CN111739057A - Free liquid level identification and extraction method based on U-net convolution neural network model - Google Patents
Free liquid level identification and extraction method based on U-net convolution neural network model Download PDFInfo
- Publication number
- CN111739057A CN111739057A CN202010495698.4A CN202010495698A CN111739057A CN 111739057 A CN111739057 A CN 111739057A CN 202010495698 A CN202010495698 A CN 202010495698A CN 111739057 A CN111739057 A CN 111739057A
- Authority
- CN
- China
- Prior art keywords
- liquid level
- neural network
- network model
- free liquid
- convolutional neural
- 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
- 239000007788 liquid Substances 0.000 title claims abstract description 96
- 238000003062 neural network model Methods 0.000 title claims abstract description 15
- 238000000605 extraction Methods 0.000 title claims abstract description 11
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 58
- 238000012549 training Methods 0.000 claims abstract description 26
- 230000006870 function Effects 0.000 claims abstract description 22
- 238000012795 verification Methods 0.000 claims abstract description 19
- 238000012360 testing method Methods 0.000 claims abstract description 18
- 238000000034 method Methods 0.000 claims abstract description 11
- 238000001514 detection method Methods 0.000 claims abstract description 6
- 230000008602 contraction Effects 0.000 claims description 28
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 claims description 27
- 230000004913 activation Effects 0.000 claims description 19
- 230000011218 segmentation Effects 0.000 claims description 19
- 238000011176 pooling Methods 0.000 claims description 10
- 239000012530 fluid Substances 0.000 claims description 5
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 claims description 3
- 238000010276 construction Methods 0.000 claims description 3
- 238000005457 optimization Methods 0.000 claims description 3
- 230000008569 process Effects 0.000 claims description 3
- 238000005259 measurement Methods 0.000 abstract description 6
- 238000012545 processing Methods 0.000 abstract description 5
- 238000013473 artificial intelligence Methods 0.000 abstract description 2
- 230000033001 locomotion Effects 0.000 description 8
- 210000000746 body region Anatomy 0.000 description 7
- 238000010586 diagram Methods 0.000 description 6
- 239000000523 sample Substances 0.000 description 5
- 238000002474 experimental method Methods 0.000 description 3
- 239000002352 surface water Substances 0.000 description 3
- 238000013528 artificial neural network Methods 0.000 description 2
- 230000007547 defect Effects 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000013507 mapping Methods 0.000 description 2
- 238000000691 measurement method Methods 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 239000000969 carrier Substances 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000002372 labelling Methods 0.000 description 1
- 239000003949 liquefied natural gas Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000008439 repair process Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000000926 separation method Methods 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
- G06T7/20—Analysis of motion
- G06T7/215—Motion-based segmentation
-
- G06T5/70—
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- 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]
Landscapes
- Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
Abstract
A free liquid level identification and extraction method based on a U-net convolution neural network model belongs to the technical field of image processing and automatic detection. Firstly, processing a liquid level image and manually marking a free liquid level to form a data set with a segmented liquid level image, and dividing the data set into a training set, a verification set and a test set. Secondly, building a U-net convolution neural network model; and (4) introducing a training set into the U-net convolutional neural network, performing feature learning on the image, and predicting the free liquid level. And thirdly, verifying the model by using the verification set and adjusting the model, and saving the optimal model when the loss function is not descending. And finally, deploying the trained U-net convolutional neural network model for automatically detecting the liquid level image in the test set and evaluating the model. The invention can extract the free liquid level of the crushing wave by a method combining non-contact measurement and artificial intelligence, and can solve the technical problem of difficult liquid level identification of the crushing wave.
Description
Technical Field
The invention belongs to the technical field of image processing and automatic detection, and relates to a method for identifying and extracting a free liquid level, in particular to a broken liquid level based on a U-net convolutional neural network model, which can be used in the technical fields of aerospace, transportation and the like.
