34 results about "Connected-component labeling" patented technology

The present invention discloses a vehicle tracking system and method, and the tracking method comprises the steps of capturing a bright object from an image by the bright object segmentation; labeling the bright object by a connected component labeling method and forming a connected component object; identifying, analyzing and combining the characteristics of the connected component object to form a lamp object by the bright object recognition; tracking the trajectory of the lamp object by a multi-vehicle tracking method; and identifying the type of a vehicle having the lamp object by the vehicle detection / recognition and counting the number of various vehicles.

An input image (e.g. a digital RGB color image) is subjected to an eye classifier that is targeted at discriminating a complete eye pattern from any non-eye patterns. The red-eye candidate list with associated bounding boxes that are generated by the red-eye classifier are received. The bounding rectangles are subjected to object segmentation. A connected component labeling procedure is then applied to obtain one or more red regions. The largest red region is then chosen for feature extraction. A number of features are then extracted from this region. Then these features are used to determine if the particular candidate red-eye object is a mouth.

A hybrid connected component labeling process for analyzing digitized or binary images includes the following steps. Firstly, a forward scan is executed to assign a forward label to each foreground pixel in the image. Then, a backward scan is executed to assign a backward label to each foreground. The backward labels are rearranged and label connection is recorded. A label allocation table including final labels and reference labels is provided for recording the use of the labels. When an object is considered as noise, the label corresponds to the pixels of the object is released by updating the label allocation table.

The invention discloses an infrared image target recognition method. The method solves the technical problem of two-dimensional infrared image segmentation by using the concept of one-dimensional image processing, i.e. each row of image pixels for cyclic segmentation processing is synthesized into two-dimensional binary images; and the concrete steps of the method are as follows: firstly, the infrared image enhancement is performed by a multiscale wavelet transformation method; secondly, the stable segmentation of the infrared targets under various imaging conditions is realized by using the extraction to the convex area in the infrared image, and after each row of the pixels is processed, the processing results are synthesized into two-dimensional binary images with sizes consistent withthat of the original images; thirdly, the binary image connected component labeling is performed to the segmented images, meanwhile, the basic characteristics are extracted, then the characteristics to be extracted are defined on the basis, and the characteristic extraction is calculated by using the basic characteristics; fourthly, a candidate target area is determined by a voting method, the accurate recognition is performed by ultimately combining the prior knowledge of the target to be recognized, and the final target is determined from the candidate targets. The invention has the advantages of accuracy, real-time and robustness.

The invention discloses a complicated background bar code image automatic partitioning method based on gradient direction consistency. The complicated background bar code image automatic partitioning method comprises the following steps: a first step, converting an input image into a grey image I(n) and calculating horizontal direction gradient Gx(n) and perpendicular direction gradient Gy(n) of each pixel point n on the grey image I(n) to obtain a gradient amplitude G(n) and a gradient angle phi (n) of each pixel point n; a second step, selecting a gradient threshold value alpha, taking the pixel point with the gradient amplitude G(n) larger than the threshold value alpha as an image edge point to obtain an edge image Ie(n); a third step, performing gradient direction consistency estimation and screening based on partitioning to obtain a candidate image block set; a fourth step, calculating an average gradient direction Phici of each candidate image block Pci, and normalizing the average gradient direction Phici to a [-phi / 2, phi / 2] interval as the gradient direction of the candidate image block Pci; a fifth step, performing connected component labeling on the candidate image block based on direction consistency; a sixth step, analyzing and screening a candidate image block connected domain; a seventh step, performing waveform extraction and matching on a candidate bar code region for follow-up decoding. The method disclosed by the invention can identify a bar code more quickly and accurately.

The present invention provides a method and system for detecting a water body area on a synthetic aperture sonar image, the method comprising: using two receiving arrays to generate two synthetic aperture sonar images of the same area; calculating the amplitude of the two images Mean value, generate a set of threshold sequences according to the amplitude mean value, and use one of them as the current image; select one of the set threshold values ​​as the set threshold value; binarize the current synthetic aperture sonar image according to the set threshold value; The transformed image is marked with connected domains, and the area is grown to obtain the water body area; repeat the above method to obtain the water body area detected by the current image under different thresholds; and select the other image of the two images as the current image to repeat the above method ; Obtain the difference between the water body regions detected by the two images when the threshold is the same, and use these differences as a standard to evaluate the water body region results, and select the best evaluated water body region results as the water body region detection results.

