CN117078668B - Chromosome segmentation method based on large-size image - Google Patents

Chromosome segmentation method based on large-size image Download PDF

Info

Publication number
CN117078668B
CN117078668B CN202311323085.2A CN202311323085A CN117078668B CN 117078668 B CN117078668 B CN 117078668B CN 202311323085 A CN202311323085 A CN 202311323085A CN 117078668 B CN117078668 B CN 117078668B
Authority
CN
China
Prior art keywords
image
network
chromosome
subgraph
size
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.)
Active
Application number
CN202311323085.2A
Other languages
Chinese (zh)
Other versions
CN117078668A (en
Inventor
李娜
牛兵
朱春阳
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Xiaona Technology Suzhou Co ltd
Original Assignee
Xiaona Technology Suzhou Co ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Xiaona Technology Suzhou Co ltd filed Critical Xiaona Technology Suzhou Co ltd
Priority to CN202311323085.2A priority Critical patent/CN117078668B/en
Publication of CN117078668A publication Critical patent/CN117078668A/en
Application granted granted Critical
Publication of CN117078668B publication Critical patent/CN117078668B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

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
    • G06T7/0012—Biomedical 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
    • G06N3/0464—Convolutional networks [CNN, ConvNet]
    • 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
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N5/00—Computing arrangements using knowledge-based models
    • G06N5/04—Inference or reasoning models
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/20—Image preprocessing
    • G06V10/26—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
    • 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
    • G06T2207/20081—Training; Learning
    • 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
    • G06T2207/20084—Artificial neural networks [ANN]
    • 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
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30004—Biomedical image processing
    • G06T2207/30024—Cell structures in vitro; Tissue sections in vitro

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • General Health & Medical Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Software Systems (AREA)
  • Evolutionary Computation (AREA)
  • Computing Systems (AREA)
  • Artificial Intelligence (AREA)
  • Data Mining & Analysis (AREA)
  • Computational Linguistics (AREA)
  • Mathematical Physics (AREA)
  • General Engineering & Computer Science (AREA)
  • Biophysics (AREA)
  • Medical Informatics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Biomedical Technology (AREA)
  • Multimedia (AREA)
  • Molecular Biology (AREA)
  • Radiology & Medical Imaging (AREA)
  • Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
  • Databases & Information Systems (AREA)
  • Quality & Reliability (AREA)
  • Image Analysis (AREA)
  • Investigating Or Analysing Biological Materials (AREA)

Abstract

The invention discloses a chromosome segmentation method based on a large-size image, which comprises a training process and an reasoning process, wherein the reasoning process comprises the following steps: taking subgraphs according to a set step length; classifying the subgraph through a classification network; the classification network is a dense convolutional neural network, wherein a full-connection layer is replaced by a trainable tensor regression layer; and segmenting the enhanced subgraph through a segmentation network to obtain an independent image of the complete chromosome. The invention has the advantages that: by adopting a deep learning technology, chromosomes can be automatically and accurately identified and segmented from a large-size image by using a cut-off graph, a classification model and a segmentation model; the tensor regression layer is used for replacing the full connection layer in the classification model, so that the training speed is remarkably improved, and the consumption of computing resources and time is reduced.

