the efficiency is higher.

The invention relates to a method and system for extracting parking spaces of underground garages in high-precision map making, comprising the following steps: S1, projecting a 3D laser point cloud including parking spaces into a 2D bird's-eye view mode image; S2, calculating an image contrast estimation index, According to the contrast estimation, different image preprocessing methods are used to obtain a binary image; S3, use the probabilistic Hough transform to detect the straight line segment of the binary image, and calculate the rotation angle of the parking line according to the detected straight line segment; S4, according to the rotation angle and the image The center point is the rotation image of the center of the circle; S5, count the number of pixels of the parking line in each row and column of the rotated image, and obtain the integral projection of the image in the horizontal and vertical directions; S6, search according to the integral projection of the image in the horizontal and vertical directions to obtain parking The coordinates of the four inner corners of the parking space are obtained; S7, the coordinates of the inner corners of the parking space are inversely transformed to the point cloud data, thereby extracting the parking space.

The present invention relates to a blue license plate segmentation and correction method, comprising the following steps: 1. For the rough positioning image of the license plate, the R channel map of the RGB channel and the V channel map of the HSV color space are used to construct a combined image, and then the highlighted blue license plate is obtained. The blue area binary image of the colored area, and then fuse it with the white area binary image highlighting the white character area of ​​the license plate to obtain a dual-area fusion binary image; 2. Perform edge detection on the dual-area fusion binary image, here On the basis of obtaining the binary image of the circumscribed contour of the license plate area, then the probabilistic Hough transform is performed on the binary image of the circumscribed contour of the license plate area to fit the contour line segment and further locate the corner points of the license plate; 3. After obtaining the corner points of the license plate, the original The license plate image is subjected to perspective transformation to obtain the rectified license plate image. This method can quickly and accurately correct the license plate under complex conditions and improve the recognition rate of the license plate.

The invention relates to a road edge detection method based on heuristic probability Hough transform. As an adaptive threshold search method, the Otsu method is robust to unstructured road region segmentation in complex environments. However, the Otsu method itself can only obtain rough segmentation results of road areas and non-road areas, but cannot obtain precise road boundaries. The method of the present invention comprises the following steps: in the double-threshold detection stage, the edge pixels connected to each other are uniformly numbered, thereby forming several subsets of the Canny edge; first, the least squares method is used for each edge point subset to carry out straight line fitting to obtain The axis of this point set. The invention is used for road edge detection.

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.