CN111462106A - Method for generating tensor for recognizing input of deep learning image and application of tensor - Google Patents
Method for generating tensor for recognizing input of deep learning image and application of tensor Download PDFInfo
- Publication number
- CN111462106A CN111462106A CN202010272361.7A CN202010272361A CN111462106A CN 111462106 A CN111462106 A CN 111462106A CN 202010272361 A CN202010272361 A CN 202010272361A CN 111462106 A CN111462106 A CN 111462106A
- Authority
- CN
- China
- Prior art keywords
- tensor
- image
- input
- deep learning
- generating
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000013135 deep learning Methods 0.000 title claims abstract description 44
- 238000000034 method Methods 0.000 title claims abstract description 37
- 238000001514 detection method Methods 0.000 claims abstract description 22
- 238000005286 illumination Methods 0.000 claims description 23
- 238000012545 processing Methods 0.000 claims description 16
- 238000013528 artificial neural network Methods 0.000 claims description 13
- 238000001228 spectrum Methods 0.000 claims description 12
- 238000001914 filtration Methods 0.000 claims description 8
- 238000003062 neural network model Methods 0.000 claims description 7
- 238000005516 engineering process Methods 0.000 claims description 6
- 230000001131 transforming effect Effects 0.000 claims description 4
- 238000007781 pre-processing Methods 0.000 claims description 3
- 238000004519 manufacturing process Methods 0.000 claims 2
- 238000011897 real-time detection Methods 0.000 abstract description 3
- 238000012797 qualification Methods 0.000 abstract 1
- 230000007547 defect Effects 0.000 description 6
- 230000008569 process Effects 0.000 description 4
- 230000011218 segmentation Effects 0.000 description 4
- 230000008859 change Effects 0.000 description 3
- 238000000605 extraction Methods 0.000 description 3
- 230000005540 biological transmission Effects 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000013527 convolutional neural network Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000009499 grossing Methods 0.000 description 1
- 238000003709 image segmentation Methods 0.000 description 1
- 238000012994 industrial processing Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0004—Industrial image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/70—Denoising; Smoothing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20024—Filtering details
- G06T2207/20032—Median filtering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20048—Transform domain processing
- G06T2207/20056—Discrete and fast Fourier transform, [DFT, FFT]
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Quality & Reliability (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a method for generating a tensor used for recognizing and inputting a deep learning image and application thereof. And arranging the amplitudes from low frequency to high frequency to form a new eigenvector, effectively reducing tensor dimensionality and simultaneously reserving main characteristics to generate tensor input of the surface detection depth learning image algorithm. The method can effectively retain effective information of the image, simultaneously reduce tensor input dimensionality and accelerate the training and reasoning of the surface detection deep learning network, thereby improving the detection speed and precision of the surface detection, realizing real-time detection and improving the qualification rate of target products.
Description
Technical Field
The invention relates to the technical field of image recognition, in particular to a generation method of a tensor used for deep learning image recognition input and application thereof.
Background
The surface defects of industrial products directly affect the quality of product appearance, and the application of surface detection technology in the industrial field is more and more emphasized by people. The traditional surface detection method depends on manual work, but the defects of over-strong labor intensity, low working efficiency, low detection accuracy, easiness in being influenced by subjective factors and the like exist in manual defect detection, so that the replacement of the traditional manual detection by the active defect detection based on machine vision becomes an important trend.
The traditional machine vision surface detection method is based on image segmentation and feature extraction algorithms, but the segmentation and feature extraction algorithms are relatively complex in calculation, the segmentation effect is related to the image, no general segmentation algorithm can realize the segmentation and feature extraction processing of all images, and the real-time detection of the images is usually difficult. In recent years, deep learning is widely applied in the field of machine vision, a deep learning network can learn essential characteristics of data from a large number of samples in a centralized manner, and an image surface defect detection method based on deep learning is also applied and developed.