Description

Chromosome segmentation method based on large-size image
Technical Field
The invention relates to the field of artificial intelligence, in particular to a chromosome segmentation method based on a large-size image.
Background
In biomedical research, chromosome segmentation is a key step in identifying, analyzing chromosome structure and function. Traditional chromosome segmentation methods are often based on steps of image preprocessing, threshold setting, region growing and the like, and the methods have unsatisfactory segmentation effects on large-size images and are easily influenced by factors such as image noise, illumination unevenness and the like.
In recent years, deep learning has achieved remarkable results in the field of image processing, and particularly convolutional neural networks exhibit excellent performance in terms of image classification, object detection, semantic segmentation, and the like. However, the existing deep learning chromosome segmentation method mainly aims at small-size images, and when the method is applied to large-size images, the problems of low calculation efficiency, high memory consumption and the like exist.
Disclosure of Invention
The invention aims to provide a chromosome segmentation method based on a large-size image, which solves or partially solves the technical problems.
In order to achieve the above purpose, the present invention provides the following technical solutions: a chromosome segmentation method based on a large-size image, comprising a training process and an reasoning process, wherein the reasoning process comprises the following steps:
s11, determining the size of the subgraph according to a preset size, and taking the size as the size of a drawing frame;
s12, sequentially moving the picture taking frames from the upper left corner of the input image according to a set step length, taking the image covered by the picture taking frames at each step as a sub-image, wherein the step length is required to ensure that adjacent sub-images have an overlapping area, and each chromosome in the input image is at least completely appeared on one of the sub-images;
s13, classifying the subgraph through a classification network, judging whether a complete chromosome exists in the subgraph, classifying the subgraph with the complete chromosome into positive classes, and equally dividing the other subgraphs into negative classes; the classification network is a dense convolutional neural network, wherein a full-connection layer is replaced by a trainable tensor regression layer; function of the tensor regression layerThe method comprises the following steps:wherein->Is tensor->Is spread by the modulus n+1,/->Is the weight tensor, and +.>，/>For the number of categories of the dataset, +.>Representing a space; />Representation->Vectorization of->Is the N-order eigenvector of the convolutional layer output, and +.>；/>N is the order for the bias vector;
s14, carrying out image enhancement on the subgraphs divided into positive classes through an image enhancement network so as to strengthen the characteristics and details of chromosomes;
s15, dividing the subgraph reinforced by the reinforcing network through a dividing network to obtain an independent image of a complete chromosome;
s16, merging the segmentation results of the subgraphs, and splicing or superposing the subgraphs according to the position information of each subgraph in the original graph to obtain the segmentation result of the complete image;
the training process comprises the following steps:
s21, manually marking the input image, and marking the outline of a chromosome in the input image;
s22, determining the size of the subgraph according to a preset size, and taking the size as the size of the drawing frame;
s23, sequentially moving the picture taking frames from the upper left corner of the input image according to a set step length, taking the image covered by the picture taking frames at each step as a sub-image, wherein the step length is required to ensure that adjacent sub-images have an overlapping area, and each chromosome in the input image is at least completely appeared on one of the sub-images;
s24, classifying the subgraph through a classification network, judging whether a complete chromosome exists in the subgraph, classifying the subgraph with the complete chromosome into positive classes, and equally dividing the other subgraphs into negative classes;
s25, carrying out image enhancement on the subgraphs divided into positive classes through an image enhancement network so as to strengthen the characteristics and details of chromosomes; the image enhancement network is trained by: downsampling the subgraph divided into positive classes to obtain a downsampled image of low resolution; training the image enhancement network by taking the downsampled image and the corresponding subgraph as paired data sets; in the training process, the pre-training weight is used for fine adjustment of the image enhancement network so as to improve the image enhancement effect;
s26, segmenting the subgraph after image enhancement of the image enhancement network through a segmentation network, and obtaining an independent image of a complete chromosome in the subgraph.
Preferably, the split network is a Yolov8 network comprising SimAM modules and a Dice loss function; wherein the SimAM module adds a parameter-less attention mechanism.