The invention discloses an online quality inspection method for diamond wire particles based on machine vision. A single-axis mechanical arm is arranged along the moving direction of the diamond wire, a camera and a microscope lens are arranged on the single-axis mechanical arm, and the movement of the mechanical arm and the diamond wire in the same direction reduces the number of cameras. The relative speed difference with the diamond wire improves the clarity of the image. Use the workstation to perform RIO cropping on the original image of the diamond wire, and delete the image area that does not contain the diamond wire information; filter the RIO cropped image to remove the noise of the image; Perform binary segmentation on the filtered image; use morphological closure operation to optimize the image after binary segmentation; use the connected region marking method based on run-length code table search to count the diamond wire particles in the optimized image, and get the diamond wire The statistical data of the particles, the image acquisition and processing can be completed when the diamond wire is moving, and the online quality inspection is realized. The detection speed is fast, and the entire quality inspection process is automatically completed.

The invention discloses a silkworm cocoon counting method based on image segmentation. In the silkworm cocoon counting method, the collected silkworm cocoon sample images are preprocessed, and then the preprocessed images are sequentially subjected to K-means clustering, binarization processing, and expansion and erosion, and finally the number of connected regions in the processed image is the number of cocoons. The silkworm cocoon counting method of the present invention successfully separates each silkworm cocoon in the silkworm cocoon sample image, effectively solves the problem of inaccurate counting caused by silkworm cocoon adhesion, and performs binary processing on the clustered image based on the adaptive threshold segmentation algorithm The accuracy of binarization is greatly improved, and then the accuracy of counting results is improved, and the connected region labeling method is used for counting, which greatly improves the accuracy and speed of counting.

The invention relates to a method for marking connected domains, in particular to a method for marking connected domains based on projection transformation. This method takes the binary image as the research object, and uses the projection transformation method to realize the extraction and analysis of the connected regions in the image. The connected domain marking method based on projective transformation realizes the connected domain marking and analysis by projecting the two-dimensional image to the one-dimensional direction, and segmenting and extracting the region according to the projection result. This method uses the idea of ​​projection transformation to transform a two-dimensional region segmentation problem into a one-dimensional line segment segmentation problem, and simplifies the connected domain labeling problem. Compared with the traditional method, the method in this paper is simple to implement and faster to process.

The invention provides a method for identifying a serial number on a bill. The method includes the steps that a current identification serial number can be generated from the serial number on the bill according to a binaryzation algorithm, a connected component labeling algorithm, a linear classi er algorithm and a tangent distance algorithm, and the current identification serial number is checked through the other serial number (if the other serial number exists) on the bill and / or a check sum valve corresponding to the serial number on the bill. Thus, the identification accuracy of the serial number on the bill can be effectively improved.

The present invention relates to a method for estimating a measure of a penumbra size of a penumbra in biological tissue wherein first and second images obtained respectively by perfusion weighted imaging (PWI) and diffusion weighted imaging (DWI) are analyzed, and wherein the analysis of the first image includes application of a level set method, and the analysis of the second image includes application of a grayscale morphological operation. In a further embodiment of the invention, a connected component labeling algorithm maybe applied on any one of the first and second image. The invention further relates to a system, a computer program product and use of a corresponding method.

The invention discloses a laser infrared fusion detection and recognition system, which comprises a nonuniformity correction SoC chip, an image rotation ASIC chip, a multistage filtering ASIC chip, a connected component labeling ASIC chip, a main DSP, a slave DSP, a main FPGA processor and a slave FPGA processor, wherein the slave FPGA processor controls each ASIC / SoC chip to complete laser and infrared image pretreatment; and the main FPGA processor controls the main DSP, the slave DSP and the labeling ASIC to help to complete long-distance, middle-distance and near-distance target fusion detection and recognition. The parallelism and the real-time performance of the system are improved, and power consumption of the system can be effectively reduced.

The present invention provides a side face detecting method and device based on head profile authentication. The method comprises a step of carrying out skin color binary processing on a collected image, carrying out communication component marking on the processed image, and obtaining a plurality of rectangular skin color areas, a step of carrying out hair color binary processing on the areas if the areas satisfy a preset skin color judgment condition, a step of adjusting an interest area according to the areas if the areas satisfy a preset hair color judgment condition, a step of detecting an edge pixel point in the interest area, a step of making a head top profile template according to the interest area, a step of using the template to carry out distribution in the interest area and calculating the edge pixel point quantities in the template in different positions, and a step of taking a quantity maximum value and determining that the rectangular skin color areas have an effective head and side face detection is successful if the maximum value reaches a preset edge pixel value. According to the method and the device, the problem of large angle side face detection is solved, the technical threshold of side face detection is lowered, and the flow of the side face detection is simplified.