CN111611893A - Intelligent measuring and judging method applying neural network deep learning - Google Patents
Intelligent measuring and judging method applying neural network deep learning Download PDFInfo
- Publication number
- CN111611893A CN111611893A CN202010407547.9A CN202010407547A CN111611893A CN 111611893 A CN111611893 A CN 111611893A CN 202010407547 A CN202010407547 A CN 202010407547A CN 111611893 A CN111611893 A CN 111611893A
- Authority
- CN
- China
- Prior art keywords
- neural network
- network layer
- information
- deep learning
- value
- 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.)
- Granted
Links
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 133
- 238000000034 method Methods 0.000 title claims abstract description 89
- 238000013135 deep learning Methods 0.000 title claims abstract description 62
- 238000012544 monitoring process Methods 0.000 claims abstract description 34
- 210000002364 input neuron Anatomy 0.000 claims abstract description 23
- 210000002569 neuron Anatomy 0.000 claims abstract description 16
- 210000004205 output neuron Anatomy 0.000 claims abstract description 16
- 238000004364 calculation method Methods 0.000 claims description 29
- 238000004422 calculation algorithm Methods 0.000 claims description 18
- 230000008569 process Effects 0.000 claims description 15
- 230000032683 aging Effects 0.000 claims description 11
- 238000005259 measurement Methods 0.000 claims description 11
- 201000010099 disease Diseases 0.000 claims description 10
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 claims description 10
- 230000001149 cognitive effect Effects 0.000 claims description 4
- 230000003902 lesion Effects 0.000 claims description 4
- 230000002250 progressing effect Effects 0.000 claims description 4
- 230000036541 health Effects 0.000 claims description 3
- 230000000750 progressive effect Effects 0.000 claims description 3
- 238000012163 sequencing technique Methods 0.000 claims description 3
- 238000012360 testing method Methods 0.000 claims description 3
- 230000002035 prolonged effect Effects 0.000 claims 1
- 238000011161 development Methods 0.000 description 6
- 230000018109 developmental process Effects 0.000 description 6
- 238000010801 machine learning Methods 0.000 description 4
- 238000013473 artificial intelligence Methods 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 230000006870 function Effects 0.000 description 3
- 210000001015 abdomen Anatomy 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 210000000038 chest Anatomy 0.000 description 2
- 230000010485 coping Effects 0.000 description 2
- 238000009795 derivation Methods 0.000 description 2
- 230000006866 deterioration Effects 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 230000035772 mutation Effects 0.000 description 2
- 208000024891 symptom Diseases 0.000 description 2
- 210000000115 thoracic cavity Anatomy 0.000 description 2
- 238000007792 addition Methods 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000019771 cognition Effects 0.000 description 1
- 238000007418 data mining Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000003631 expected effect Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000003058 natural language processing Methods 0.000 description 1
- 238000003909 pattern recognition Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000013519 translation Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/40—Scenes; Scene-specific elements in video content
- G06V20/41—Higher-level, semantic clustering, classification or understanding of video scenes, e.g. detection, labelling or Markovian modelling of sport events or news items
-
- 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/047—Probabilistic or stochastic 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
- G06N3/06—Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons
- G06N3/061—Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons using biological neurons, e.g. biological neurons connected to an integrated circuit
-
- 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
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/52—Surveillance or monitoring of activities, e.g. for recognising suspicious objects
- G06V20/53—Recognition of crowd images, e.g. recognition of crowd congestion
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
- G16H50/20—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for computer-aided diagnosis, e.g. based on medical expert systems
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biophysics (AREA)
- General Physics & Mathematics (AREA)
- Software Systems (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Evolutionary Computation (AREA)
- Mathematical Physics (AREA)
- Artificial Intelligence (AREA)
- Medical Informatics (AREA)
- Multimedia (AREA)
- Neurology (AREA)
- Public Health (AREA)
- Microelectronics & Electronic Packaging (AREA)
- Databases & Information Systems (AREA)
- Probability & Statistics with Applications (AREA)
- Pathology (AREA)
- Epidemiology (AREA)
- Primary Health Care (AREA)
- Image Analysis (AREA)
Abstract
The invention provides an intelligent measuring and judging method applying neural network deep learning, wherein the neural network comprises an input neural network layer and an output neural network layer, a plurality of layers of hidden neural network layers presenting arrangement exist between the input neural network layer and the output neural network layer, the input neural network layer comprises a plurality of input neurons, the hidden neural network layer comprises a plurality of hidden neurons, the output neural network layer comprises a plurality of output neurons, and the intelligent measuring and judging method comprises the following contents: A. a deep learning method; B. a measuring and judging method. The invention omits a large amount of labels on the target information value, and can reduce the labor burden to the maximum extent. And the characteristic information is integrally identified, so that the monitoring of all target information values is completed, the complexity of creating a plurality of identifiers is saved, and the problem of target information value omission is avoided. The special recognizer of the specific target information value can be learned and trained, and the corresponding field is continuously expanded. And is continuously evolving into a higher precision, larger scale system.