Preferably, the input/output formula of the SimAM module is as follows:wherein->For outputting (I)>For input, energy matrix->Is energy +.>Summarizing all channels and dimensions; energy->The method meets the following conditions:wherein, mean->Variance->，/>For input +.>Element of (a)>Representing neurons,/->For the balance coefficient, M is the total number of elements in X, and i is a variable.
Preferably, the Dice loss functionThe solution formula of (2) is: />Wherein->For a real bounding box->Is a predicted bounding box.
Preferably, the tensor regression layer applies a low rank decomposition to the weight tensor to reduce the number of training parameters.
PreferablyThe tensor regression layer applies CP decomposition to the weight tensor, its functionThe method comprises the following steps:wherein->Is tensorFactor matrix of CP decomposition of ∈,>KR product.
Preferably, the enhancement network is an enhanced super-resolution network, and features and details of the chromosome are enhanced to improve performance of the segmentation model.
Preferably, the step length satisfies that the overlapping area of the adjacent subgraphs is greater than or equal to 20% of the area of the subgraph.
The invention has the advantages that:
by adopting a deep learning technology, chromosomes can be automatically and accurately identified and segmented from a large-size image by using a cut-off graph, a classification model and a segmentation model;
the segmentation result is further optimized through the image enhancement module and the modification of the segmentation network structure and the loss function, and the accuracy of chromosome segmentation is improved;
the system is flexible in design, and can be optimized and adjusted for different types of large-size chromosome images;
the tensor regression layer is used for replacing the full connection layer in the classification model, so that the training speed is remarkably improved, and the consumption of computing resources and time is reduced.
Drawings
FIG. 1 is a schematic diagram of the principles of the present invention;
FIG. 2 is a tensor network structure based on a dense convolutional neural network;
FIG. 3 is a schematic diagram of a compact block.
Detailed Description
The technical scheme of the invention is further described in detail below with reference to the specific embodiments.
In the description of the present invention, it should be noted that the positional or positional relationship indicated by the terms such as "inner", "outer", "upper", "lower", "horizontal", etc. are based on the positional or positional relationship shown in the drawings, are merely for convenience of describing the present invention and simplifying the description, and do not indicate or imply that the apparatus or element in question must have a specific orientation, be constructed and operated in a specific orientation, and thus should not be construed as limiting the present invention.
As shown in fig. 1, the chromosome segmentation method based on the large-size image comprises a training process and an reasoning process, wherein the reasoning process comprises the following steps:
s11, determining the size of the subgraph according to a preset size, and taking the size as the size of a drawing frame;
s12, sequentially moving the picture taking frames from the upper left corner of the input image according to a set step length, taking the image covered by each picture taking frame as a sub-image, wherein the step length needs to ensure that adjacent sub-images have an overlapping area, and each chromosome in the input image at least completely appears on one sub-image;
s13, classifying the subgraphs through a classification network, judging whether complete chromosomes exist in the subgraphs, classifying the subgraphs with the complete chromosomes into positive classes, and equally dividing the subgraphs with the complete chromosomes into negative classes; the classification network is a dense convolutional neural network, wherein the full-connection layer is replaced by a trainable tensor regression layer; function of tensor regression layerThe method comprises the following steps: />Wherein->Is tensor->Modulo n+1 expansion of (2)，/>Is the weight tensor, and +.>，/>For the number of categories of the dataset, +.>Representing a space; />Representation->Vectorization of->Is the N-order eigenvector of the convolutional layer output, and +.>；/>Is a bias vector;
s14, performing image enhancement on the subgraphs divided into positive classes through an image enhancement network so as to strengthen the characteristics and details of chromosomes;
s15, dividing the subgraph reinforced by the reinforcing network through the dividing network to obtain an independent image of a complete chromosome, wherein chromosome information obtained by dividing is obtained;
s16, merging the segmentation results of the subgraphs, and splicing or superposing the subgraphs according to the position information of each subgraph in the original graph to obtain the segmentation result of the complete image, namely the chromosome information obtained by segmentation;
the training process comprises the following steps:
s21, manually marking the input image, and marking chromosomes and contours in the input image;
s22, determining the size of the subgraph according to a preset size, and taking the size as the size of the drawing frame;
s23, sequentially moving the picture taking frames from the upper left corner of the input image according to a set step length, taking the image covered by each picture taking frame as a sub-image, wherein the step length is required to ensure that adjacent sub-images have an overlapping area, and each chromosome in the input image at least completely appears on one sub-image;
s24, classifying the subgraphs through a classification network, judging whether complete chromosomes exist in the subgraphs, classifying the subgraphs with the complete chromosomes into positive classes, and equally dividing the subgraphs with the complete chromosomes into negative classes; the input of the subsequent steps can be effectively reduced, and the efficiency is improved;
s25, performing image enhancement on the subgraphs divided into positive classes through an image enhancement network so as to strengthen the characteristics and details of chromosomes; the image enhancement network is trained by: downsampling the subgraphs divided into positive classes to obtain a downsampled image of low resolution; taking the downsampled image and the corresponding subgraph as a paired data set training image enhancement network; in the training process, the pre-training weight is used for fine adjustment of the image enhancement network so as to improve the image enhancement effect;
s26, dividing the subgraph after image enhancement of the image enhancement network through a division network to obtain an independent image of a complete chromosome.
The split network is a Yolov8 network comprising a SimAM module and a Dice loss function; wherein the SimAM module adds a parameter-less attention mechanism.
The input and output formulas of the SimAM module are as follows:wherein->For outputting (I)>For input, energy matrix->Is energy +.>Summarizing all channels and dimensions; energy->The method meets the following conditions: />Wherein, mean->Variance->，/>For input +.>Element of (a)>Representing neurons,/->For the balance coefficient, M is the total number of elements in X, and i is a variable.
Dice loss functionThe solution formula of (2) is: />Wherein->For a real bounding box->Is a predicted bounding box.
The invention uses tensor network layer to replace the full connection layer in the classification network, thereby reducing the parameters of the classification networkA number; the problem of multi-mode structure information loss caused by flattening the characteristic tensor at the full-connection layer can be effectively solved, and the memory of a computer can be greatly saved while the performance of a model is ensured. The tensor regression layer applies a low rank decomposition to the weight tensors to reduce the number of training parameters. Specifically, the tensor regression layer applies a CP decomposition to the weight tensor, its functionThe method comprises the following steps:wherein->Is tensorFactor matrix of CP decomposition of ∈,>KR product.
Further, the enhancement network is an enhanced super-resolution network, and features and details of the chromosome are enhanced to improve the performance of the segmentation model.
Further, the step length satisfies that the overlapping area of the adjacent subgraphs is larger than or equal to 20% of the area of the subgraphs, and the practice proves that for common chromosome images, the setting can ensure that each chromosome can appear in at least one subgraph.
As shown in fig. 2, a tensor network structure based on a dense convolutional neural network (densnet) is shown.
The original structure of the DenseNet network is that after the characteristics are extracted, the obtained output is subjected to global average pooling once, and then the full connection layer is connected to classify the image data. Such global averaging pooling destroys the original structure of the data, loses the spatial structure information of the features, and the subsequent full-connection layer is also provided with a large number of trainable parameters, and also increases the difficulty in training the model. In this work, therefore, the present invention wishes to be able to optimize the structure of the DenseNet network, embedding a tensor regression layer into the network as a trainable layer of the model, allowing classification of data by joint learning features. Therefore, the invention directly replaces the global average pooling and full connection layer of the original network with the tensor regression layer, and applies low-rank constraint to the regression weight. Intuitively, the advantage of the tensor regression layer comes from the ability to effectively use the spatial structure information of the data and greatly reduce the training parameters of the model.
Dense block (DenseBlock) the deep convolutional network typically reduces the feature map size by pooling or convolutional layers, but dense connectivity of DenseNet requires the feature map size to be consistent. To solve this problem, dense blocks are defined in which feature patterns of each layer are uniform in size, connected on channels, and the number of channels to which the input feature patterns are added isThen->The number of channels of the layer is->Here +.>Which is a super parameter, becomes a growth rate in DensNet. Fig. 3 shows a dense block with a layer number of 4 and a growth rate of 3.
Transition layer (Transition): the transition layer is mainly to connect two adjacent dense blocks and reduce the feature map size. The transition layer comprises a 1x1 convolution kernel and a 2x2 global average pooling layer, which plays the role of a compression model.
The tensor network layer is a trainable tensor regression network.
The foregoing description is only illustrative of the present invention and is not intended to limit the scope of the invention, and all equivalent structures or equivalent processes or direct or indirect application in other related technical fields are included in the scope of the present invention.

Claims (8)

1. A chromosome segmentation method based on a large-size image, comprising a training process and an reasoning process, wherein the reasoning process comprises the following steps:
s11, determining the size of the subgraph according to a preset size, and taking the size as the size of a drawing frame;
s12, sequentially moving the picture taking frames from the upper left corner of the input image according to a set step length, taking the image covered by the picture taking frames at each step as a sub-image, wherein the step length is required to ensure that adjacent sub-images have an overlapping area, and each chromosome in the input image is at least completely appeared on one of the sub-images;
s13, classifying the subgraph through a classification network, judging whether a complete chromosome exists in the subgraph, classifying the subgraph with the complete chromosome into positive classes, and equally dividing the other subgraphs into negative classes; the classification network is a dense convolutional neural network, wherein a full-connection layer is replaced by a trainable tensor regression layer; function of the tensor regression layerThe method comprises the following steps:
wherein->Is tensor->Is spread by the modulus n+1,/->Is the weight tensor, and +.>，/>For the number of categories of the dataset, +.>Representing a space; />Representation->Vectorization of->Is the N-order eigenvector of the convolutional layer output, and +.>；/>N is the order for the bias vector;
s14, carrying out image enhancement on the subgraphs divided into positive classes through an image enhancement network so as to strengthen the characteristics and details of chromosomes;
s15, dividing the subgraph reinforced by the reinforcing network through a dividing network to obtain an independent image of a complete chromosome;
s16, merging the segmentation results of the subgraphs, and splicing or superposing the subgraphs according to the position information of each subgraph in the original graph to obtain the segmentation result of the complete image;
the training process comprises the following steps:
s21, manually marking the input image, and marking the outline of a chromosome in the input image;
s22, determining the size of the subgraph according to a preset size, and taking the size as the size of the drawing frame;
s23, sequentially moving the picture taking frames from the upper left corner of the input image according to a set step length, taking the image covered by the picture taking frames at each step as a sub-image, wherein the step length is required to ensure that adjacent sub-images have an overlapping area, and each chromosome in the input image is at least completely appeared on one of the sub-images;
s24, classifying the subgraph through a classification network, judging whether a complete chromosome exists in the subgraph, classifying the subgraph with the complete chromosome into positive classes, and equally dividing the other subgraphs into negative classes;
s25, carrying out image enhancement on the subgraphs divided into positive classes through an image enhancement network so as to strengthen the characteristics and details of chromosomes; the image enhancement network is trained by: downsampling the subgraph divided into positive classes to obtain a downsampled image of low resolution; training the image enhancement network by taking the downsampled image and the corresponding subgraph as paired data sets; in the training process, the pre-training weight is used for fine adjustment of the image enhancement network so as to improve the image enhancement effect;
s26, segmenting the subgraph after image enhancement of the image enhancement network through a segmentation network, and obtaining an independent image of a complete chromosome in the subgraph.
2. The large-size image-based chromosome segmentation method according to claim 1, wherein: the split network is a Yolov8 network comprising a SimAM module and a Dice loss function; wherein the SimAM module adds a parameter-less attention mechanism.
3. The large-size image-based chromosome segmentation method according to claim 2, wherein: the input and output formulas of the SimAM module are as follows:
wherein->For outputting (I)>For input, energy matrix->Is energy +.>Summarizing all channels and dimensions; energy->The method meets the following conditions:
wherein, mean->Variance->，/>For input +.>Element of (a)>Representing neurons,/->For the balance coefficient, M is the total number of elements in X, and i is a variable.
4. The large-size image-based chromosome segmentation method according to claim 2, wherein: the Dice loss functionThe solution formula of (2) is:
wherein->For a real bounding box->Is a predicted bounding box.
5. The large-size image-based chromosome segmentation method according to claim 1, wherein: the tensor regression layer applies a low rank decomposition to the weight tensors to reduce the number of training parameters.
6. The large-size image-based chromosome segmentation method as set forth in claim 5, wherein: the tensor regression layer applies CP decomposition to the weight tensor, its functionThe method comprises the following steps:
wherein->Tensor>Factor matrix of CP decomposition of ∈,>KR product.
7. The large-size image-based chromosome segmentation method according to claim 1, wherein: the enhancement network is an enhanced super-resolution network, and the characteristics and details of the chromosome are enhanced to improve the performance of the segmentation model.
8. The large-size image-based chromosome segmentation method according to claim 1, wherein: the step length satisfies that the overlapping area of the adjacent subgraphs is more than or equal to 20% of the area of the subgraphs.
CN202311323085.2A 2023-10-13 2023-10-13 Chromosome segmentation method based on large-size image Active CN117078668B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202311323085.2A CN117078668B (en) 2023-10-13 2023-10-13 Chromosome segmentation method based on large-size image

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202311323085.2A CN117078668B (en) 2023-10-13 2023-10-13 Chromosome segmentation method based on large-size image

Publications (2)

Publication Number Publication Date
CN117078668A CN117078668A (en) 2023-11-17
CN117078668B true CN117078668B (en) 2024-02-20

Family

ID=88702817

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202311323085.2A Active CN117078668B (en) 2023-10-13 2023-10-13 Chromosome segmentation method based on large-size image

Country Status (1)

Country Link
CN (1) CN117078668B (en)

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112215847A (en) * 2020-09-30 2021-01-12 武汉大学 Method for automatically segmenting overlapped chromosomes based on counterstudy multi-scale features
CN112365482A (en) * 2020-11-16 2021-02-12 上海北昂医药科技股份有限公司 Crossed chromosome image example segmentation method based on chromosome trisection feature point positioning
CN115601374A (en) * 2019-08-22 2023-01-13 杭州德适生物科技有限公司（Cn） Chromosome image segmentation method
CN116051896A (en) * 2023-01-28 2023-05-02 西南交通大学 Hyperspectral image classification method of lightweight mixed tensor neural network
CN116485767A (en) * 2023-04-26 2023-07-25 长安大学 Pavement crack image detection method and system based on image classification and segmentation

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115601374A (en) * 2019-08-22 2023-01-13 杭州德适生物科技有限公司（Cn） Chromosome image segmentation method
CN112215847A (en) * 2020-09-30 2021-01-12 武汉大学 Method for automatically segmenting overlapped chromosomes based on counterstudy multi-scale features
CN112365482A (en) * 2020-11-16 2021-02-12 上海北昂医药科技股份有限公司 Crossed chromosome image example segmentation method based on chromosome trisection feature point positioning
CN116051896A (en) * 2023-01-28 2023-05-02 西南交通大学 Hyperspectral image classification method of lightweight mixed tensor neural network
CN116485767A (en) * 2023-04-26 2023-07-25 长安大学 Pavement crack image detection method and system based on image classification and segmentation

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
AS-PANet:改进路径增强网络的重叠染色体实例分割;林成创;赵淦森;尹爱华;丁笔超;郭莉;陈汉彪;;中国图象图形学报(第10期);全文 *

Also Published As

Publication number Publication date
CN117078668A (en) 2023-11-17

Similar Documents

Publication Publication Date Title
CN111768432B (en) Moving target segmentation method and system based on twin deep neural network
Luo et al. Fire smoke detection algorithm based on motion characteristic and convolutional neural networks
CN108830855B (en) Full convolution network semantic segmentation method based on multi-scale low-level feature fusion
CN109493350A (en) Portrait dividing method and device
Zhou et al. YOLO-CIR: The network based on YOLO and ConvNeXt for infrared object detection
CN109886159B (en) Face detection method under non-limited condition
Liu et al. Smooth filtering identification based on convolutional neural networks
WO2023036157A1 (en) Self-supervised spatiotemporal representation learning by exploring video continuity
CN115631369A (en) Fine-grained image classification method based on convolutional neural network
CN114627154B (en) Target tracking method deployed in frequency domain, electronic equipment and storage medium
Xiang et al. Crowd density estimation method using deep learning for passenger flow detection system in exhibition center
Sun et al. Image semantic segmentation for autonomous driving based on improved U-Net
Wang et al. CDFF: a fast and highly accurate method for recognizing traffic signs
CN118212572A (en) Road damage detection method based on improvement YOLOv7
Huynh et al. An efficient model for copy-move image forgery detection
Luo et al. RBD-Net: robust breakage detection algorithm for industrial leather
CN117078668B (en) Chromosome segmentation method based on large-size image
Li et al. An efficient method for DPM code localization based on depthwise separable convolution
Özyurt et al. A new method for classification of images using convolutional neural network based on Dwt-Svd perceptual hash function
Zheng et al. A novel semantic segmentation algorithm for RGB-D images based on non-symmetry and anti-packing pattern representation model
CN111008986B (en) Remote sensing image segmentation method based on multitasking semi-convolution
CN112926670A (en) Garbage classification system and method based on transfer learning
Li et al. A fish image segmentation methodology in aquaculture environment based on multi-feature fusion model
Mukherjee et al. CNN-based real-time parameter tuning for optimizing denoising filter performance
Lipetski et al. A combined HOG and deep convolution network cascade for pedestrian detection

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
GR01 Patent grant
GR01 Patent grant