90 results about "Texture recognition" patented technology

The invention relates to a garment texture recognition and classification method based on LBP and GLCM. The classification method for the texture types in fashionable garment images comprises the following steps: S1, extracting LBP texture features; S2, designing a GLCM texture image feature statistic extraction function; S3, calculating a texture feature value; and S4, training the feature vectors by using an SVM algorithm, and classifying the texture pictures. The method has the advantages that the challenge of texture recognition in fashionable clothes is well solved in the actual production process, the constraint of manual recognition is liberated, convenience is brought to later data analysis and algorithm development optimization, and the method has important practical significance.

A dynamic texture recognition method on the basis of an image sequence is used for realizing dynamic texture recognition by comparison of three-dimension HMT (Hidden Markov Tree) model parameters and includes performing Surfacelet transformation to the image sequence to obtain a coefficient matrix, particularly realizing multiscale decomposition to the image sequence by pyramid decomposition, decomposing three-dimension signals to different directions through 3D-DFB(3D directional filter banks) which are in series connection with two 2D-DFB, saving sub-band data acquired after the Surfacelet transformation through the three-dimension matrix, generating a coefficient matrix through extracted feature vectors; building a three-dimension HMT model to the coefficient matrix by details of realizing distributed modeling of coefficient by utilizing a Gauss mixture model, realizing inter-scale continuous modeling of the coefficient by utilizing the three-dimension HMT model, and then solving parameters of the HMT model by an EM algorithm. Corresponding expansion schemes are provided for handling the situation that processed data contain different types of dynamic textures. The dynamic texture recognition method is easy, high in adaptability and good in recognition effect.

The invention discloses an identity authentication method based on a finger vein, which comprises the following steps: pre-processing the received finger vein image to be identified to obtain a pre-processed image; Convolution calculation is carried out on the preprocessed image, and characteristic point map and vein pattern map are generated according to the obtained convolution value. Accordingto the eigenvalue size and coordinate information of the feature points in the feature point map, the feature points in the feature point map are selected globally according to the preset conditions,and the feature points with high eigenvalue in the global distribution are taken as the feature points to be matched. The vein pattern and template image are matched and compared according to the feature points to be matched, and the identity authentication result is generated according to the matched result. This method extracts feature points from global and local features to obtain globally distributed feature points with high eigenvalues, which can improve the accuracy of vein texture recognition and generate high-precision recognition results. The invention also discloses an identity authentication device e based on a finger vein and equipment, which have the beneficial effects mentioned above.

The invention discloses an effective multiscale texture recognition method. The method comprises the steps that an image pyramid of an input image is calculated firstly, then an LBP operator is applied to the image pyramid with various scales, next, the image pyramid of each scale generates a feature vector, multiscale information is integrated through similarity fusion on each scale according to the D-S evidence principle, and particularly, the similarity of the tested image and a target sample is calculated by fusing the similarity between the tested image and the sample of each scale. By means of the effective multiscale texture recognition method, the identification precision of a public data set Brodatz'salbum and an MIT video texture database (VisTex) reaches 96.43% and 91.67%. Meanwhile, the method has a certain robustness to image rotation invariance and has a certain application value in the practical application.

A three-dimensional multilayer skin texture recognition system and method based on hyperspectral imaging. Three-dimensional facial model associated with an object may be acquired from a three-dimensional image capturing device. A face reconstruction approach may be implemented to reconstruct and rewarp the three-dimensional facial model to a frontal face image. A hyperspectral imager may be employed to extract a micro structure skin signature associated with the skin surface. The micro structure skin signature may be characterized utilizing a weighted subtraction of reflectance at different wavelengths that captures different layers under the skin surface via a multilayer skin texture recognition module. The volumetric skin data associated with the face skin can be classified via a volumetric pattern.

The invention relates to a texture recognition method based on a deep self-attention network and local feature coding. The method comprises the following steps: designing a deep self-attention module with four stages according to characteristics of a texture image, merging local image blocks in the first three stages to increase a receptive field, and limiting a self-attention calculation in local space with a fixed size; in the last stage, canceling local image block merging, globally calculating self-attention, and obtaining a relation between local blocks; therefore, better extracting texture features of a local area, and keeping global features not lost. According to a PET network provided by the invention, texture information in a local area in an image is fully combined, and two-dimensional features output by a backbone network are remodeled into a three-dimensional feature map; block descriptors of multiple scales are densely sampled in a feature map through a moving window, and a group of multi-scale local representations is obtained; finally, local feature coding and fusion are carried out on the multi-scale block features, and a fixed-scale texture representation is generated for final classification.

A trunk texture recognition method based on a four-channel convolutional neural network comprises the following steps: S1, collecting a trunk image data set to ensure the integrity of an image; s2, preprocessing the image; and S3, training a multi-classifier by using the four-channel convolutional neural network. According to the invention, richer image feature information can be obtained, features with higher discrimination can be learned automatically, and the classification precision and robustness of the classifier obtained by training are higher.

The invention discloses a video smoke detection method based on a convolutional neural network, and belongs to the field of picture recognition. Existing fire smoke identification algorithms have strong scene pertinence and are particularly susceptible to environmental interference. The invention discloses a video smoke detection method based on a convolutional neural network. The method comprisesthe following steps: preprocessing an acquired video image; performing suspected smoke area extraction on the preprocessed video image; carrying out smoke feature description on the obtained suspected smoke area; and based on a convolutional neural network smoke texture recognition framework, performing smoke recognition on the to-be-detected area obtained in the previous step by utilizing a convolutional neural network method. The moving target in the foreground image is input into the CNN model for smoke identification, so that the smoke detection efficiency is improved while the static object interference is reduced.