Description
Technical Field
The invention belongs to the technical field of computer artificial intelligence, relates to an algorithm for neural network deep learning, and particularly relates to an intelligent testing and judging method applying neural network deep learning.
Background
Deep learning is a new research direction in the field of machine learning, and is introduced into machine learning to make it closer to the original target, artificial intelligence.
Deep learning is the intrinsic law and expression level of the learning sample data, and the information obtained in the learning process is very helpful for the interpretation of data such as characters, images and sounds. The final aim of the method is to enable the machine to have the analysis and learning capability like a human, and to recognize data such as characters, images and sounds. Deep learning is a complex machine learning algorithm, and achieves the effect in speech and image recognition far exceeding the prior related art. Deep learning has achieved many achievements in search technology, data mining, machine learning, machine translation, natural language processing, multimedia learning, speech, recommendation and personalization technologies, and other related fields. The deep learning enables the machine to imitate human activities such as audio-visual and thinking, solves a plurality of complex pattern recognition problems, and makes great progress on the artificial intelligence related technology.
In the prior art, deep learning is used for identification and judgment of human body scanning images in the medical field. However, the existing calculation method for deep learning of the neural network has the following disadvantages:
1. in order to learn, a large amount of learning data is needed, and the learning data needs a large amount of labels of professional medical personnel, so that time and labor are wasted, and effective medical resources are wasted;
2. the object of recognition is an image, and the time-series data hardly correspond to each other. In the current method, the learned time-series data is learned by cutting out a part of all data at a fixed length. However, in many medical images, a body area to be scanned is defined, and the number of scans differs depending on a doctor and a machine. That is, the first frame and the last frame must be fixed, and each of them must be matched with learning data (frame positions of time-series data).
3. One recognizer can only correspond to one disease, a plurality of recognizers are manufactured for recognizing a plurality of diseases, and the number of cases in the recognizers needs to be consistent, otherwise, the problem of heavy recognition is generated, and the serious consequence of focus omission exists.
4. In the medical field, the site of pain symptoms and the actual site of onset are often different from each other, and as a result, the doctor needs to confirm all the positions, and the expected effect is not obtained.
5. The existing calculation method for deep learning of the neural network cannot effectively monitor and judge the continuity information with the logical relationship, and only can learn, record and compare the independent information values.
Disclosure of Invention
The invention aims to solve the problems in the prior art, and provides an intelligent measuring and judging method for deep learning by applying a neural network, which predicts a development situation by using an evolution logic after learning and recognizing the evolution logic and obtains a judgment conclusion by comparing the development situation with an actual development situation.
The purpose of the invention can be realized by the following technical scheme: an intelligent measuring and judging method applying neural network deep learning, wherein the neural network comprises an input neural network layer and an output neural network layer, a plurality of layers of hidden neural network layers presenting arrangement exist between the input neural network layer and the output neural network layer, the input neural network layer comprises a plurality of input neurons, the hidden neural network layers comprise a plurality of hidden neurons, the output neural network layer comprises a plurality of output neurons, and the intelligent measuring and judging method comprises the following contents:
A. the deep learning method comprises the following steps:
1) the number of hidden neural network layers is initially set, and the number of hidden neurons in each hidden neural network layer is set; the number of output neurons in an output neural network layer is initially set;
2) collecting a large number of information groups serving as reference values, wherein each information group comprises a plurality of continuity information values presented into logic, and respectively inputting the continuity information values of each information group into an input neural network layer to enable logic calculation between two adjacent information values to form an input neuron; carrying out parallel computation on each input neuron of an input neural network layer to obtain each hidden neuron of a first hidden neural network layer, and obtaining n +1 hidden neural network layers step by step through computation until finally obtaining a conclusion probability value of each output neuron in an output neural network layer;
3) in the step 2), a forward propagation algorithm is used for deduction, a modified weight is obtained through a backward propagation algorithm, a conclusion value meeting the learning expectation is obtained by combining the forward propagation algorithm and the backward propagation algorithm, and the evolution process of the integral deep learning forms a learning database;
B. the measuring and judging method comprises the following steps:
1) determining a first information value and a last information value of the judged sample, wherein the judged information group is a continuous information value presenting progressive logic between the first information value and the last information value;
2) inputting the first information value into an input neuron of an input neural network layer to learn and recognize, and extracting the evolution process of the same type from a learning database to be used as a reference calculation model;
3) calculating a 1+1 predicted information value according to the first information value according to a reference calculation model, inputting a 1+1 actual information value into an input neural network layer, and comparing the 1+1 predicted information value with the 1+1 actual information value to obtain a judgment output value;
4) taking the 1+1 actual information value as a new learning cognitive object, calculating an n +1 predicted information value according to a reference calculation model, inputting the n +1 actual information value into an input neural network layer, comparing the n +1 predicted information value with the n +1 actual information value to obtain a judgment output value, and gradually and circularly progressing until evolving to form a final information value;
5) and comprehensively calculating the judgment output value to obtain a plurality of different conclusion probability values, and artificially designating one conclusion probability value according to the expression of the conclusion probability values.