In order to obtain a high-quality surface image, the stability and reliability of the deep learning surface detection algorithm are improved. The machine vision solution generally adopts illumination of multiple light sources in annular arrangement or structural illumination similar to the mechanism, and images of the same target object in different modes obtained through multiple times of shooting form a high-dimensional image tensor as a network input vector. However, the generated high-dimensional image tensor makes a training data set too large, the difficulty of deep learning network algorithm training is improved, the algorithm operation efficiency is reduced, and the requirement on real-time performance is difficult to meet.
Therefore, a scheme for training the deep learning surface detection algorithm more effectively, reducing the size of the training data set, increasing the inference speed of the deep learning detection algorithm, and realizing the real-time detection of the surface quality is needed.
Disclosure of Invention
The invention aims to provide a generation method for identifying an input tensor by a deep learning image for surface detection and application thereof, which are used for solving the problems of overlarge dimensionality and low detection efficiency of the input image tensor based on a deep learning surface detection method.
In order to achieve the purpose, the invention adopts the following scheme: a generation method of tensor used for deep learning image identification input and application thereof comprise the following steps:
s1: acquiring image tensor data of the surface of a shot target by shooting different illumination modes;
s2: preprocessing the acquired target surface image;
s3: fusing the preprocessed surface images to form a high-dimensional input image tensor;
s4: and carrying out Fourier transform on the obtained high-dimensional input image tensor to produce a new input tensor.
As a preferable embodiment of the present invention, the step S1 specifically includes the following steps:
s 11: illuminating the target surface by arranging a light source structure for annularly arranging multiple light sources for illumination;
s 12: the illumination of the target surface is changed into different modes;
s 13: shooting the same target object in different illumination modes for multiple times to obtain multiple surface images;
s 14: the step s13 is performed in one loop, and the obtained surface image constitutes the image input tensor.
As a preferable embodiment of the present invention, the step S2 specifically includes the following steps:
s 21: converting the acquired surface image into a grayscale image;
s 22: processing the acquired gray level image by adopting a median filtering method, filtering noise by adopting a digital signal processing technology, filtering target surface image data by combining multiple filtering technologies, and denoising noise and non-defect noise introduced in the image data transmission process;
s 23: the surface images acquired in the different illumination modes in step s13 are processed according to steps s21 and s 21.
As a preferable embodiment of the present invention, the step S4 specifically includes the following steps:
s41, arranging the amplitude values of the new input tensor from low frequency to high frequency after Fourier transform as a new input tensor, acquiring an amplitude spectrum corresponding to a frequency domain, and expanding the image tensor with high dimensionality in a third dimension by utilizing a Fourier transform algorithm or performing series expansion in other forms to acquire the frequency spectrum information of the image tensor;
s 42: acquiring frequency distribution information of an image tensor according to the frequency spectrum information, only reserving amplitude information of the image tensor, neglecting phase information of a Fourier coefficient, and reducing redundant image information in the main characteristic of the reserved image tensor; wherein: arranging the amplitude values from low to high frequency { S1,S2,…,SkAnd generating a new input tensor as a new feature expression, wherein the newly generated input image tensor is used for training and reasoning the deep neural network model.
A generation method of a tensor for recognizing an input by a deep learning image is used for collecting a low-dimensional surface image data set.
Further, the gathering of the low-dimensional surface image dataset includes: the method comprises the steps of collecting a plurality of input image data by utilizing an annular arrangement multi-light-source illumination mode, generating a training set of a deep neural network, processing and converting the original image data by the image tensor generation method, generating a new image tensor, and effectively reducing the size of the training data set.
A generation method of a tensor for recognizing an input by a deep learning image produces an application of a tensor for recognizing an input by a deep learning image, which is applied to an inference of a deep neural network of a low-dimensional surface image tensor.
Further, the reasoning of the depth neural network of the low-dimensional surface image tensor comprises: the method comprises the steps of processing and transforming image data acquired in real time to generate a new image tensor, inputting a trained deep neural network for reasoning, and outputting a detection result of the surface quality of a target to obtain the low-dimensionality surface image tensor.
