62 results about "Fourier descriptor" patented technology

The invention discloses a gesture identification method and a gesture identification system, wherein the method comprises the following steps: step S1, a gesture detecting step: detecting a gesture in video stream in real time, marking a region in which the gesture is detected as an interested region; step S2, a gesture segmenting step: processing the interested region by utilizing skin color segmentation, and then performing edge detecting and outline extracting to obtain a point sequence of a hand-shaped outline; step S3, a gesture identifying step: firstly extracting a Fourier descriptor of the hand-shaped outline and then mapping the Fourier descriptor to a new vector of a characteristic space by utilizing PCA (principal components analysis), comparing the distance between the new vector and a gesture clustering center obtained by training, and judging a gesture type represented by the vector. The gesture identification method and the gesture identification system disclosed by the invention can improve the gesture identification accuracy and the gesture identification efficiency, and can effectively avoid background color interference.

The invention discloses a hybrid neural network-based gesture recognition method. For a gesture image to be recognized and a gesture image training sample, first a pulse coupling neural network is used to detect to obtain noise points, then a composite denoising algorithm is used to process the noise points, then a cell neural network is used to extract edge points in the gesture image, connected regions are obtained according to the extracted edge points, curvature is used to perform fingertip detection on each connected region to obtain undetermined fingertip points, interference of a face part is eliminated to obtain a gesture region, then the gesture region is partitioned according to gesture shape features, Fourier descriptors which keep phase information are obtained according to contour points of the partitioned gesture region, and first multiple Fourier descriptors are selected as gesture features; and a BP neural network is trained according to gesture features of the gesture image training sample, and the gesture features of the gesture image to be recognized are input to the BP neural network for recognition. The hybrid neural network-based gesture recognition method provided by the invention improves the accuracy rate of gesture recognition through utilization of various neural networks.

The invention relates to a Fourier descriptor and gait energy image fusion feature-based gait identification method. The method comprises the steps of performing graying preprocessing on a single frame of image, updating a background in real time by using a Gaussian mixture model, and obtaining a foreground through a background subtraction method; performing binarization and morphological processing on each frame, obtaining a minimum enclosing rectangle of a moving human body, performing normalization to a same height, and obtaining a gait cycle and key 5 frames according to cyclic variation of a height-width ratio of the minimum enclosing rectangle; extracting low-frequency parts of Fourier descriptors of the key 5 frames to serve as features I; centralizing all frames in the cycle to obtain a gait energy image, and performing dimension reduction through principal component analysis to serve as features II; and fusing the features I and II and performing identification by adopting a support vector machine. According to the method, the judgment whether a current human behavior is abnormal or not can be realized; the background is accurately modeled by using the Gaussian mixture model, and relatively good real-time property is achieved; and the used fused feature has strong representability and robustness, so that the abnormal gait identification rate can be effectively increased.

The invention discloses an automatic building identification method based on a Fourier descriptor. The method is used for automatically identifying illegal buildings of illegal land use. The method comprises the following steps: an original building RGB color image is converted into a gray edge image through filtering, enhancement and edge detection; all the contour of the gray edge image is extracted according to a Fourier descriptor method; the contour points are written in the plural form, discrete Fourier transform is carried out, and then modulus operation is performed on the value after transformation to obtain a normalized Fourier descriptor; the extracted contour data is simplified according to a DP algorithm; a building shape standard template library is established according to the simplified building contour; the Euclidean distance with the standard template library is calculated, and the Euclidean distance is used to represent the similarity of two Fourier descriptors; and finally, whether the extracted contour is a typical building shape feature is judged. Therefore, the purpose of automatically identifying an illegal building in a video image is achieved.

The invention provides a method and a device for retrieving a three-dimensional object, wherein, the retrieval method comprises the following steps: obtaining the characteristic view of a three-dimensional object, and extracting the first Shock diagram skeleton descriptor and the first Fourier descriptor of the characteristic view; extracting the second Shock diagram skeleton descriptor and the second Fourier descriptor of an input picture; calculating the first similarity metric of the first and the second diagram skeleton descriptors; calculating the second similarity metric of the first and the second Fourier descriptors; and obtaining a mixed similarity metric according to the first similarity metric and the second similarity metric, confirming that the characteristic view corresponding to the maximum mixed similarity metric is the view most similar to the input picture, and further confirming that a three-dimensional object corresponding to the characteristic view is the three-dimensional object ready for retrieving. The invention effectively combines the Shock diagram skeleton descriptors and the Fourier descriptors, and realizes excellent 3D-object retrieval performance.

