55 results about "Image cube" patented technology

The present invention relates to a human behavior identification method based on a three-dimensional convolutional neural network and a transfer learning model. The method comprises: performing sampling frame by frame of video, stacking obtained continuous single-frame images on a time dimension to form an image cube with a certain size, and taking the image cube as the input of the three-dimensional convolutional neural network; performing training of a basic multi-classification three-dimensional convolutional neural network model when implementation is performed, selecting part of classes of input samples from a test result to construct a sub-data set, training a plurality of dichotomy models on the basis of the sub-data set, and selecting a plurality of models with the best dichotomy result; and finally, employing the transfer learning to transfer the knowledge learned by the models to an original multi-classification model to perform retraining of the transferred multi-classification model. Therefore, the multi-classification identification accuracy is improved and human behavior identification with high accuracy is realized.

Seismic data are assembled and stored for a set of cross-correlation lag times to form an array of common image gathers over depth levels of interest. The two dimensional gathers assembled over different lag times form a three-dimensional cube of common image data. The data are analyzed to determine the travel time shift required to equalize upgoing and downgoing wavefields. Events in the common image gathers are then modeled using Green's functions to generate a data set representing the data resulting from processing had a precise velocity model been obtainable from the seismic data. The generated data are then processed with inversion techniques to form a velocity model for seismic data analysis.

The invention relates to a system for rebuilding spectrum of high spectrum intervention data and a method thereof. Based on the interference imaging mechanism of an HJ-1A high spectrum camera, the system adopts a particular data processing method to rebuild intervention data into spectrum data, which solves the practical engineering problems in processing ground system building data. The system comprises a relative radiation correction unit, an interference data de-noising unit, an apodization processing unit, a spectrum recovering unit and a spectrum checking unit; wherein, the spectrum recovering unit converts an interference data cube into a spectrum image data cube by the symmetrization processing of interference data; the spectrum checking unit carries out wavelength positioning to a real object spectrum image cube by checking the position of the characteristic spectral line of an imaging spectrometer. The technical scheme of the invention fills the blank of data processing of satellite-bone interference imaging spectrometers in China, and satisfies user demand for the data of satellite-bone interference imaging spectrometers.

The invention discloses a comprehensive processing method of multi-spectral imaging information of core microscopy, which is used for comprehensively analyzing and processing the multi-spectral imaging data of the microscopic fluorescence, infrared and Raman of the core with oil in the geological logging of oil. At first, the color image of full color light of the polished section of the core is acquired with the fluorescence microscope; the three-spectral imaging analysis system of the microscopic fluorescence, infrared and Raman is then used for measuring the core sample in the geological logging to acquire the corresponding imaging ''spectral cube'' or spectral ''image cube''; the pixel level blending technology is used for data processing and comprehensive analysis of the three-spectral imaging information; a DWT is used for transforming; noise in the image is eliminated; the sub-images of each level are blended in a weighing way; the DWT is used for inverse transformation to acquire the blended multi-spectral image. The experiment has proven that the data blending technology of the spectral information of the image formed by multi microscopic spectra has higher practical value in the mineralogy, petrology and geochemistry of the core, when compared with each independent spectral imaging method.

A method of training a user to identify an object in an image, by querying a computer system having a lexicon of photo-interpreters to formulate object extraction rules. A database to be queried by an expert photo analyst is generated and a programming language is provided to generate extraction rules against that database. Graphical query results are displayed to determine whether an object has been identified. First, a hyperspectral image cube is represented by a set of fraction planes and texture transforms. Mean spectral reading values are obtained and used to build a pseudo multivariate distribution of the values. Recognizable features are then extracted from the hyperspectral image cube. A textural transform of a hyperspectral image cube results. The hundreds of image bands reduced to a few texture transforms can be used to train users of the image recognition system.

The invention relates to a method for fusing multi-scale spectrum projection remote sensing images. The method comprises the following steps of: performing accurate space aligning on the original panchromatic image and the original multispectral image; generating a Gauss image cube consisting of the original multispectral image of each waveband with a low space resolution and the original panchromatic image with a high space resolution by introducing an improved Gauss scale space theory; extracting the spectral characteristics of the multispectral image of each waveband and the panchromatic image by using the laminar image of the Gauss image cube; and projecting the spectral characteristics onto the original panchromatic image according to a weighted fusion strategy to obtain a fused multispectral image having the high space resolution and a high spectral resolution. The method has the advantages of increasing the space resolution of the image, keeping the spectral information of the original multispectral image to the maximum extent, acquiring a better balance between the space resolution and the spectral resolution and obtaining the fused image with higher quality and practical value.

The invention relates to an injurious insect image automatic recognition method based on a deep restricted Boltzmann machine. The method comprises a training process and a testing process. According to the training process, data of a training image set are preprocessed, preprocessed training images are grouped to construct a training image cube, features of each training image group are extracted through a restricted Boltzmann machine algorithm, and trained training image set feature data are obtained through feedback adjustment. According to the testing process, a test image to be recognized is input and is preprocessed, features of the test image are extracted through the restricted Boltzmann machine algorithm, and feature data of the test image with small errors are obtained through feedback adjustment; the varieties of injurious insects are recognized, and a preventive method is given out. The invention discloses an injurious insect image automatic recognition system based on the deep restricted Boltzmann machine. The injurious insect recognition rate and procedure robustness are improved, and the actual application value of injurious insect recognition in agricultural production is improved.

The invention discloses a hyperspectral-analysis-based copper quality detection method, which comprises the following steps of: (1) constructing a characteristic image cube of a copper material to be detected; (2) extracting contents and spectral information of ingredients of the copper material to be detected from the characteristic image cube; and (3) evaluating the quality of the copper material to be detected. According to the method disclosed by the invention, the concept of the characteristic image cube is introduced based on the traditional hyperspectral image processing technology, meanwhile, the image characteristics and the spectral characteristics of the copper material to be detected are acquired, and the copper material is subjected to characteristic extraction and quality detection by using a fused PSO-nsNMF (Particle Swarm Optimization-based non-smooth Nonnegative Matrix Factorization) algorithm, therefore, the accuracy in the detection of copper quality is ensured. Theinvention further discloses a hyperspectral-analysis-based copper quality detection system which comprises a hyperspectral imaging unit and an image processing unit; the hyperspectral-analysis-based copper quality detection system disclosed by the invention is used for carrying out real-time online acquisition and data processing on an image of a copper sample by utilizing a computer and combining with a hyperspectral imaging instrument, therefore, the timeliness in the detection of copper quality is ensured.

The invention discloses a hyperspectral image classification method of a graph convolution network based on a multi-graph structure. The hyperspectral image classification method comprises the following steps: reading a hyperspectral image cube and a training set; rearranging the hyperspectral data to obtain a pixel data matrix; constructing a forced nearest neighbor connection matrix and a spatial nearest neighbor connection matrix according to the pixel data matrix; performing convolution on the forced nearest neighbor connection matrix, the spatial nearest neighbor weight matrix and the pixel data matrix through a graph convolution network, and obtaining a feature matrix; and splicing the feature matrixes, classifying the feature matrixes AF of the pixels by using a softmax classifier,and finally verifying a hyperspectral image classification result.

The invention discloses a hyperspectral imaging system combining chromatic aberration blurred imaging and image reconstruction technologies, and the system comprises a chromatic aberration blurred imaging subsystem which comprises an imaging lens and a CMOS camera, the central axis of the imaging lens and the central axis of the CMOS camera are arranged in an aligned manner, and the imaging lens has axial high chromatic aberration in a visible light band or an infrared light band; the CMOS camera is used for collecting chromatic aberration blurred images under different image distances; an image processing unit is used for stacking the chromatic aberration blurred images to form a chromatic aberration blurred image cube, and based on a spectral image cube degradation model, in combinationwith the wavelength image distance relation curve and the point spread function library, the spectral image cube is reconstructed by using a sequence image deconvolution algorithm. The hyperspectral imaging system disclosed by the invention has the advantages of simple structure, high light energy utilization rate, high detection speed and the like, so that a high-sensitivity, rapid and miniaturized detection means is provided for spatial spectral information perception and specific identification of a target object in a special occasion.

The invention discloses a multi-image encryption method based on a three-dimensional scrambling model and fractional Fourier transform (FrFT), and belongs to the field of information encryption. At present, the data volume of network information interaction is increasingly increased, and image batch transmission and storage are often needed in practical application. In order to protect the contentof multiple images, multiple interactive images are constructed into an image cube, and the image cube is segmented into an image cube set. A three-dimensional scrambling model of an image is established by utilizing two-dimensional Zigzag transformation and a chaotic system, and pixel scrambling operation is carried out on a cubic block set of the image. Then, FrFT (Fast Fourier Transform) is carried out on the scrambling result to generate k encrypted images. Experiments show that the method can protect the content of multiple interactive images at the same time, and is high in safety, goodin encryption effect and efficient.

The invention provides a multi-channel-based wordart generation method and system. The method includes the following steps that: the offset images of a source word image and a target word image are extracted; the source word image, the offset image of the source word image and a source wordart image are combined into a multi-channel source image cube according to channels; a target wordart image is initialized, and the initialized target wordart image, the target word image, and the offset image of the target word image are combined into a multi-channel target image cube according to channels;similar image blocks of each image block in the target image cube are iteratively searched in the source image cube, and the wordart image channels of the similar image blocks are adopted to update the corresponding channels of the target image cube until convergence occurs or the update degree of the target image cube is smaller than a set threshold; and the wordart image channels of the targetimage cube are extracted, and the target wordart image can be generated.

The disclosure features methods that include obtaining multispectral image information for a sample that includes a fluorescent dye, calculating from an image cube of the information a first spectrum and a second spectrum, and calculating a pure spectrum of the fluorescent dye in the sample based on the first and second spectra, where a relative contribution of light emission from the fluorescent dye to the second spectrum is larger than a relative contribution of light emission from the fluorescent dye to the first spectrum, where calculating the first and second spectra includes identifying corresponding first and second sets of pixel intensity values in the image cube, and where identifying the first set of pixel intensity values includes designating one or more layers of the image cube as a first layer set.

The invention discloses a hyperspectral image wave band selection method and system based on neighbor subspace division, and the method comprises the steps: S11, inputting a hyperspectral image cube, calculating a correlation coefficient between any two adjacent wave bands included in the input hyperspectral image cube to obtain a vector of the correlation coefficient; S12, searching for all correlation coefficient extreme points according to the obtained vectors of the correlation coefficients, screening out correlation coefficient minimum points from all the searched correlation coefficient extreme points, and determining an optimally divided hyperspectral band subspace through the screened correlation coefficient minimum points; and S13, sorting the subspaces according to the number of the wavebands in the subspaces, calculating the information entropy of the wavebands in each subspace, and selecting a required number of characteristic wavebands from the subspaces according to the calculated information entropy. According to the invention, it can be ensured that the finally selected wave band contains relatively complete information.

Disclosed are various embodiments for interpretation of flow cytometry data. Flow cytometry data sets are combined to form a multidimensional image cube. The machine learning functions and genetic algorithm interpret the multidimensional image cube to produce a confidence value that is assigned to a function for the targeted condition. The confidence value determines the relative degree to which a targeted condition is present or absent in a flow cytometry data set. Such confidence value can be used to diagnose and interpret results from flow cytometry data.

The invention discloses a general acquisition method for a weak contrast collimation image target center of a comprehensive diagnosis system, which adopts a unified method to acquire optical target centers of different types of images, and is simpler and more practical compared with the existing method. The method mainly comprises the following steps: 1, constructing an original image into a multi-dimensional image cube; 2, performing dimension reduction processing on the multi-dimensional image cube to obtain a first-dimensional image after dimension reduction; 3, performing K-means classification on the first-dimensional image after dimension reduction to obtain a binary image; 4, judging whether the binarized image is a schlieren small ball image or not, and if so, obtaining a schlierensmall ball target separation image; if the image is a pinhole image, feature points on the circular contour of the pinhole image needing to be obtained; if the image is another image, directly acquiring a target center of the image by using a gravity center method; and 5, carrying out circle fitting on the pinhole image and the schlieren small ball image by using a least square method to calculate the circle center of the images so as to obtain the center of the optical target.

The invention relates to the technical field of high-dimensional image processing, in particular to an image semi-supervised classification method and device and a computer readable storage medium. The invention provides an image semi-supervised classification method. The method comprises the following steps: reading hyperspectral image cube data and a training set; rearranging the hyperspectral image cube data into a data matrix; calculating a full connection weight matrix; calculating k neighbors; forcing a sparse connection matrix of the nearest neighbor attribute; constructing a sparse connection matrix of mutual k neighbor attributes; obtaining a sparse connection matrix with complementary neighbor attributes; sparsifying the weight matrix by using a sparse connection matrix with complementary neighbor attributes to obtain a complementary neighbor sparse weight matrix; constructing an initial label matrix; and implementing semi-supervised image classification to obtain a classification result of all pixels of the hyperspectral image cube data.

The invention discloses a multi-image encryption method based on a three-dimensional scrambling model and fractional Fourier transform (FrFT), and belongs to the field of information encryption. At present, the data volume of network information interaction is increasingly increased, and the image batch transmission and storage requirements are increased. The invention provides the multi-image encryption method based on the three-dimensional scrambling model and FrFT in order to protect the security of network transmission of content of multiple images. Under the inspiration of scrambling andrestoring of a three-dimensional magic cube, a plurality of images form an image cube structure, a three-dimensional row, column and layer scrambling mode is designed, and a three-dimensional scrambling model is established. The k original images are scrambled by using the scrambling model, and FrFT is performed on a scrambled image set to generate k encrypted images. Experiments show that the method can encrypt a plurality of interactive images at the same time, and is good in encryption effect, high in safety and efficient.

The invention discloses a multi-image scrambling method based on chaos and block rotation, and belongs to the field of information encryption. At present, network transmission and storage of batch images become increasingly frequent. In order to protect network transmission and storage security of multi-image content, the invention defines concepts of cube rotation and cube revolution, and designs a multi-image scrambling method based on chaos and block rotation. The method comprises the following steps: firstly, combining k original images into an image cube, and carrying out blocking operation to form a cube set; secondly, performing cube rotation and cube revolution on the cubes with the concentrated vertical cubes; thirdly, scrambling the vertical block set by using chaos; and finally, combining the scrambled cubic block sets to generate k scrambled images. Experimental results show that the method can encrypt a plurality of interactive images at the same time, and is good in encryption effect, high in security and efficient.

The invention relates to an image processing technology, and provides a light source determination method and device, electronic equipment and a storage medium. The method comprises the steps that a plurality of gray level images of an object carrying a target defect are obtained, and different gray level images correspond to different light sources; generating a pseudo hyperspectral image cube according to the plurality of grayscale images; determining defect information and spectral wavelength corresponding to the target defect from the pseudo hyperspectral image cube; generating a spectral characteristic curve according to the defect information and the spectral wavelength; and determining a target light source from the spectral characteristic curve according to a hyperspectral image analysis algorithm. The light source selection efficiency can be improved.