Background
At present, in the engineering fields of ocean, transportation, aerospace and the like, the identification and extraction of the free liquid level are widely required to be solved. For example, there is much research work on wave energy converters at sea, with conversion efficiency being assessed by the rate of change of volume of the free surface. Thus, the result of capturing the free liquid level may directly affect the design and performance of the device. In addition, in deep sea engineering, the design of large transportation facilities liquefied natural gas carriers (FLNG) vessels also requires estimation of the free liquid level movement inside the storage tanks. In summary, how to accurately capture the free liquid level is an urgent problem to be solved by engineering science.
At present, the identification and extraction method of the free liquid level is divided into a wave height instrument contact measurement method and an optical camera non-contact measurement method. The wave height meter is used for measuring the wave height at a fixed point and taking the wave height as the height of the free liquid level. The non-contact measurement is based on the appearance of optical image measurement technology, images of free liquid level movement are shot, and the images are processed to extract the free liquid level.
There are three main errors in extracting the free liquid level using a wave height meter: (1) due to the working principle of the wave probe, the resistance type or capacitance type wave height instrument can only measure continuous liquid. Thus, if the free level of liquid is contaminated with some air, or the free level breaks suddenly at the level of the wave height gauge, the wave height is inaccurate. This means that the free liquid surface with the breaking waves or entrained air cannot be accurately captured using the wave height instrument contact method. (2) The wave height meter converts the electrical signal into a height signal, and calibrates the height signal before the experiment all the time, wherein the calibration coefficient depends on the water quality (dielectric constant), the environmental temperature and the test condition, and the physical conditions can also cause errors in measurement. (3) The presence of the wave height meter can interfere with the flow field around it, and the results are inaccurate when measuring very narrow areas of the free surface. In addition, the free level motion in a closed tank cannot be interpolated by only two to three wave probes due to the non-linearity of the free level of the liquid. The larger the number of wave probes, the more accurate the motion state of the captured free liquid surface. However, it is difficult to balance the number of wave detectors and their impact on the ambient flow field. The current methods of contactless measurement for extracting free liquid level have limitations for extracting broken liquid level or free liquid level with entrained air.
China starts late in the deep sea technical field, and is vigorously developing deep sea energy from 'eleven-five' and emphasizes sustainable development, so that how to efficiently extract the free liquid level of liquid is a hotspot and an economic concern in the fluid field. At present, no technology for realizing free liquid level identification and extraction by utilizing a neural network exists at home and abroad.
Disclosure of Invention
Aiming at the defects of the prior art, the invention provides a free liquid level identification and extraction method based on a U-net convolutional neural network model, which can identify and extract an accurate free liquid level result through a liquid level image.
In order to achieve the purpose, the technical scheme of the invention is as follows:
a free liquid level identification and extraction method based on a U-net convolutional neural network model comprises the following steps:
in the first step, a sample training set, a verification set and a test set are constructed.
1.1) acquiring a fluid free liquid level motion image as an original image in a mode of shooting free liquid level motion through experiments, manually marking the free liquid level in the original image, and dividing the image into a water body area and an air area.
1.2) carrying out global threshold segmentation on the image, and carrying out denoising processing after the image is inverted.
1.3) marking the water body area and the air area as black and white areas respectively.
1.4) the segmented image and the corresponding original image together constitute a data set.
1.5) manually dividing a data set into a training set, a verification set and a test set, wherein the training set is used for model construction; the verification set is used for verifying the model; the test set is used for testing of the model. The classification criteria of the three data sets are that the more the number of verification set samples is, the more reliable the model result is under the condition of sufficient experimental data.
And secondly, building a U-net convolutional neural network model.
The built U-net convolutional neural network model is a coder-decoder structure and consists of a contraction path and an expansion path.
The contraction path comprises four contraction blocks, each contraction block comprises two convolution layers and a maximum pooling layer, wherein the convolution kernel size of the maximum pooling layer is 2 multiplied by 2, and the step length is 2; the convolution layer has a convolution kernel size of 3 × 3 and a step size of 1. In addition, the convolutional layer includes a linear active cell and padding (padding ═ 1). The linear activation unit is a Relu activation function, and padding is 1.