The intelligent measuring and judging method applying the neural network deep learning improves the traditional deep learning method, and realizes continuous associated information judgment by finishing calculation of a deduction function after learning, thereby avoiding omission caused by interval independent data judgment.
In the above intelligent measuring and judging method using neural network deep learning, the intelligent measuring and judging method further includes:
A. the deep learning method comprises the following steps: manually setting a plurality of specific information groups, and inputting the continuity information value in each specific information group into an input neural network layer to enable the logic calculation between two adjacent information values to form an input neuron; carrying out parallel computation on each input neuron of an input neural network layer to obtain each hidden neuron of a first hidden neural network layer, and obtaining n +1 hidden neural network layers step by step through computation until finally obtaining a conclusion probability value of each output neuron in an output neural network layer; the evolution process of the integral deep learning forms a specific learning database;
B. the measuring and judging method comprises the following steps: taking any one or any of the multi-type evolution processes in the specific learning database as a reference calculation model; and comparing the reference calculation model with the actual information values one by one until the judged independent sample is finished to obtain a conclusion probability value.
Specific target attributes can be searched in the judged sample through the specific learning database so as to complete targeted search and identification.
In the above intelligent measuring and judging method applying neural network deep learning, in step a, step 2) of the deep learning method, a large number of reference values are continuity information values of spatial attributes; the plurality of reference values are continuity information values of the time attribute.
In the above-mentioned intelligent measurement and judgment method applying neural network deep learning, the intelligent measurement and judgment method is applied to the human body image focus judgment in the medical field, and the spatial attribute is a plurality of position sequencing scanning images of a section of human body structure (for example, the thoracic cavity); the time attribute is a scanning image of the same position or range position (such as the heart) of the human body, which changes over time.
The continuity information value of the spatial attribute can monitor the structural forms of different positions of the human body so as to judge whether a focus exists in the human body at a certain time and the specific position of the focus. The continuous information value of the time attribute can monitor the change of a single focus of a human body along with the time extension so as to monitor the development situation of the focus under normal conditions, and if the focus is monitored to be suddenly increased and deteriorated, a warning can be provided so as to avoid the death rate increase of a patient caused by the sudden deterioration without self-examination.
In the above-mentioned intelligent measurement and judgment method applying neural network deep learning, the intelligent measurement and judgment method is applied to internal structure aging monitoring in the building field, and the time attribute is a monitoring image in which the same position or range position (for example, a load-bearing beam) of the internal structure of the building is aged over time.
Public buildings have a certain service life, the structures in the buildings are aged and corroded to a certain degree along with the lapse of time, the aging process of the buildings under normal conditions is mastered by utilizing the continuity information value of the time attribute, and once the local mutation enhancement of the building structures is monitored, a warning can be provided to avoid major accidents such as collapse.
In the above intelligent measuring and judging method applying neural network deep learning, in step a, step 2) of the deep learning method, a large number of reference values include continuity information values of different types; the opposite type of continuity information value is included in the plurality of reference values.
In the above intelligent determination method using neural network deep learning, the intelligent determination method is applied to the determination of human body image focus in the medical field, and the different types include human body structure scanning images of symmetrical body group, human body structure scanning images of fat group, human body structure scanning images of lean group, human body structure scanning images of tall group, and human body structure scanning images of short group; the opposite types comprise human body structure scanning images of healthy people and human body structure scanning images of different sick people.
In the above-mentioned intelligent determination method using neural network deep learning, which is applied to the determination of the lesion in the human body image in the medical field, the scanned image of the information group as the reference value is labeled with sex, age, scanned part (for example, chest, abdomen, etc.), health or disease state. The disease state can be marked on a specific disease name.