In summary, compared with the prior art, the invention has the beneficial effects that:
according to the method for generating the input tensor of the neural network, the traditional high-dimensionality surface image tensor is input, the input image tensor more suitable for training a deep convolutional neural network model is generated, the deep neural network model is obtained through training, and meanwhile the reasoning speed of a deep learning algorithm is improved. The generated image tensor is efficiently utilized in both training and reasoning of the neural network.
Drawings
Fig. 1 is a flowchart illustrating generation of tensors for recognizing an input in a deep learning image according to an embodiment of the present invention.
Fig. 2 is a flowchart of deep neural network model training for a deep learning image generation tensor according to an embodiment of the present invention.
Fig. 3 is a flowchart illustrating deep learning algorithm inference for generating tensors from deep learning images according to an embodiment of the present invention.
Detailed Description
The following detailed description provides many different embodiments, or examples, for implementing the invention. Of course, these are merely embodiments or examples and are not intended to be limiting. In addition, repeated reference numbers, such as repeated numbers and/or letters, may be used in various embodiments. These iterations are for simplicity and clarity of describing the present invention and are not intended to represent a particular relationship between the various embodiments and/or configurations discussed.
Furthermore, spatially relative terms, such as "below," "lower," "inside-out," "above," "upper," and the like, may be used herein to facilitate describing the relationship of element(s) or feature(s) to other element(s) or feature(s) of the drawings and may include different orientations of the device in use or operation and the orientation depicted in the drawings. The devices may be turned to different orientations (rotated 90 degrees or at other orientations) and the spatially relative descriptors used therein should be interpreted accordingly and are not to be construed as limiting the invention, and the terms "first" and "second" are used for descriptive purposes only and are not intended to indicate or imply relative importance or to implicitly indicate the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature.
The invention will be further described with reference to the following description and embodiments in conjunction with the accompanying drawings:
a method for generating a tensor for recognizing an input by a deep learning image as shown in fig. 1 includes the following steps: .
S1: acquiring image tensor data of the surface of a shot target by shooting different illumination modes; wherein, the step S1 specifically includes the following steps:
s 11: illuminating the target surface by arranging a light source structure for annularly arranging multiple light sources for illumination;
s 12: changing the illumination of the target surface into different illumination modes;
s 13: shooting the same target object in different illumination modes for multiple times to obtain multiple surface images;
s 14: the step s13 is performed in one loop, and the obtained surface image constitutes the image input tensor.
S2: preprocessing the acquired target surface image; wherein the step S2 specifically includes the following steps:
s 21: the acquired surface image is converted into a gray image, for the target surface image, the characteristics of the contour, the gradient and the like of the surface image are paid more attention in the industrial processing process, and in order to improve the processing speed and reduce the interference of a plurality of color channels on the image contour and the gradient characteristics, the acquired color image is converted into the gray image. (ii) a
s 22: the acquired gray level image is processed by adopting a median filtering method, noise is filtered by adopting a digital signal processing technology, and the noise is inevitably introduced in the image acquisition and transmission process in engineering application. For noise, due to the existence of the randomness characteristic, the noise can be filtered by a digital signal processing technology, and then the surface image with noise interference eliminated is subjected to subsequent processing. This step is performed to improve the accuracy of the results obtained in the subsequent steps, and is mainly performed by smoothing the image, which is to eliminate the high-frequency noise in the image acquisition process. Compared with experimental results, the method adopts a median filtering method;
s 23: the surface image acquired in the different illumination modes in step s13 is processed according to steps s21 and s 21.
S3: fusing the preprocessed surface images in different illumination modes to form a high-dimensional input image tensor;
s4: performing fourier transform on the obtained high-dimensional input image tensor to generate a new input tensor, wherein the step S4 specifically includes the following steps:
s41, arranging the amplitude values of the new input tensor from low frequency to high frequency after Fourier transform as a new input tensor, and acquiring an amplitude spectrum corresponding to a frequency domain; the fourier transform is a basic method for signal analysis, and can transform a signal from a time domain to a frequency domain to study the frequency spectrum structure and change rule of the signal. The frequency reflects the change in the signal strength in the spatial domain, the amplitude corresponds to the spatial domain signal contrast, and the phase represents the frequency shift relative to the original signal. The direction of each point in the two-dimensional frequency spectrum is perpendicular to the direction of the change of the image intensity in the space domain. Since the phase spectrum contains little new image structure information, only the amplitude spectrum of the image block is considered in this embodiment.