The invention specifically relates to a dynamic texture recognition method based on multi-task learning and is designed so as to improve dynamic texture recognition accuracy. The dynamic texture recognition method based on multi-task learning is characterized by, to begin, carrying out chaotic feature vector extraction on each pixel time sequence in a dynamic texture video, so that the video is changed into a chaotic feature vector matrix; then, carrying out video modeling through a bag-of-word model to obtain histogram features; and converting the identification problem into group sparse representation, and calculating through an ADMM method. The method can carry out dynamic texture recognition through a multi-task learning method, can be widely applied to various civilian and military systems, such as a video monitoring system, a video conference system, an industrial product detection system, a robot vision navigation system and a military target detection and classification system,and has a wide market prospect and application value.

The invention relates to a garment pattern texture recognition method and system based on a neural network. The system comprises an image feature extraction network, a garment position feature extraction network, a texture feature extraction module, a garment position regression module and a pattern texture prediction module. The method has the advantages that the specific position of the garmentin the image is judged, the specific position content of the garment serves as the ROI to be input into the neural network, background interference can be greatly eliminated, and the recognition accuracy is improved; by utilizing the system to analyze popular elements of the current fashion garment industry, the current fashion trend is analyzed, meanwhile, the system can also provide design inspiration for fashion designers, help the designers to design garment products meeting the psychology of consumers, and improve the cognition degree and satisfaction degree of the consumers; data labeling on 700-800 image frames is completed per minute, so that the efficiency is greatly improved; by using the system, the identification accuracy is improved by about 20%.

The invention discloses a car brake pad surface defect detecting and automatic sorting system based on machine vision. The system comprises a machine vision system. The machine vision system comprisesa camera, a camera lens and a light source. The machine vision system is connected to a CPU. The CPU is connected with an upper computer and a sorting arm. The upper computer involves image preprocessing, image segmentation, feature extraction and mode recognition, mode recognition involves color recognition, shape recognition, texture recognition and multi-feature fusion recognition, the machinevision system detects a to-be-sorted target image and transfers information to the CPU, and the CPU processes the received information. The car brake pad surface defect detecting and automatic sorting system based on machine vision is reasonable in design, brake pads are conveniently detected through image acquisition, processing and recognition, the sorting arm is controlled to act by means of detecting results, brake pads with detects are sorted to an unqualified area, qualified products leave a factory, the purpose of automatic sorting is achieved, and using is facilitated.

The invention provides a color film substrate and a display device. The color film substrate comprises an underlayer substrate, a color film layer, a plurality of texture recognition units; the colorfilm layer is arranged on one side of the underlayer substrate and comprises a black matrix and a plurality of color filters; the plurality of texture recognition units are arranged on the underlayersubstrate, each texture recognition unit comprises a thin film transistor and an optical sensor, and the orthographic projections of the texture recognition units on the substrate are located in the orthographic projection of the black matrix on the substrate. The texture recognition units are designed on the color film substrate, so that the optical path between an optical sensor and the surfaceof a finger can be effectively shortened, an acquired optical signal is enhanced, and meanwhile, the noise of the optical signal transmitted between LCD layers is greatly reduced; and meanwhile, LCD light transmittance is not influenced.

The invention discloses an unsupervised content-preserved domain adaptation method for multiple CT lung texture recognition, which belongs to the field of image processing and computer vision. This method enables the deep network model of lung texture recognition trained in advance on one type of CT data (on the source domain), when applied to another CT image (on the target domain), under the premise of only obtaining target domain CT image and not requiring manually label the typical lung texture, the adversarial learning mechanism and the specially designed content consistency network module can be used to fine-tune the deep network model to maintain high performance in lung texture recognition on the target domain. This method not only saves development labor and time costs, but also is easy to implement and has high practicability.

The invention provides a defective pixel detection and correction device based on texture recognition. The device comprises an image sensor data input unit for inputting image data of an image sensor;a chromatic aberration compensation unit connected to the image sensor data input unit, and used for counting the chromatic aberration of each pixel of the image data in an m*n data window, selectinga median value of the chromatic aberration to obtain a chromatic aberration value of the current pixel, subtracting the chromatic aberration value, and performing chromatic aberration compensation; adefective pixel detection unit for calculating a total variation of N pixels around the current pixel, including pixels of the same channel or different channels, determining the texture intensity ofthe current pixel according to the calculation result, and detecting defective pixels; and a defective pixel correction unit for collecting the defective pixels according to the detection result of the defective pixel detection unit. The invention further provides a defective pixel detection and correction method based on texture recognition.

The invention provides a terminal control method and system based on facial texture recognition. The method includes: acquiring current first facial texture information of a user; acquiring a target operation, which corresponds to the first facial texture information, in a facial texture database according to the first facial texture information; and executing the target operation corresponding tothe first facial texture information. Through the method, quick control which is on a terminal and is based on the facial texture information can be realized, and thus use efficiency and security ofthe terminal are improved.

The invention provides an image texture recognition method and system. The method specifically comprises the following steps: step 1, acquiring image data in an operation process; step 2, encoding the image data; step 3, carrying out point-by-point calculation on gray values of local areas of the images according to the relation between the image space position information and the image gray values, respectively counting the occurrence times of different LBP values so as to describe texture features of the images in the areas, and carrying out texture feature extraction on the encoded data; step 4, inputting the extracted texture features into an image recognition model for recognition and classification; and step 5, outputting an identification and classification result for assisting the industrial operation process. The texture features of the image data are analyzed, and the image features are better extracted, so that the image analysis result better meets the requirement of real-time operation.