In the above-mentioned intelligent measurement and judgment method applying neural network deep learning, the intelligent measurement and judgment method is applied to monitoring of internal structure aging in the building field, and the structure name, specification, position and service time are marked in the monitoring image of the information group as the reference value.
In the above-mentioned intelligent measuring and judging method using neural network deep learning, in step a, step 3) of the deep learning method, the learning database at least includes 300 information groups.
Compared with the prior art, the intelligent measuring and judging method applying the neural network deep learning has the following advantages:
1. a large number of labels on the target information values are omitted, and the labor burden can be reduced to the maximum extent.
2. And the characteristic information is integrally identified, so that the monitoring of all target information values is completed, the complexity of creating a plurality of identifiers is saved, and the problem of target information value omission is avoided.
3. And predicting the next information value according to the true value, comparing the next information value with the actual information value, outputting a difference region to perform target information value discrimination, forming function learning derivation on the continuous information value with logic, and coping with the target information value without learning so as to reduce omission.
4. The special recognizer of the specific target information value can be learned and trained, and the corresponding field is continuously expanded. And is continuously evolving into a higher precision, larger scale system.
Drawings
Fig. 1 is a deep learning diagram according to a first embodiment of the present invention.
Detailed Description
The following are specific embodiments of the present invention and are further described with reference to the drawings, but the present invention is not limited to these embodiments.
Example one
As shown in fig. 1, the intelligent measurement and judgment method using neural network deep learning is applied to the human body image lesion judgment in the medical field, the neural network includes an input neural network layer and an output neural network layer, a plurality of hidden neural network layers presenting arrangement exist between the input neural network layer and the output neural network layer, the input neural network layer includes a plurality of input neurons, the hidden neural network layer includes a plurality of hidden neurons, the output neural network layer includes a plurality of output neurons, and the intelligent measurement and judgment method includes the following contents:
A. the deep learning method comprises the following steps:
1) the number of hidden neural network layers is initially set, and the number of hidden neurons in each hidden neural network layer is set; the number of output neurons in an output neural network layer is initially set;
2) collecting a large number of people as reference values, collecting a plurality of continuous human body structure scanning images presented in logic by each person, and respectively putting all the human body structure scanning images of each person into an input neural network layer to enable logic calculation between two adjacent images to form an input neuron; carrying out parallel computation on each input neuron of an input neural network layer to obtain each hidden neuron of a first hidden neural network layer, and obtaining n +1 hidden neural network layers step by step through computation until finally obtaining a conclusion probability value whether each output neuron in the output neural network layer is ill or not;
3) in the step 2), a forward propagation algorithm is used for deduction, a modified weight is obtained through a backward propagation algorithm, a conclusion value meeting the learning expectation is obtained by combining the forward propagation algorithm and the backward propagation algorithm, and the evolution process of the integral deep learning forms a learning database;
B. the measuring and judging method comprises the following steps:
1) determining a first scanning image and a last scanning image of the determined specimen, wherein the set of the determined scanning images is a continuous scanning image presenting logic between the first scanning image and the last scanning image;
2) inputting the first scanning image into an input neuron of an input neural network layer to learn and recognize, and extracting the evolution process of the same type from a learning database to be used as a reference calculation model;
3) calculating a 1+1 predicted scanned image from the first scanned image according to a reference calculation model, enabling a 1+1 actual scanned image to enter an input neural network layer, and comparing the 1+1 predicted scanned image with the 1+1 actual scanned image to obtain a judgment output value;
4) taking the 1+1 actual scanning image as a new learning cognitive object, calculating an n +1 predicted scanning image according to a reference calculation model, inputting the n +1 actual scanning image into an input neural network layer, comparing the n +1 predicted scanning image with the n +1 actual scanning image to obtain a judgment output value, and gradually and circularly progressing until evolving to form a final scanning image;
5) and comprehensively calculating the judgment output value to obtain a plurality of different conclusion probability values, and artificially designating one conclusion probability value according to the expression of the conclusion probability values.
The intelligent measuring and judging method applying the neural network deep learning improves the traditional deep learning method, and realizes continuous associated information judgment by finishing calculation of a deduction function after learning, thereby avoiding omission caused by interval independent data judgment.
The intelligent measuring and judging method further comprises the following steps:
A. the deep learning method comprises the following steps: manually setting a plurality of specific focus images, and putting the continuous scanning sheets in each specific focus image into an input neural network layer to enable the logic calculation between two adjacent scanning sheets to form an input neuron; carrying out parallel computation on each input neuron of an input neural network layer to obtain each hidden neuron of a first hidden neural network layer, and obtaining n +1 hidden neural network layers step by step through computation until finally obtaining a conclusion probability value of each output neuron in an output neural network layer; the evolution process of the integral deep learning forms a specific focus database;
B. the measuring and judging method comprises the following steps: taking any one or any of various evolution processes in a specific focus database as a reference calculation model; and comparing the reference calculation model with the actual scanned images one by one until the judged independent sample is finished to obtain a conclusion probability value.
Specific disease symptoms can be searched in detected patients through a specific lesion database so as to complete targeted search and identification.
In step 2) of the deep learning method, a large number of reference values are continuous scanning images with spatial attributes, and the spatial attributes are a plurality of position sequencing scanning images of a section of human body structure (such as a chest cavity); the plurality of reference values are continuous scanning images with time attributes, and the time attributes are scanning images of the same position or range position (such as the heart) of the human body changing with time.
The continuous scanning image of the spatial attribute can monitor the structural forms of different positions of the human body so as to judge whether a focus exists in the human body at a certain time and the specific position of the focus. The time-based continuous scanning image can monitor the change of a single focus of a human body along with the time extension so as to monitor the development situation of the focus under normal conditions, and if the focus is monitored to be suddenly increased and deteriorated, an alarm can be provided so as to avoid the death rate increase of a patient caused by the sudden deterioration without self-examination.
In step 2) of the deep learning method, the plurality of reference values comprise different types of continuous scanning images, wherein the different types comprise body structure scanning images of symmetrical body-type crowds, body structure scanning images of fat-type crowds, body structure scanning images of thin-type crowds, body structure scanning images of tall-type crowds and body structure scanning images of short crowds; the plurality of reference values include opposite types of continuous scan images, the opposite types include scan images of human structures of healthy people and scan images of human structures of different sick people.
The scanned images as reference values are labeled with sex, age, scanned region (e.g., chest, abdomen, etc.), health, or disease state. The disease state can be marked on a specific disease name.
In step 3) of the deep learning method, the learning database at least comprises 300 human body structure scanning image sets.
The intelligent measuring and judging method applying the neural network deep learning is applied to the medical field and has the following advantages:
1. a large number of labels on the focus are omitted, and the labor burden can be reduced to the maximum extent.
2. The whole identification is carried out on all the human body structure scanning images, the monitoring on all the focuses is completed, the complexity of creating a plurality of identifiers is saved, and the focus omission problem is avoided.
3. And predicting the next scanning image according to the truth value, comparing the next scanning image with the actual image, outputting a difference region for identifying the focus, completing the learning derivation on the development and evolution of the human body structure, and coping with the focus without learning so as to reduce the omission problem.
4. Can learn and train special recognizer of specific focus, and expand corresponding field continuously. And is continuously evolving into a higher precision, larger scale system.
Example two
The intelligent measuring and judging method applying the neural network deep learning is applied to the internal structure aging monitoring in the building field, the neural network comprises an input neural network layer and an output neural network layer, a plurality of layers of hidden neural network layers presenting in arrangement exist between the input neural network layer and the output neural network layer, the input neural network layer comprises a plurality of input neurons, the hidden neural network layer comprises a plurality of hidden neurons, the output neural network layer comprises a plurality of output neurons, and the intelligent measuring and judging method comprises the following contents:
A. the deep learning method comprises the following steps:
1) the number of hidden neural network layers is initially set, and the number of hidden neurons in each hidden neural network layer is set; the number of output neurons in an output neural network layer is initially set;
2) collecting a large number of monitoring images of building structure aging under normal state as reference values, collecting a plurality of time continuous aging monitoring images presented in logic for each building, and putting each aging monitoring image into an input neural network layer to enable logic calculation between two adjacent images to form an input neuron; carrying out parallel computation on each input neuron of an input neural network layer to obtain each hidden neuron of a first hidden neural network layer, and obtaining n +1 hidden neural network layers step by step through computation until finally obtaining a conclusion probability value whether each output neuron in the output neural network layer is normally aged or not;
3) in the step 2), a forward propagation algorithm is used for deduction, a modified weight is obtained through a backward propagation algorithm, a conclusion value meeting the learning expectation is obtained by combining the forward propagation algorithm and the backward propagation algorithm, and the evolution process of the integral deep learning forms a learning database;
B. the measuring and judging method comprises the following steps:
1) determining the earliest monitoring image and the last monitoring image of the judged sample, wherein the judged scanning image set is a time continuous monitoring image presenting progressive logic between the earliest monitoring image and the last monitoring image;
2) inputting the earliest monitored image into an input neuron of an input neural network layer for learning and cognition, and extracting an evolution process from a learning database to be used as a reference calculation model;
3) calculating a 1+1 prediction monitoring image from the early monitoring image according to a reference calculation model, enabling a 1+1 actual monitoring image to enter an input neural network layer, and comparing the 1+1 prediction monitoring image with the 1+1 actual monitoring image to obtain a judgment output value;
4) taking the 1+1 actual monitoring image as a new learning cognitive object, calculating an n +1 predicted monitoring image according to a reference calculation model, inputting the n +1 actual monitoring image into an input neural network layer, comparing the n +1 predicted monitoring image with the n +1 actual monitoring image to obtain a judgment output value, and gradually and circularly progressing until evolving to form a final monitoring image;
5) and comprehensively calculating the judgment output value to obtain a plurality of different conclusion probability values, and artificially designating one conclusion probability value according to the expression of the conclusion probability values.
The monitoring image as a reference value is labeled with a structure name, specification, position, and use time.
The plurality of reference values are continuous monitoring images of time attributes, and the time attributes are monitoring images of the same position or range position (for example, a bearing beam) of the internal structure of the building, which prolongs the aging situation with time.
Public buildings have a certain service life, the structures in the buildings are aged and corroded to a certain degree along with the lapse of time, the aging process of the buildings under normal conditions is mastered by utilizing continuous monitoring images of time attributes, and once the local mutation and enhancement of the building structures are monitored, a warning can be provided to avoid major accidents such as collapse.
The specific embodiments described herein are merely illustrative of the spirit of the invention. Various modifications or additions may be made to the described embodiments or alternatives may be employed by those skilled in the art without departing from the spirit or ambit of the invention as defined in the appended claims.
Claims (10)
1. An intelligent measuring and judging method applying neural network deep learning, wherein the neural network comprises an input neural network layer and an output neural network layer, a plurality of layers of hidden neural network layers presenting arrangement exist between the input neural network layer and the output neural network layer, the input neural network layer comprises a plurality of input neurons, the hidden neural network layers comprise a plurality of hidden neurons, and the output neural network layer comprises a plurality of output neurons, the intelligent measuring and judging method is characterized by comprising the following steps:
A. the deep learning method comprises the following steps:
1) the number of hidden neural network layers is initially set, and the number of hidden neurons in each hidden neural network layer is set; the number of output neurons in an output neural network layer is initially set;
2) collecting a large number of information groups serving as reference values, wherein each information group comprises a plurality of continuity information values presented into logic, and respectively inputting the continuity information values of each information group into an input neural network layer to enable logic calculation between two adjacent information values to form an input neuron; carrying out parallel computation on each input neuron of an input neural network layer to obtain each hidden neuron of a first hidden neural network layer, and obtaining n +1 hidden neural network layers step by step through computation until finally obtaining a conclusion probability value of each output neuron in an output neural network layer;
3) in the step 2), a forward propagation algorithm is used for deduction, a modified weight is obtained through a backward propagation algorithm, a conclusion value meeting the learning expectation is obtained by combining the forward propagation algorithm and the backward propagation algorithm, and the evolution process of the integral deep learning forms a learning database;
B. the measuring and judging method comprises the following steps:
1) determining a first information value and a last information value of the judged sample, wherein the judged information group is a continuous information value presenting progressive logic between the first information value and the last information value;
2) inputting the first information value into an input neuron of an input neural network layer to learn and recognize, and extracting the evolution process of the same type from a learning database to be used as a reference calculation model;
3) calculating a 1+1 predicted information value according to the first information value according to a reference calculation model, inputting a 1+1 actual information value into an input neural network layer, and comparing the 1+1 predicted information value with the 1+1 actual information value to obtain a judgment output value;
4) taking the 1+1 actual information value as a new learning cognitive object, calculating an n +1 predicted information value according to a reference calculation model, inputting the n +1 actual information value into an input neural network layer, comparing the n +1 predicted information value with the n +1 actual information value to obtain a judgment output value, and gradually and circularly progressing until evolving to form a final information value;
5) and comprehensively calculating the judgment output value to obtain a plurality of different conclusion probability values, and artificially designating one conclusion probability value according to the expression of the conclusion probability values.
2. The intelligent measurement and judgment method applying the neural network deep learning according to claim 1, further comprising:
A. the deep learning method comprises the following steps: manually setting a plurality of specific information groups, and inputting the continuity information value in each specific information group into an input neural network layer to enable the logic calculation between two adjacent information values to form an input neuron; carrying out parallel computation on each input neuron of an input neural network layer to obtain each hidden neuron of a first hidden neural network layer, and obtaining n +1 hidden neural network layers step by step through computation until finally obtaining a conclusion probability value of each output neuron in an output neural network layer; the evolution process of the integral deep learning forms a specific learning database;
B. the measuring and judging method comprises the following steps: taking any one or any of the multi-type evolution processes in the specific learning database as a reference calculation model; and comparing the reference calculation model with the actual information values one by one until the judged independent sample is finished to obtain a conclusion probability value.
3. The intelligent testing and judging method applying the neural network deep learning according to claim 1, wherein in step 2) of the deep learning method, a large number of reference values are continuity information values of spatial attributes; the plurality of reference values are continuity information values of the time attribute.
4. The intelligent measurement and judgment method applying neural network deep learning according to claim 3, wherein the intelligent measurement and judgment method is applied to human body image focus judgment in the medical field, and the spatial attributes are a plurality of position sequencing scanning images of a segment of human body structure; the time attribute is a scanning image of the situation that the same position or range position of the human body changes along with the time extension.
5. The method according to claim 3, wherein the time attribute is a monitoring image of an aging situation of the same position or range position of the building internal structure, which is prolonged with time.
6. The intelligent testing and judging method applying the neural network deep learning according to claim 1, wherein in step 2) of the deep learning method, a large number of reference values comprise different types of continuity information values; the opposite type of continuity information value is included in the plurality of reference values.
7. The method according to claim 6, wherein the method is applied to the identification of human body image focus in the medical field, and the different types include the human body structure scanning image of a symmetrical body group, the human body structure scanning image of a fat group, the human body structure scanning image of a lean group, the human body structure scanning image of a tall group, and the human body structure scanning image of a short group; the opposite types comprise human body structure scanning images of healthy people and human body structure scanning images of different sick people.
8. The method of claim 1, wherein the method is applied to the identification of lesion in human body image in the medical field, and the scanned image of the information group as the reference value is labeled with sex, age, scanned part, health or disease state.
9. The method of claim 1, wherein the structure name, specification, location and service time are marked on the monitoring image of the information group as the reference value in the aging monitoring of the internal structure in the building field.
10. The method according to claim 1, wherein in step 3) of the deep learning method, the learning database comprises at least 300 information groups.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010407547.9A CN111611893B (en) | 2020-05-14 | 2020-05-14 | Intelligent measuring and judging method applying neural network deep learning |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010407547.9A CN111611893B (en) | 2020-05-14 | 2020-05-14 | Intelligent measuring and judging method applying neural network deep learning |
Publications (2)
Publication Number | Publication Date |
---|---|
CN111611893A true CN111611893A (en) | 2020-09-01 |
CN111611893B CN111611893B (en) | 2024-03-19 |
Family
ID=72198009
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202010407547.9A Active CN111611893B (en) | 2020-05-14 | 2020-05-14 | Intelligent measuring and judging method applying neural network deep learning |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN111611893B (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112017776A (en) * | 2020-10-27 | 2020-12-01 | 平安科技(深圳)有限公司 | Disease prediction method based on dynamic graph and medical knowledge map and related equipment |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2015172560A1 (en) * | 2014-05-16 | 2015-11-19 | 华南理工大学 | Central air conditioner cooling load prediction method based on bp neural network |
CN105701540A (en) * | 2016-01-11 | 2016-06-22 | 清华大学 | Self-generated neural network construction method |
CN105787557A (en) * | 2016-02-23 | 2016-07-20 | 北京工业大学 | Design method of deep nerve network structure for computer intelligent identification |
CN108073985A (en) * | 2016-11-14 | 2018-05-25 | 张素菁 | A kind of importing ultra-deep study method for voice recognition of artificial intelligence |
CN109255119A (en) * | 2018-07-18 | 2019-01-22 | 五邑大学 | A kind of sentence trunk analysis method and system based on the multitask deep neural network for segmenting and naming Entity recognition |
US20190065956A1 (en) * | 2017-08-31 | 2019-02-28 | International Business Machines Corporation | Facilitating neural networks |
-
2020
- 2020-05-14 CN CN202010407547.9A patent/CN111611893B/en active Active
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2015172560A1 (en) * | 2014-05-16 | 2015-11-19 | 华南理工大学 | Central air conditioner cooling load prediction method based on bp neural network |
CN105701540A (en) * | 2016-01-11 | 2016-06-22 | 清华大学 | Self-generated neural network construction method |
CN105787557A (en) * | 2016-02-23 | 2016-07-20 | 北京工业大学 | Design method of deep nerve network structure for computer intelligent identification |
CN108073985A (en) * | 2016-11-14 | 2018-05-25 | 张素菁 | A kind of importing ultra-deep study method for voice recognition of artificial intelligence |
US20190065956A1 (en) * | 2017-08-31 | 2019-02-28 | International Business Machines Corporation | Facilitating neural networks |
CN109255119A (en) * | 2018-07-18 | 2019-01-22 | 五邑大学 | A kind of sentence trunk analysis method and system based on the multitask deep neural network for segmenting and naming Entity recognition |
Non-Patent Citations (1)
Title |
---|
任红格;李冬梅;李福进: "动态神经网络分类器主动学习算法及其智能控制应用", 计算机应用与软件, vol. 33, no. 7 * |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112017776A (en) * | 2020-10-27 | 2020-12-01 | 平安科技(深圳)有限公司 | Disease prediction method based on dynamic graph and medical knowledge map and related equipment |
CN112017776B (en) * | 2020-10-27 | 2021-01-15 | 平安科技(深圳)有限公司 | Disease prediction method based on dynamic graph and medical knowledge map and related equipment |
Also Published As
Publication number | Publication date |
---|---|
CN111611893B (en) | 2024-03-19 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Singh et al. | Stroke prediction using artificial intelligence | |
Rigas et al. | Biometric identification based on the eye movements and graph matching techniques | |
Walsh et al. | Deep learning–based outcome prediction in progressive fibrotic lung disease using high-resolution computed tomography | |
Liu et al. | Risk scoring for prediction of acute cardiac complications from imbalanced clinical data | |
CN113113130A (en) | Tumor individualized diagnosis and treatment scheme recommendation method | |
Hashemi et al. | Delve into Multiple Sclerosis (MS) lesion exploration: A modified attention U-Net for MS lesion segmentation in Brain MRI | |
Yousefi et al. | Recognizing patterns of visual field loss using unsupervised machine learning | |
Kaya et al. | A new approach for congestive heart failure and arrhythmia classification using angle transformation with LSTM | |
Hassan et al. | Comparative analysis of machine learning algorithms in detection of brain tumor | |
CN116616770A (en) | Multimode depression screening and evaluating method and system based on voice semantic analysis | |
Akter | Dementia Identification for Diagnosing Alzheimer's Disease using XGBoost Algorithm | |
CN116705300A (en) | Medical decision assistance method, system and storage medium based on sign data analysis | |
Sirigineedi et al. | The Early Detection Of Alzheimer's Illness Using Machine Learning And Deep Learning Algorithms | |
CN111611893A (en) | Intelligent measuring and judging method applying neural network deep learning | |
Maulana et al. | Performance analysis and feature extraction for classifying the severity of atopic dermatitis diseases | |
Srivastav et al. | Machine Learning Models for Early Brain Stroke Prediction: A Performance Analogy | |
Ganitidis et al. | Stratification of carotid atheromatous plaque using interpretable deep learning methods on B-mode ultrasound images | |
CN117334337A (en) | Cancer patient pain intelligent evaluation and early warning system based on image recognition technology | |
Nivedita et al. | DBN with a Developed Version of Thermal Exchange Optimization (TEO) Model for Skin Cancer Detection and Classification | |
Phankokkruad | Evaluation of deep transfer learning models in glaucoma detection for clinical application | |
Sanjay et al. | Deep learning techniques for early detection of Alzheimer’s disease: A review | |
Niemann et al. | Combining subgroup discovery and clustering to identify diverse subpopulations in cohort study data | |
Stepanyan et al. | Neural network modeling and correlation analysis of brain plasticity mechanisms in stroke patients | |
Maistrou et al. | Technology of automated medical diagnostics using fuzzy linguistic variables and consensus ranking methods | |
Alagarsamy et al. | A Novel Technique for Prophecy of Brain Strokes |
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 | ||
TA01 | Transfer of patent application right | ||
TA01 | Transfer of patent application right |
Effective date of registration: 20231227 Address after: Room 603, No. 512 Yunchuang Road, Jiangling Street, Wujiang District, Suzhou City, Jiangsu Province, 215000 Applicant after: Longliqiang Artificial Intelligence Technology (Suzhou) Co.,Ltd. Address before: Room 302-16, Jiudingfeng Building, No. 888 Changbai Mountain Road, Qingdao Area, China (Shandong) Pilot Free Trade Zone, Qingdao City, Shandong Province, 266555 Applicant before: Qingdao Hanlin Huili Technology Co.,Ltd. |
|
GR01 | Patent grant | ||
GR01 | Patent grant |