And converting the high-dimensional image tensor from a space domain to a frequency domain, and performing two-dimensional fast Fourier transform on the image tensor to acquire a magnitude spectrum corresponding to the frequency domain. The fourier transform can be expressed by the following mathematical formula:
eix=cosx+i sinx
in the formula, F is a spatial-domain value and F is a frequency-domain value.
s 42: the frequency distribution information of the image tensor is obtained according to the frequency spectrum information, only the amplitude information of the frequency distribution information is reserved, the phase information of the Fourier coefficient is ignored, and in the image processing process, only the amplitude information is used in the embodiment, because the amplitude image contains almost all the geometric image information needed by the original image, the main characteristics of the image tensor are reserved, and meanwhile, the redundant image information is reduced. Wherein: arranging the amplitudes from low to high frequency (S)1,S2,…,SkAnd generating a new input tensor as a characteristic representation of the new input tensor, wherein the newly generated input image tensor is used for training a preset deep learning neural network model or is used as an input of deep learning network reasoning.
Fig. 2 is a flowchart of a deep neural network model training process of a deep learning image generation tensor according to another embodiment of the present invention, which collects a plurality of input image data by using an illumination mode with multiple light sources arranged in a ring shape to generate a training set of a deep neural network, and performs processing transformation on the original image data by using the image tensor generation method to generate a new image tensor, so as to effectively reduce the size of the training data set and obtain a collection method of a low-dimensional surface image data set;
in this embodiment, an input tensor generation method may be identified through the acquired multiple target image data and the deep learning image of fig. 1, and a new image tensor is acquired to form a training data set for training a preset deep learning network model.
Fig. 3 is a deep learning algorithm inference flow chart of a deep learning image generation tensor according to another embodiment of the present invention, which is a deep learning algorithm inference method for obtaining a low-dimensional surface image tensor by processing and transforming image data acquired in real time to generate a new image tensor, inputting a trained depth neural network for inference, and outputting a detection result of a target surface quality;
in this embodiment, a new input image tensor is generated by using the tensor generation method for recognizing input by using the deep learning image shown in fig. 1, which is acquired in real time, and the trained deep learning neural network model is input, so as to realize real-time target surface quality detection.
While there have been shown and described the fundamental principles and principal features of the invention and advantages thereof, it will be understood by those skilled in the art that the invention is not limited by the embodiments described above, which are given by way of illustration of the principles of the invention, but is susceptible to various changes and modifications without departing from the spirit and scope of the invention as defined by the appended claims. The scope of the invention is defined by the appended claims and equivalents thereof.
Claims (8)
1. A method for generating a tensor for recognizing an input in a deep learning image, comprising the steps of:
s1: acquiring image tensor data of the surface of a shot target by shooting different illumination modes;
s2: preprocessing the acquired target surface image;
s3: fusing the preprocessed surface images to form a high-dimensional input image tensor;
s4: and carrying out Fourier transform on the obtained high-dimensional input image tensor to produce a new input tensor.
2. The method for generating a tensor for deep learning image recognition input according to claim 1, wherein the step S1 specifically includes the following steps:
s 11: illuminating the target surface by arranging a light source structure for annularly arranging multiple light sources for illumination;
s 12: changing the illumination of the target surface into different illumination modes;
s 13: shooting the same target object in different illumination modes for multiple times to obtain multiple surface images;
s 14: the step s13 is performed in one loop, and the obtained surface image constitutes the image input tensor.
3. The method for generating a tensor for deep learning image recognition input as claimed in claim 2, wherein the step S2 specifically includes the following steps:
s 21: converting the acquired surface image into a grayscale image;
s 22: processing the acquired gray level image by adopting a median filtering method, and filtering noise by adopting a digital signal processing technology;
s 23: the surface images acquired in the different illumination modes in step s13 are processed according to steps s21 and s 21.