The expansion path comprises four expansion blocks corresponding to the contraction block, wherein each expansion block comprises a deconvolution layer with convolution kernel size of 2 multiplied by 2 and step length of 1 and two convolution layers with convolution kernel size of 3 multiplied by 3 and step length of 1. The convolution layer of the extension path also includes a linear activation unit Relu activation function and padding (padding ═ 1) as the contraction path.
Between the contraction path and the expansion path, the feature information of the contraction block is transferred into the corresponding expansion block through the operations of copying and cutting. And adopting a convolution layer with the convolution kernel size of 1 multiplied by 1 after the fourth contraction block, wherein the convolution layer comprises a linear activation unit and padding, and mapping the output result into the probability from 0 to 1 through a sigmoid activation function.
The U-net convolutional neural network model comprises 23 convolutional layers and 4 maximum pooling layers.
And thirdly, introducing a training set into the U-net convolutional neural network, performing feature learning on the images in the data set, predicting the free liquid level through an original image U-net convolutional neural network model, obtaining feedback information through the artificially labeled free liquid level image by the model, adjusting parameters of the U-net convolutional neural network model, and outputting the free liquid level image by the U-net convolutional neural network model. Wherein the water body area and the air area are respectively represented by black and white, and the junction of the two areas is a free liquid level.
And fourthly, verifying the U-net convolutional neural network model by using a verification set, and storing the U-net convolutional neural network model meeting the output condition.
4.1) the U-net convolution neural network model utilizes a cross entropy loss function to calculate the error between the real free liquid level and the predicted free liquid level, and the smaller the loss function is, the better the prediction result is.
The cross entropy loss function expression is as follows:
wherein m represents the number of samples; k represents the classification number, two classification water and air are adopted in the invention, and K is 1 and 2; omega (x)i) Is the weight value of the important pixel; p is a radical ofk(xi) Representing the probability value calculated by Sigmoid;
4.2) inputting a verification set, and adjusting parameters of the U-net convolutional neural network model by using an Adam optimization algorithm.
4.3) iterating the network until the value of the loss function reaches the minimum and is not descending, and saving the U-net convolution neural network model at the moment.
4.4) carrying out regional hollow detection on the predicted free liquid level image:
if a black block exists in a white part representing an air area or a black block exists in a black part representing a water body area in the free liquid level segmentation result of the U-net convolutional neural network model, the fact that the predicted free liquid level image has errors is indicated, the error image needs to be repaired, and the step 4.5 is carried out);
if the white part representing the air area in the free liquid level segmentation result of the U-net convolutional neural network model does not have a black block, or the black part representing the water body area does not have a white block, performing step 4.6);
and 4.5) after the picture segmentation error part in the U-net convolutional neural network model is manually repaired, forming a data set by the picture segmentation error part and the corresponding original image, recombining the data set with the original training set into a new training set, inputting the new training set into the U-net convolutional neural network model for training, and repeating the operation processes of the third step to the fourth step.
4.6) storing the U-net convolution neural network model until the output result of the U-net convolution neural network model meets the requirement.
And fifthly, evaluating the segmentation result of the model on the liquid level by using the test set.
And deploying the trained U-net convolutional neural network model for automatically detecting the concentrated liquid level image, and evaluating the model result by using the Dice coefficient. The Dice coefficient describes the error magnitude of the real area and the prediction area, and the expression is as follows:
wherein V (A) represents the area of the water body region in the artificially labeled liquid level image, namely the area of the real water body region in the original image; v (B) represents the water body area in the output image of the U-net convolutional neural network model, and V (A &' B) represents the water body area correctly predicted by the U-net convolutional neural network model.
The invention has the beneficial effects that: compared with the prior art, the invention avoids the defect of using a wave height instrument to measure the free liquid level; the invention can extract the free liquid level of the crushing wave by a method combining non-contact measurement and artificial intelligence, and can solve the technical problem of difficult identification and extraction of the free liquid level of the crushing liquid in the existing free liquid level separation method.
Drawings
FIG. 1 is a flow chart of the present invention.
FIG. 2 is a schematic diagram of the segmentation of free liquid surface water body area and air area by liquid surface image and artificial labeling according to the present invention.
FIG. 2(a) is an original image of the free liquid level;
fig. 2(b) is a schematic diagram of the division of the free liquid surface water body region and the air region which are marked manually.
FIG. 3 is a schematic diagram of the U-net neural network of the present invention.
FIG. 4 is a schematic diagram of the image hollow detection and repair process of the present invention:
FIG. 4(a) is an original image of the free liquid level;
FIG. 4(b) is a schematic diagram of a manually labeled free liquid surface water region and air region segmentation;
FIG. 4(c) is a U-net convolutional neural network prediction graph;
fig. 4(d) is a schematic diagram of the division of the artificially restored water body region and air region.
Detailed Description
The following detailed description of embodiments of the invention is provided in connection with the accompanying drawings.
A free liquid level identification and extraction method based on a U-net convolutional neural network model comprises the following steps:
referring to fig. 1, the fluid free liquid level identification method based on intelligent segmentation comprises the following steps:
in the first step, a sample training set, a verification set and a test set are constructed.
1.1) acquiring a fluid free liquid level motion image as an original image in a mode of shooting free liquid level motion through experiments, manually marking the free liquid level in the original image, and dividing the image into a water body area and an air area.
1.2) carrying out global threshold segmentation on the image, and carrying out denoising processing after the image is inverted.
1.3) marking the water body area and the air area as black and white areas, respectively, see fig. 2.
1.4) the segmented image and the corresponding original image together constitute a data set.
1.5) manually dividing a data set into a training set, a verification set and a test set, wherein the training set is used for model construction; the verification set is used for verifying the model; the test set is used for testing of the model. The classification criteria of the three data sets are that the more the number of verification set samples is, the more reliable the model result is under the condition of sufficient experimental data.
(2) And building a U-net convolutional neural network, and referring to fig. 3.
The built U-net convolutional neural network model is a coder-decoder structure and consists of a contraction path and an expansion path.
The puncturing path comprises four puncturing blocks, each puncturing block comprises two convolutional layers and a maximum pooling layer, wherein the convolutional kernel size of the convolutional layer is 3 × 3, the step size is 1, and the convolutional layers comprise a linear activation unit (Relu activation function) and padding (padding ═ 1). The convolution kernel size of the maximum pooling layer is 2 × 2 with a step size of 2. .
The contraction path comprises four contraction blocks, each contraction block comprises two convolution layers and a maximum pooling layer, wherein the convolution kernel size of the maximum pooling layer is 2 multiplied by 2, and the step length is 2; the convolution layer has a convolution kernel size of 3 × 3 and a step size of 1. In addition, the convolutional layer includes a linear active cell and padding (padding ═ 1). The linear activation unit is a Relu activation function, and padding is 1.
The expansion path comprises four expansion blocks corresponding to the contraction block, wherein each expansion block comprises a deconvolution layer with convolution kernel size of 2 multiplied by 2 and step length of 1 and two convolution layers with convolution kernel size of 3 multiplied by 3 and step length of 1. The convolution layer of the extension path also includes a linear activation unit Relu activation function and padding (padding ═ 1) as the contraction path.
Between the contraction path and the expansion path, the feature information of the contraction block is transferred into the corresponding expansion block through the operations of copying and cutting. And adopting a convolution layer with the convolution kernel size of 1 multiplied by 1 after the fourth contraction block, wherein the convolution layer comprises a linear activation unit and padding, and mapping the output result into the probability from 0 to 1 through a sigmoid activation function.
The U-net convolutional neural network model comprises 23 convolutional layers and 4 maximum pooling layers.
And thirdly, introducing a training set into the U-net convolutional neural network, performing feature learning on the images in the data set, predicting the free liquid level through an original image U-net convolutional neural network model, obtaining feedback information through the artificially labeled free liquid level image by the model, adjusting parameters of the U-net convolutional neural network model, and outputting the free liquid level image by the U-net convolutional neural network model. Wherein the water body area and the air area are respectively represented by black and white, and the junction of the two areas is a free liquid level.
And fourthly, verifying the U-net convolutional neural network model by using a verification set, and storing the U-net convolutional neural network model meeting the output condition.
4.1) the U-net convolution neural network model utilizes a cross entropy loss function to calculate the error between the real free liquid level and the predicted free liquid level, and the smaller the loss function is, the better the prediction result is.
The cross entropy loss function expression is as follows:
wherein m represents the number of samples; k represents the classification number, two classification water and air are adopted in the invention, and K is 1 and 2; omega (x)i) Is the weight value of the important pixel; p is a radical ofk(xi) Representing the probability value calculated by Sigmoid;
4.2) inputting a verification set, and adjusting parameters of the U-net convolutional neural network model by using an Adam optimization algorithm.
4.3) iterating the network until the value of the loss function reaches the minimum and is not descending, and saving the U-net convolution neural network model at the moment.
4.4) carrying out regional hollow detection on the predicted free liquid level image, wherein a black block exists in a white part representing an air region or a white block exists in a black part representing a water body region in a free liquid level segmentation result of a partial U-net convolutional neural network model, and referring to FIG. 4.
And 4.5) manually repairing the part of the picture segmentation error in the U-net convolutional neural network model, forming a data set with the corresponding original image, recombining the data set with the original training set into a new training set, inputting the new training set into the U-net convolutional neural network model for training, and repeating the operation processes of the third step and the fourth step. And storing the U-net convolutional neural network model until the output result of the U-net convolutional neural network model meets the requirement.
And fifthly, evaluating the segmentation result of the model on the liquid level by using the test set.
And deploying the trained U-net convolutional neural network model for automatically detecting the concentrated liquid level image, and evaluating the model result by using the Dice coefficient. The Dice coefficient describes the error magnitude of the real area and the prediction area, and the expression is as follows:
wherein V (A) represents the area of the water body region in the artificially labeled liquid level image, namely the area of the real water body region in the original image; v (B) represents the water body area in the output image of the U-net convolutional neural network model, and V (A &' B) represents the water body area correctly predicted by the U-net convolutional neural network model.
The above-mentioned embodiments only express the embodiments of the present invention, but not should be understood as the limitation of the scope of the invention patent, it should be noted that, for those skilled in the art, many variations and modifications can be made without departing from the concept of the present invention, and these all fall into the protection scope of the present invention.
Claims (3)
1. A free liquid level identification and extraction method based on a U-net convolutional neural network model is characterized by comprising the following steps:
firstly, constructing a sample training set, a verification set and a test set;
1.1) taking a fluid free liquid level moving image as an original image, manually marking the free liquid level in the original image, and dividing the image into a water body area and an air area;
1.2) carrying out global threshold segmentation on the image, and carrying out denoising treatment after the image is inverted;
1.3) marking the water body area and the air area as black and white areas respectively;
1.4) forming a data set by the segmented image and the corresponding original image;
1.5) manually dividing a data set into a training set, a verification set and a test set, wherein the training set is used for model construction; the verification set is used for verifying the model; the test set is used for testing the model;
secondly, building a U-net convolution neural network model;
the built U-net convolutional neural network model is a coder-decoder structure and consists of a contraction path and an expansion path;
the contraction path comprises four contraction blocks, each contraction block comprises two convolution layers and a maximum pooling layer, wherein the convolution kernel size of the maximum pooling layer is 2 multiplied by 2, and the step length is 2; the convolution kernel size of the convolution layer is 3 multiplied by 3, and the step length is 1; in addition, the convolutional layer comprises a linear active unit and filling; the expansion path comprises four expansion blocks corresponding to the contraction block, wherein each expansion block comprises a deconvolution layer with convolution kernel size of 2 multiplied by 2 and step length of 1 and two convolution layers with convolution kernel size of 3 multiplied by 3 and step length of 1; the convolution layer of the expansion path is the same as the contraction path, and also comprises a linear activation unit and a filling;
between the contraction path and the expansion path, the characteristic information of the contraction block is transferred into the corresponding expansion block through the operations of copying and cutting; a convolution layer with a convolution kernel size of 1 multiplied by 1 is adopted after the fourth contraction block, the convolution layer comprises a linear activation unit and padding, and the output result is mapped into the probability from 0 to 1 through a sigmoid activation function;
thirdly, introducing a training set into the U-net convolutional neural network, performing feature learning on images in a data set, predicting the free liquid level through an original image U-net convolutional neural network model, obtaining feedback information through the model through a manually marked free liquid level image, adjusting parameters of the U-net convolutional neural network model, and outputting the free liquid level image through the U-net convolutional neural network model; wherein the water body area and the air area are respectively represented by black and white, and the junction of the two areas is a free liquid level;
fourthly, verifying the U-net convolutional neural network model by using a verification set, and storing the U-net convolutional neural network model meeting the output condition;
4.1) calculating the error between the real free liquid level and the predicted free liquid level by adopting a cross entropy loss function through a U-net convolution neural network model, wherein the smaller the loss function is, the better the prediction result is;
4.2) inputting a verification set, and adjusting parameters of the U-net convolutional neural network model by using an Adam optimization algorithm;
4.3) iterating the network until the value of the loss function reaches the minimum and is not descending, and storing the U-net convolution neural network model at the moment;
4.4) carrying out regional hollow detection on the predicted free liquid level image:
if a black block exists in a white part representing an air area or a black block exists in a black part representing a water body area in the free liquid level segmentation result of the U-net convolutional neural network model, the fact that the predicted free liquid level image has errors is indicated, the error image needs to be repaired, and the step 4.5 is carried out);
if the white part representing the air area in the free liquid level segmentation result of the U-net convolutional neural network model does not have a black block, or the black part representing the water body area does not have a white block, performing step 4.6);
4.5) after the picture segmentation error part in the U-net convolutional neural network model is manually repaired, the picture segmentation error part and the corresponding original image form a data set, the data set and the original training set are recombined into a new training set, the new training set is input into the U-net convolutional neural network model for training, and the operation processes of the third step to the fourth step are repeated;
4.6) storing the U-net convolution neural network model until the output result of the U-net convolution neural network model meets the requirement;
fifthly, evaluating the segmentation result of the model on the liquid level by using the test set;
and deploying the trained U-net convolutional neural network model for automatically detecting the concentrated liquid level image, and evaluating the model result by using the Dice coefficient.
2. The method for identifying and extracting the free liquid level based on the U-net convolutional neural network model as claimed in claim 1, wherein the linear activation units in the contraction path and the expansion path are both Relu activation functions.
3. The method for identifying and extracting free liquid level based on the U-net convolutional neural network model as claimed in claim 1 or 2, wherein the cross entropy loss function expression in the step 4.1) is as follows:
wherein m represents the number of samples; k represents the classification number, two classification water and air are adopted in the invention, and K is 1 and 2; omega (x)i) Is the weight value of the important pixel; p is a radical ofk(xi) Representing the probability value calculated by Sigmoid.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010495698.4A CN111739057B (en) | 2020-06-03 | 2020-06-03 | Free liquid level identification and extraction method based on U-net convolution neural network model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010495698.4A CN111739057B (en) | 2020-06-03 | 2020-06-03 | Free liquid level identification and extraction method based on U-net convolution neural network model |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111739057A true CN111739057A (en) | 2020-10-02 |
| CN111739057B CN111739057B (en) | 2022-10-18 |
Family
ID=72648438
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010495698.4A Active CN111739057B (en) | 2020-06-03 | 2020-06-03 | Free liquid level identification and extraction method based on U-net convolution neural network model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111739057B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112233117A (en) * | 2020-12-14 | 2021-01-15 | 浙江卡易智慧医疗科技有限公司 | New coronary pneumonia CT detects discernment positioning system and computing equipment |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109308695A (en) * | 2018-09-13 | 2019-02-05 | 镇江纳兰随思信息科技有限公司 | Based on the cancer cell identification method for improving U-net convolutional neural networks model |
| CN110659718A (en) * | 2019-09-12 | 2020-01-07 | 中南大学 | Small convolution nuclear cell counting method and system based on deep convolution neural network |
-
2020
- 2020-06-03 CN CN202010495698.4A patent/CN111739057B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109308695A (en) * | 2018-09-13 | 2019-02-05 | 镇江纳兰随思信息科技有限公司 | Based on the cancer cell identification method for improving U-net convolutional neural networks model |
| CN110659718A (en) * | 2019-09-12 | 2020-01-07 | 中南大学 | Small convolution nuclear cell counting method and system based on deep convolution neural network |
Non-Patent Citations (1)
| Title |
|---|
| 赵浩: "基于TensorFlow的卷积神经网络图像分类实践策略研究", 《价值工程》 * |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112233117A (en) * | 2020-12-14 | 2021-01-15 | 浙江卡易智慧医疗科技有限公司 | New coronary pneumonia CT detects discernment positioning system and computing equipment |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111739057B (en) | 2022-10-18 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US20210319265A1 (en) | Method for segmentation of underground drainage pipeline defects based on full convolutional neural network | |
| CN110176001B (en) | Grad-CAM algorithm-based high-speed rail contact net insulator damage accurate positioning method | |
| CN109376606B (en) | Power inspection image tower foundation fault detection method | |
| CN111444939B (en) | Small-scale equipment component detection method based on weak supervision cooperative learning in open scene of power field | |
| CN110826588A (en) | Drainage pipeline defect detection method based on attention mechanism | |
| KR102346676B1 (en) | Method for creating damage figure using the deep learning-based damage image classification of facility | |
| CN111815576B (en) | Method, device, equipment and storage medium for detecting corrosion condition of metal part | |
| CN111598098A (en) | Water gauge water line detection and effectiveness identification method based on full convolution neural network | |
| CN113487600B (en) | Feature enhancement scale self-adaptive perception ship detection method | |
| CN112001411A (en) | Dam crack detection algorithm based on FPN structure | |
| CN111739057B (en) | Free liquid level identification and extraction method based on U-net convolution neural network model | |
| CN114022715A (en) | Method and device for detecting lead sealing defect of high-voltage cable terminal | |
| CN116840135A (en) | Steel gate accelerated degradation test bed with simultaneous effects of corrosion and fatigue and post-degradation running state evaluation method | |
| CN116310850A (en) | Remote sensing image target detection method based on improved RetinaNet | |
| CN111507398A (en) | Transformer substation metal instrument corrosion identification method based on target detection | |
| CN114022435A (en) | Method for establishing pipeline leakage identification system based on machine vision | |
| Bruno et al. | Decay detection in historic buildings through image-based deep learning | |
| CN113344148A (en) | Marine ship target identification method based on deep learning | |
| CN116630323B (en) | Automatic calculation method, system, medium and equipment for corrosion depth of dense metal | |
| CN113052103A (en) | Electrical equipment defect detection method and device based on neural network | |
| CN112767351A (en) | Transformer equipment defect detection method based on sensitive position dependence analysis | |
| CN112215301A (en) | Image straight line detection method based on convolutional neural network | |
| Li et al. | Surface defect detection algorithm of aluminum based on improved faster RCNN | |
| Ding et al. | A measurement system for the tightness of sealed vessels based on machine vision using deep learning algorithm | |
| CN112085779B (en) | Wave parameter estimation method and device |
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 |