The invention discloses a monocular static gesture recognition method based on multi-feature fusion. The method comprises the following steps: gesture image collection: collecting an RGB image containing gesture by a monocular camera; image preprocessing: using human skin color information for skin color segmentation, using morphological processing and combining with the geometric characteristicsof the hand, separating the hand from the complex background, and locating the palm center and removing the arm region of the hand through the distance transformation operation to obtain the gesture binary image; gesture feature extraction: calculating the ratio of perimeter to area, Hu moment and Fourier descriptor feature of gesture and forming gesture feature vector; gesture recognition: usingthe input gesture feature vector to train the BP neural network to achieve static gesture classification. The invention combines the skin color information and the geometrical characteristics of the hand, and realizes accurate gesture segmentation under monocular vision by using morphological processing and distance transformation operation. By combining various gesture features and training BP neural network, a gesture classifier with strong robustness and high accuracy is obtained.

The invention discloses a static gesture recognition method which combines depth information and skin color information. The method comprises the following steps: acquiring RGB image and depth image by kinect; using the depth threshold and human skin color information to get the binary image of the hand. Distance transform operation combined with palm cutting circle and threshold method is used tojudge whether there is arm region in hand image. The arm region is removed by exclusive OR operation between images to obtain binary gesture image. Calculating the number of Fourier descriptors and fingertips to form the eigenvector of the gesture; Support vector machine is used for gesture classification to achieve the purpose of gesture recognition. The invention realizes hand partial cutting by combining depth information and skin color information, and overcomes the influence of skin color region in complex background. By removing the arm region, the disturbance of the arm to the classification accuracy of the system can be overcome. The number of fingertips and Fourier descriptor features are computed and input to support vector machine for gesture recognition.

The invention relates to a Fourier descriptor and BP (Back Propagation) neural network-based garment style identification method. The method comprises the steps of preprocessing a garment image to obtain an outer contour of the garment; performing Fourier description of the outer contour of the garment, and performing data preprocessing; and performing BP neural network-based garment style identification. The preprocessing of the garment image refers to a process that the garment image is subjected to segmentation processing to obtain a garment region, and the garment image is subjected to edge detection to obtain a contour image of the garment; the Fourier description of the outer contour of the garment refers to a process that a standardized Fourier descriptor eigenvector of a contour shape of the garment is extracted, and the data preprocessing refers to normalization processing and principal component analysis performed on the standardized Fourier descriptor eigenvector; and the BP neural network-based garment style identification refers to garment style identification performed on a principal component matrix by using a three-layer BP neural network. The method can achieve the identification accuracy of 81%, is good in robustness and generalization ability, and can be suitable for identification of garment styles in garment images.

The invention discloses a human face identification algorithm based on an image sensor imaging system. On the basis of three-dimension human face data obtained through a three-dimension human face imaging system of a relevant image sensor (CIS), the algorithm firstly expresses the three-dimension human face data through an isobathic line and a Fourier descriptor, extracts the characteristics through an improved manifold learning method, and finally realizes classified identification through a nearest neighbor classifier based on Euclidean distance. According to the invention, the three-dimension data utilized in the human face identification is converted into two-dimensional data (isobathic line) to be processed, thereby reducing the data processing complexity on the basis of maintaining the original human face information; and meanwhile, the improved manifold learning method is used for extracting the characteristics, so that the identification efficiency and the identification speed are greatly improved.

A system for matching nodes of different spatial data sources. The system matches or ties a node from one data source to another data source. The present invention uses a highly accurate exclusion technique for excluding potential corresponding nodes of a second data source. Potential corresponding nodes are selected using node degree comparison, adjacency, fourier descriptor analysis of shapes formed by connecting paths between a node under investigation and a secondary node associated with that node under investigation, an analysis of total length to secondary nodes, an angle analysis to eliminate reflections, a secondary node degree analysis, and a surrounding node correlation coefficient comparison. Using one or more of these techniques in various combinations can result in the selection of a potential corresponding node with a high degree of accuracy.

The invention relates to an intrusion detection algorithm based on a Fourier descriptor and a histogram of oriented gradient, which comprises the steps of 1) acquiring a continuous video frame image; 2) carrying out inter-frame difference on the video frame image, performing a series of morphological operations, and acquiring an actual motion region; 3) judging whether the motion region is a human body target or not, intercepting the video frame image and a binary image corresponding to the actual motion region acquired in the step 2), extracting HOG (histogram of oriented gradient) features of the video frame image in a sliding window mode, carrying out judgment by adopting an HOG classifier, extracting FD (Fourier descriptor) features of a specified dimension in the binary image, carrying out judgment by adopting an FD classifier, and carrying out an AND operation on results acquired by the two classifiers so as to acquire a final judgment result; and 4) according to the judgment result of the step 3), carrying out follow-up behavior logic judgment if the motion region is judged to be a human body target, and giving out an alarm. The intrusion detection algorithm provided by the invention is high in robustness for illumination, the image quality and environmental interference, and the detection precision is high.

The invention discloses a three-dimensional model searching method based on a sketching. The three-dimensional model searching method based on the sketching comprises the following steps: (1) generating multi-perspective profile diagrams; (2) extracting a placeholder diagram characteristic, a distance conversion characteristic, a profile signature characteristic, a Fourier descriptor, a Hu moment characteristic and a Poisson characteristic of each multi-perspective profile diagram; (3) obtaining a form of a novel characteristic; (4) by means of the form of the novel characteristic, extracting characteristics of a corresponding dimension from all the characteristics of the multi-perspective profile diagrams obtained in the step 2 to form a novel characteristic; (5) extracting a sketching characteristic as the step 2; (6) obtaining a novel characteristic of the sketching from characteristics of a sketching profile diagram in the same mode as the step 4 by means of the form of the novel characteristic obtained from the step 3 in an off-line period; and (7) finding out a novel characteristic of the profile diagram most similar to the novel characteristic of the sketching through a k-d tree characteristic matching method and determining a three-dimensional model generating the profile diagram. The three-dimensional model searching method based on the sketching lowers sensitiveness on setting of parameters, and improves the searching effect.

The invention provides a typical ship target identification method based on graded invariance features. At first, the binary entropy and the normalized inertia moment of each image ship target are extracted as the primary feature; and then each image is subjected to wavelet decomposition to form four sub-images, and the weighted Hu moment, Zernike moment, and Fourier descriptor of the each sub-image ship target are extracted to be the secondary feature; the polar coordinate shape matrix of each image ship target is taken as the trinary feature; and all of the features are modified to have the properties of translation, rotation, and scaling invariance. The experimental result of a recognition classifier shows that the algorithm can describe the typical ship targets in satellite remote sensing images in details step by step, and the recognition accuracy is high. The method can be applied to the typical ship target recognition of the satellite remote sensing image database, and is an engineering method that is high in universality.

Provided is a method for detecting a profile phenotype of arabidopsis. The method includes the steps that a calibration board is placed in a planting pot of the arabidopsis, an RGB image of the arabidopsis is collected through a camera, the collecting image is preprocessed, automatic correction and calibration of the image are achieved, image correction is used for correcting deformation of the image, image calibration is used for obtaining a real size of a unit pixel, the preprocessed image is divided, the arabidopsis is separated from the background and extracted from the image, after the arabidopsis image is separated, a profile phenotype parameter of the arabidopsis is extracted, and an overall profile of the arabidosis is quantitatively described through an oval Fourier descriptor. The differences of the overall profile and growth orientation of the arabidopsis with different genes can be described through analysis of the profile phenotype parameters, and accordingly the function of different genes and influences of the different genes on the arabidopsis can be concluded.

The invention discloses a marine vortex recognition method based on MKL multi-feature fusion. The marine vortex recognition method comprises the following steps: 1) carrying out data preprocessing ona data set based on a synthetic aperture radar image; 2) inputting the preprocessed synthetic aperture radar images into a feature extractor in batches, and extracting gray level co-occurrence matrixfeatures, Fourier descriptor features and Harris features; 3) constructing different types of kernel function sets, obtaining a training set of gray level co-occurrence matrix features, Fourier descriptor features and Harris features, and performing multi-feature fusion based on multi-kernel learning on the training set to obtain a data set; and 4) constructing a classifier model which is used forclassifying the data set. According to the invention, a plurality of feature fusion strategies are adopted, and a plurality of different types of features are applied to the identification of the marine vortex, so that the limitation of the data processing capability and the limitation of the conventional artificial visual and threshold setting method on the identification of the marine vortex inthe prior art are overcome.

The invention discloses a method for predicting self-interaction effect of protein. The method comprises the steps of selection and establishment of a data set, generation of a PSSM matrix, extractionof a feature value by a Fourier descriptor, construction of a training set and a testing set and construction of a classifier model. According to the method, the feature value of a sample set is extracted by the Fourier descriptor, the number of times of multiplication required for discrete Fourier transform of a computing data set of a computer is greatly reduced, and the computing quantity is reduced. A model can be constructed by a random projection method, the prediction precision is greatly improved, and a good prediction effect can be achieved. The method is low in computing cost and small in power consumption; the self-interaction effect of the protein can be predicted effectively; and the prediction effect can reach 93% or above.

The invention discloses a traffic sign detection and identification method under the complex environment. According to the method, employing a color segmentation method of a normalization RGB space can reduce the image processing time, and algorithm timeliness can be guaranteed; employing a convex hull processing method can solve a problem that a traffic sign is partially shielded; contour analysis is carried out by employing a method extracting contour Fourier descriptors, and a problem that the traffic sign generates rotation, translation and scale change can be solved. Through adding the contour data with projection distortion to a standard database, a problem that the traffic sign generates projection distortion can be solved. Compared with the traditional method, the method has advantages of high robustness and good timeliness, and the method can be applied to intelligent automobiles or pilotless automobiles.

A method for monitoring cardiac function of a patient during a magnetic resonance imaging (MRI) procedure, including: acquiring an MR image sequence of the patient's heart during a cardiac phase; segmenting a left ventricle of the patient's heart in the MR image sequence, wherein the segmentation produces endocardial and epicardial contours; representing at least one of the contours in polar or radial coordinates and computing its Fourier transform, wherein the Fourier transform produces Fourier descriptors for the contour; putting a vector of the Fourier descriptors into a classifier, wherein the classifier determines whether the contour reflects normal wall motion in the cardiac phase or whether the contour reflects abnormal wall motion in the cardiac phase; and alerting a medical practitioner when abnormal wall motion is detected.

The invention discloses a three-dimensional face image recognition method based on the moment Fourier descriptor, and relates to the technical field of three-dimensional imaging technology. For a three-dimensional face depth map under different postures, a differential geometric correlation theory is used for correcting to the median plane, and the contour line features of a human face are extracted to turn the three-dimensional face into a two-dimensional curve which is easy to process. A novel method for combining a torque and fourier descriptor is provided for better describing the two-dimensional curve. The extracted curve features are used for carrying out face recognition. The problem of low accuracy and slow recognition speed of a conventional three-dimensional face recognition method can be solved, the calculation time can be effectively shortened, and the recognition rate is high.

The invention discloses a distributed optical fiber vibration signal recognition algorithm based on two-dimensional matrix feature recognition, and the algorithm comprises the following steps: a light signal is acquired and is transformed as an original electrical signal, the original electrical signal is filtered, two-dimensional data generation is performed on the waveforms, image dilation is performed on the generated two-dimensional data through a morphological processing method, hole filling is performed on the two-dimensional data through the morphological processing method, an MATLAB software system is utilized to realize local area filling of an image, region segmentation is performed on the generated two-dimensional data through the morphological processing method, character representation and description are performed on a recognized region, the image skeleton is extracted, a Fourier descriptor of the image is extracted, and the recognition of the vibration signal is performed through the Fourier descriptor. According to the invention, the recognition effect is higher than before, the data is more accurate than before, and the reliability is high.

The invention discloses an arrester health state determination method based on volt-ampere characteristic curve analysis. According to the method, first, an aging and moisture absorbing test is performed, a power grid voltage and a corresponding leakage current are collected, a volt-ampere characteristic curve of an MOA (Metal Oxide Surge Arrester) under aging and moisture absorbing conditions isdrawn, edge analysis of a Fourier descriptor is performed, and a Fourier descriptor data table under the aging and moisture absorbing conditions is created; second, a power grid voltage and a corresponding leakage current of the to-be-tested arrester are collected, a volt-ampere characteristic curve of the to-be-tested MOA is drawn, edge analysis of the Fourier descriptor is performed, and a Fourier descriptor data table under to-be-tested conditions is determined; and last, the Fourier descriptor data table under the to-be-tested conditions and the Fourier descriptor data table under the aging and moisture absorbing conditions are compared to judge the health condition of the arrester. Through the method, the health condition of the arrester can be analyzed quantitatively from the volt-ampere characteristic curves.

The invention relates to an image segmentation and identification technology, and aims at providing a combined segmentation and Fourier descriptor-based underwater target image identification method.The method comprises the following steps of: segmenting an image by utilizing grayscale information and color information on the basis of combination of a grayscale information-based multi-threshold segmentation algorithm and a color information-based HSV space segmentation algorithm, so as to improve the segmentation correctness; and identifying the shape of a target after the segmentation is completed. According to the method, Fourier descriptors are selected to depict shape features, shape determination is carried out by adoption of feature comparison and classification, a Fourier descriptor library is established, and identification problems are converted into clustering problems, so that the method is higher in expandability when being compared with the method of directly carrying outshape identification by utilizing feature operators. By adoption of the multi-threshold segmentation algorithm and the HSV color segmentation algorithm, the method is relatively mature in development, wide in application and easy to grasp. By utilizing the Fourier descriptor library to carry out classification and identification, the method is capable of expanding a classification library conveniently and is suitable for different scenes.

The invention discloses a human body identity identification method based on gait trajectory curve characteristics. The method comprises the steps that according to curve charts of gait acceleration data in a time domain, a gait identity identification problem is converted into a trajectory curve shape matching problem; from the aspect of iconography, in a paper, a Fourier descriptor is used for describing overall trajectory curve characteristics of rough outlines of gait curves, a concept of a direction angle descriptor is provided to further describe fine local trajectory curve characteristics of the gait curves, and finally the overall trajectory curve characteristics and the fine local trajectory curve characteristics are combined to complete matching of the gait trajectory curves. Itis indicated through results that the gait trajectory curve characteristics provided in the paper can be well applied to identity identification.

The invention provides a target-matching object recognition method and device and a storage device. The method comprises the steps that a three-dimensional model of a target object is established; a multi-view picture base of the three-dimensional model is generated; Fourier descriptors of picture contours in the picture base are calculated, and a matching base is generated; a real-time object image is acquired, and the image is preprocessed to obtain a comparison image; and the comparison image is matched with an image in the matching base, and a matching result is finally generated. The target-matching object recognition device and the storage device are used for implementing the method. Through the target-matching object recognition method and device and the storage device, good robustness is achieved while the expected object image is effectively recognized.

The invention provides a plant identification method based on elliptical Fourier descriptors and weighted sparse representation. The plant identification method is mainly and technically characterized by comprising the steps that leaf images are preprocessed, wherein all the colored leaf images are converted into grayscale images, the leaf images are separated from the background through an Otsu segmentation algorithm and converted into binary images, and small holes of the leaf images are eliminated through an erosion algorithm; edge detection is conducted by adopting a Canny edge detector; centroids of boundaries are calculated; the Fourier descriptors are calculated; a complete dictionary is constructured, wherein Fourier descriptor vectors of all leaf image data sets are divided into training sets and test sets, and the complete dictionary is composed of the Fourier descriptor vectors of all the training sets; and optimization is conducted by a weighted sparse representation classifier. According to the plant identification method, by adopting the elliptical Fourier descriptors, good robustness is achieved on noise and other factors, and by applying the weighted sparse representation classifier (WSRC) to plant species identification and particularly to a low-dimension space, the identification rate is obviously increased.