4. The method for generating a tensor for deep learning image recognition input according to claim 1, wherein the step S4 specifically includes the following steps:
s41, arranging the amplitude values of the new input tensor from low frequency to high frequency after Fourier transform as a new input tensor, and acquiring an amplitude spectrum corresponding to a frequency domain;
s 42: acquiring frequency distribution information of an image tensor according to the frequency spectrum information, only keeping amplitude information of the image tensor, and neglecting phase information of a Fourier coefficient; wherein: arranging the amplitude values from low to high frequency { S1,S2,…,SkAnd generating a new input tensor as a new feature expression for training and reasoning of the deep neural network model.
5. Use of a tensor for deep learning image recognition input produced based on the tensor for deep learning image recognition input production method of any one of claims 1 to 4, characterized in that: the deep learning image identifies the collection of low dimensional surface image datasets to which the tensor used for input applies.
6. Use according to claim 5, characterized in that: the gathering of the low-dimensional surface image dataset includes: the method comprises the steps of collecting a plurality of input image data by utilizing an annular arrangement multi-light-source illumination mode, generating a training set of a deep neural network, processing and transforming the original image data by the image tensor generation method, generating a new image tensor, and effectively reducing the size of the training data set.
7. Use of a tensor for deep learning image recognition input produced based on the tensor for deep learning image recognition input production method of any one of claims 1 to 4, characterized in that: the deep learning image identifies an inference of a deep neural network in which a tensor used for an input is applied to a low-dimensional surface image tensor.
8. Use according to claim 7, characterized in that: the reasoning of the depth neural network of the low-dimensional surface image tensor comprises the following steps: the method comprises the steps of processing and transforming image data acquired in real time to generate a new image tensor, inputting a trained deep neural network for reasoning, outputting a detection result of the surface quality of a target, and obtaining a low-dimensionality surface image tensor deep neural network reasoning method.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010272361.7A CN111462106A (en) | 2020-04-09 | 2020-04-09 | Method for generating tensor for recognizing input of deep learning image and application of tensor |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010272361.7A CN111462106A (en) | 2020-04-09 | 2020-04-09 | Method for generating tensor for recognizing input of deep learning image and application of tensor |
Publications (1)
Publication Number | Publication Date |
---|---|
CN111462106A true CN111462106A (en) | 2020-07-28 |
Family
ID=71680604
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202010272361.7A Pending CN111462106A (en) | 2020-04-09 | 2020-04-09 | Method for generating tensor for recognizing input of deep learning image and application of tensor |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN111462106A (en) |
Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20090276181A1 (en) * | 2008-04-30 | 2009-11-05 | Commissariat A L'energie Atomique | Method of Measuring an Anisotropic Surface Diffusion Tensor or Surface Energy Anisotropies |
US20140226437A1 (en) * | 2011-07-19 | 2014-08-14 | Christopher Chambers | System and method for moment tensor migration imaging |
CN107369155A (en) * | 2017-07-24 | 2017-11-21 | 广东工业大学 | A kind of cloth surface defect detection method and its system based on machine vision |
US20170344880A1 (en) * | 2016-05-24 | 2017-11-30 | Cavium, Inc. | Systems and methods for vectorized fft for multi-dimensional convolution operations |
CN108267311A (en) * | 2018-01-22 | 2018-07-10 | 北京建筑大学 | A kind of mechanical multidimensional big data processing method based on tensor resolution |
CN108305297A (en) * | 2017-12-22 | 2018-07-20 | 上海交通大学 | A kind of image processing method based on multidimensional tensor dictionary learning algorithm |
CN108734187A (en) * | 2017-04-20 | 2018-11-02 | 中山大学 | A kind of multiple view spectral clustering based on tensor singular value decomposition |
CN108873063A (en) * | 2017-05-16 | 2018-11-23 | 大庆市弘洋石油科技开发有限公司 | A kind of method and device of microseism Moment tensor inversion |
-
2020
- 2020-04-09 CN CN202010272361.7A patent/CN111462106A/en active Pending
Patent Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20090276181A1 (en) * | 2008-04-30 | 2009-11-05 | Commissariat A L'energie Atomique | Method of Measuring an Anisotropic Surface Diffusion Tensor or Surface Energy Anisotropies |
US20140226437A1 (en) * | 2011-07-19 | 2014-08-14 | Christopher Chambers | System and method for moment tensor migration imaging |
US20170344880A1 (en) * | 2016-05-24 | 2017-11-30 | Cavium, Inc. | Systems and methods for vectorized fft for multi-dimensional convolution operations |
CN108734187A (en) * | 2017-04-20 | 2018-11-02 | 中山大学 | A kind of multiple view spectral clustering based on tensor singular value decomposition |
CN108873063A (en) * | 2017-05-16 | 2018-11-23 | 大庆市弘洋石油科技开发有限公司 | A kind of method and device of microseism Moment tensor inversion |
CN107369155A (en) * | 2017-07-24 | 2017-11-21 | 广东工业大学 | A kind of cloth surface defect detection method and its system based on machine vision |
CN108305297A (en) * | 2017-12-22 | 2018-07-20 | 上海交通大学 | A kind of image processing method based on multidimensional tensor dictionary learning algorithm |
CN108267311A (en) * | 2018-01-22 | 2018-07-10 | 北京建筑大学 | A kind of mechanical multidimensional big data processing method based on tensor resolution |
Non-Patent Citations (2)
Title |
---|
GRIGORYAN 等: "Tensor transform-based quaternion fourier transform algorithm", 《INFORMATION SCIENCES》 * |
刘彬 等: "运动相关电位分类算法比较和语义范式分析", 《计算机与现代化》 * |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Wang et al. | A new intelligent bearing fault diagnosis method using SDP representation and SE-CNN | |
Bruno et al. | Fractal dimension applied to plant identification | |
CN109918417B (en) | Time sequence data self-adaptive segmentation, dimension reduction and characterization method based on wavelet transformation and application | |
CN112288645B (en) | Skull face restoration model construction method and restoration method and system | |
CN109711411B (en) | Image segmentation and identification method based on capsule neurons | |
CN111858997B (en) | Cross-domain matching-based clothing template generation method | |
CN105488512A (en) | Sift feature matching and shape context based test paper inspection method | |
Zenkova et al. | Pseudo-phase mapping of speckle fields using 2D Hilbert transformation | |
Liang et al. | Automatic defect detection of texture surface with an efficient texture removal network | |
CN111598822A (en) | Image fusion method based on GFRW and ISCM | |
Zuluaga et al. | Blind microscopy image denoising with a deep residual and multiscale encoder/decoder network | |
CN111462106A (en) | Method for generating tensor for recognizing input of deep learning image and application of tensor | |
Chen et al. | Research on pupil center location based on improved Hough transform and edge gradient algorithm | |
CN111739047A (en) | Tongue image segmentation method and system based on bispectrum reconstruction | |
CN108876711B (en) | Sketch generation method, server and system based on image feature points | |
Shaheed et al. | Novel image quality assessment and enhancement techniques for finger vein recognition | |
Liu et al. | Graph representation learning for spatial image steganalysis | |
Li et al. | Face aging effect simulation model based on multilayer representation and shearlet transform | |
Tabatabaei et al. | One dimensional second order derivative local binary pattern for hand gestures classification using sEMG signals | |
CN112529081A (en) | Real-time semantic segmentation method based on efficient attention calibration | |
Xie et al. | Retinopathy diagnosis using semi-supervised multi-channel generative adversarial network | |
Xie et al. | Transferring deep gaussian denoiser for compressed sensing MRI reconstruction | |
CN113936165B (en) | CT image processing method, terminal and computer storage medium | |
CN114120011B (en) | Method for improving image matching precision and speed by using group intelligent algorithm | |
Déau et al. | PREFAB-GEN: AD HOC Image Generation for Pre-Manufacturing of Tires Using Image-To-Image Translation |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination |