CN112686801A - Water quality monitoring method based on aerial image and series echo state network - Google Patents
Water quality monitoring method based on aerial image and series echo state network Download PDFInfo
- Publication number
- CN112686801A CN112686801A CN202110007265.4A CN202110007265A CN112686801A CN 112686801 A CN112686801 A CN 112686801A CN 202110007265 A CN202110007265 A CN 202110007265A CN 112686801 A CN112686801 A CN 112686801A
- Authority
- CN
- China
- Prior art keywords
- water quality
- image
- state network
- echo state
- noise
- 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
Images
Landscapes
- Image Processing (AREA)
Abstract
The water quality monitoring method based on aerial photography images and a series echo state network comprises the following steps of 1, acquiring water quality aerial photography experiment image data; step 2, adding salt and pepper noise on the basis of image data; step 3, filtering noise in the image data by using median filtering, and solving the problem of motion image blurring; step 4, extracting the color characteristics of the water quality image signal; step 5, training a series echo state network by using the sample set; and 6, embedding the trained series echo state network model into software for practical application. According to the method, the influence of a noise environment on data collected by a sensor is simulated, the noise in image data is filtered by using median filtering, isolated noise points in the image are eliminated, the problem of image motion blur caused by the motion of an unmanned aerial vehicle can be eliminated, and in order to effectively solve the problem of nonlinearity of aerial images and water quality types, the water quality classification model of the aerial images is solved through a series echo state network.
Description
Technical Field
The invention relates to the field of water quality monitoring, in particular to a water quality monitoring method based on aerial images and a series echo state network.
Background
Water is the origin of human life, and the life activities of people can not leave water resources, so that the water is an important pillar for maintaining human life and is also a necessary resource for human production and labor, and along with the continuous development of industry, people need to monitor the water quality in real time, and protect and treat water pollution as early as possible so as to ensure the safe water use of people. Therefore, the water quality monitor is very important in water quality detection, scientific and reasonable water quality data and information can be provided for people, the water quality data and the information are used as the basis for preventing and treating water pollution and protecting water resource ecology, and the water quality monitor plays a very important role in environmental protection and water monitoring. Through water quality detection, reasonable protection is carried out on water resources so as to avoid influencing ecological balance and human health.
The water quality detection is to acquire specific data and information aiming at water quality pollution and ensure that a water body has no problem. There are many methods for monitoring water, which can be divided into: chemical analysis, ion chromatography, spectrophotometry, electrochemical analysis, atomic spectrometry, etc. However, these test methods are expensive and cannot monitor water quality in real time.
Disclosure of Invention
In order to solve the problems, the invention provides a water quality monitoring method based on aerial images and a series echo state network on the basis of collecting water quality aerial images by an unmanned aerial vehicle and simulating a noise environment. In order to reduce the influence of environmental noise on data acquired by the sensor as much as possible, the method reduces the noise influence, eliminates image motion blur and enhances the robustness of a model obtained by training through smooth filtering. In order to achieve the purpose, the invention provides a water quality monitoring method based on aerial images and a series echo state network, which comprises the following specific steps:
step 1, acquiring experimental image data: detecting a place needing water quality monitoring by using an unmanned aerial vehicle, transmitting collected water quality monitoring images back to a server, and establishing a water quality monitoring image sample set;
step 2, adding salt and pepper noise on the basis of image data: adding salt and pepper noise to the water quality detection image data acquired in the step 1, and simulating the interference of signals in a noise environment by adding the salt and pepper noise;
and 3, filtering noise in the image data by using median filtering: the median filtering algorithm is utilized to inhibit salt and pepper noise in the water quality image, eliminate isolated noise points in the image and simultaneously eliminate the problem of image motion blur caused by the motion of the unmanned aerial vehicle;
step 4, extracting the color characteristics of the water quality image signal: extracting an image color characteristic vector on the basis of the water quality image subjected to median filtering, and establishing a water quality image characteristic sample set;
and 5, training a series echo state network by using the sample set: taking the extracted water quality sample characteristic data as input, taking a corresponding water quality grade label as input, and training a series echo state network;
and 6, embedding the trained series echo state network model into software for practical application, and calculating the water quality grade of the reservoir to be detected.
Further, the process of simulating data acquisition in a noisy environment in step 2 can be expressed as:
when the unmanned aerial vehicle is used for water quality images of the drainage reservoir, the unmanned aerial vehicle is possibly greatly interfered, the interference of the water quality images under shooting is simulated through the added salt and pepper noise, and the formula of the added salt and pepper noise is as follows:
wherein, ImaxAnd IminIs the maximum value and the minimum value of the pixel points of the water quality image, and p is the noise appearing in the imageProbability, p is in the range of 15% -25%, ixyThe actual value of the collected image pixel point (x, y) is obtained, and the f (x, y) is the image pixel point value after the salt and pepper noise is added.
Further, the process of filtering noise in the image data using median filtering in step 3 can be expressed as:
the median filter formula is:
y(n)＝med[x(n-N)...x(n)x(n+N)] (2)
in the formula, x (N-N).. x (N) x (N + N) represents pixel points in the water quality image, y (N) is image pixel points after median filtering, med [ ] is the value of all pixel points in a median filtering window and takes the median, and median filtering can eliminate salt and pepper noise in the image and solve the problem of motion image fuzzification caused by unmanned aerial vehicle flight.
Further, the process of extracting the color features of the water quality image signal in step 4 can be expressed as:
converting RGB color in the aerial image into HSV space through color space, wherein H represents hue H in the color space to be in the range of 0 DEG, 360 DEG, S represents saturation S to be in the range of 0,1, and V represents brightness V to be in the range of 0,1, then carrying out non-uniform quantization on HSV component, and carrying out H quantization 7 level, S quantization 2 level and V quantization 2 level and using the following formula to extract color characteristics of the aerial image:
G＝QsQvH+QvS+V (3)
wherein Q issAnd QvQuantization levels, Q, of S and V, respectivelys＝3、QvA color feature G of 72 dimensions can be obtained.
Further, the process of training the tandem echo state network using the sample set in step 5 can be expressed as:
step 5.1, establishing an aerial image training sample set: labeling labels E (i) of the color feature samples G (i) obtained in the step 4 to form a water quality aerial photography image sample training set D (i) ((G) (i), E (i));
step 5.2, designing a series echo state network, wherein the series echo state network consists of an input layer, two reserve pools and an output layer, and the input layer consists of 72Neurons were organized with the first reservoir consisting of N1A second reservoir consisting of N2Each neuron consists of 1 neuron, and the output layer consists of
Step 5.3, initialize the network, connect the training sample characteristics G (i) with the weight matrix W through the inputinEntering a reserve pool, E (i) connecting the weights W through feedbackbackEntering a reserve pool, and acquiring the system state and the output state according to the following sequence:
x1(i+1)＝a1f1(WinG(i+1)+W1x1(i)) (4)
x2(i+1)＝a2f2(Woutx1(i+1)+W2x2(i)+Wbacky(i)) (5)
y(i)＝g(Wout[x2(i),G(i)]) (6)
wherein x is1(i) And x2(i) Is a system parameter of the first reserve pool and the second reserve pool, a1And a1Is the regulating factor of two storage tanks, f1(. and f)2(. g) is a stimulus function sigmoid of a reserve pool node, and is a stimulus function tanh, W of a reserve pool output unit1And W2A connection weight matrix, W, representing neurons inside the reservoiroutRepresenting a matrix of output values;
and 5.4, training the echo state network through the training sample set, and adjusting parameters in the network to obtain a trained series echo state network model.
The water quality monitoring method based on the aerial image and the series echo state network has the advantages that: the invention has the technical effects that:
1. the invention simulates the severe environment of the unmanned aerial vehicle during flying, realizes the water quality detection function under the interference of the noise environment, and enhances the stability, reliability and robustness of the series echo state network;
2. the invention filters the noise in the image data by using median filtering, eliminates isolated noise points in the image, and can eliminate the problem of image motion blur caused by the motion of the unmanned aerial vehicle;
3. the invention solves the water quality classification model of the aerial image through the series echo state network, and effectively solves the problem of nonlinearity of the aerial image and the water quality classification.
Drawings
FIG. 1 is a flow chart of the present invention;
FIG. 2 is a diagram of a tandem echo state network architecture according to the present invention;
Detailed Description
The invention is described in further detail below with reference to the following detailed description and accompanying drawings:
the invention provides a water quality monitoring method based on aerial images and a series echo state network, which aims to solve the problem of water quality detection of aerial images and improve the stability and accuracy of a training model. FIG. 1 is a flow chart of the present invention. The steps of the present invention will be described in detail with reference to the flow chart.
Step 1, acquiring experimental image data: detecting a place needing water quality monitoring by using an unmanned aerial vehicle, transmitting collected water quality monitoring images back to a server, and establishing a water quality monitoring image sample set;
step 2, adding salt and pepper noise on the basis of image data: adding salt and pepper noise to the water quality detection image data acquired in the step 1, and simulating the interference of signals in a noise environment by adding the salt and pepper noise;
the process of simulating data acquisition in a noisy environment in step 2 can be expressed as:
when the unmanned aerial vehicle is used for water quality images of the drainage reservoir, the unmanned aerial vehicle is possibly greatly interfered, the interference of the water quality images under shooting is simulated through the added salt and pepper noise, and the formula of the added salt and pepper noise is as follows:
wherein, ImaxAnd IminIs the maximum value and the minimum value of the pixel points of the water quality imageThe value p is the probability of noise occurrence of the image, the value range of p is 15% -25%, ixyThe actual value of the collected image pixel point (x, y) is obtained, and the f (x, y) is the image pixel point value after the salt and pepper noise is added.
And 3, filtering noise in the image data by using median filtering: the median filtering algorithm is utilized to inhibit salt and pepper noise in the water quality image, eliminate isolated noise points in the image and simultaneously eliminate the problem of image motion blur caused by the motion of the unmanned aerial vehicle;
the process of filtering noise in the image data using median filtering in step 3 can be expressed as:
the median filter formula is:
y(n)＝med[x(n-N)...x(n)x(n+N)] (2)
in the formula, x (N-N) · x (N + N) represents pixel points in a water quality image, y (N) is an image pixel point after median filtering, and med [ ] is a median value of all pixel points in a median filtering window, wherein the median filtering can eliminate salt and pepper noise in the image and can solve the problem of motion image fuzzification caused by unmanned aerial vehicle flight.
Step 4, extracting the color characteristics of the water quality image signal: extracting an image color characteristic vector on the basis of the water quality image subjected to median filtering, and establishing a water quality image characteristic sample set;
the process of extracting the color features of the water quality image signal in the step 4 can be expressed as follows:
converting RGB color in the aerial image into HSV space through color space, wherein H represents hue H in the color space to be in the range of 0 DEG, 360 DEG, S represents saturation S to be in the range of 0,1, and V represents brightness V to be in the range of 0,1, then carrying out non-uniform quantization on HSV component, and carrying out H quantization 7 level, S quantization 2 level and V quantization 2 level and using the following formula to extract color characteristics of the aerial image:
G＝QsQvH+QvS+V (3)
wherein Q issAnd QvQuantization levels, Q, of S and V, respectivelys＝3、QvA color feature G of 72 dimensions can be obtained.
And 5, training a series echo state network by using the sample set: taking the extracted water quality sample characteristic data as input, taking a corresponding water quality grade label as input, and training a series echo state network;
the process of training the tandem echo state network using the sample set in step 5 can be expressed as:
step 5.1, establishing an aerial image training sample set: labeling labels E (i) of the color feature samples G (i) obtained in the step 4 to form a water quality aerial photography image sample training set D (i) ((G) (i), E (i));
step 5.2, designing a series echo state network, wherein the series echo state network consists of an input layer, two reserve pools and an output layer, the input layer consists of 72 neurons, and the first reserve pool consists of N1A second reservoir consisting of N2Each neuron consists of 1 neuron, and the output layer consists of
Step 5.3, initialize the network, connect the training sample characteristics G (i) with the weight matrix W through the inputinEntering a reserve pool, E (i) connecting the weights W through feedbackbackEntering a reserve pool, and acquiring the system state and the output state according to the following sequence:
x1(i+1)＝a1f1(WinG(i+1)+W1x1(i)) (4)
x2(i+1)＝a2f2(Woutx1(i+1)+W2x2(i)+Wbacky(i)) (5)
y(i)＝g(Wout[x2(i),G(i)]) (6)
wherein x is1(i) And x2(i) Is a system parameter of the first reserve pool and the second reserve pool, a1And a1Is the regulating factor of two storage tanks, f1(. and f)2(. g) is a stimulus function sigmoid of a reserve pool node, and is a stimulus function tanh, W of a reserve pool output unit1And W2A connection weight matrix, W, representing neurons inside the reservoiroutRepresenting a matrix of output values;
and 5.4, training the echo state network through the training sample set, and adjusting parameters in the network to obtain a trained series echo state network model.
And 6, embedding the trained series echo state network model into software for practical application, and calculating the water quality grade of the reservoir to be detected.
The above description is only a preferred embodiment of the present invention, and is not intended to limit the present invention in any way, but any modifications or equivalent variations made according to the technical spirit of the present invention are within the scope of the present invention as claimed.
Claims (5)
1. The water quality monitoring method based on the aerial image and the series echo state network comprises the following specific steps:
step 1, acquiring water quality aerial photography experiment image data: detecting a place needing water quality monitoring by using an unmanned aerial vehicle, transmitting collected water quality monitoring images back to a server, and establishing a water quality monitoring image sample set;
step 2, adding salt and pepper noise on the basis of image data: adding salt and pepper noise to the water quality detection image data acquired in the step 1, and simulating the interference of signals in a noise environment by adding the salt and pepper noise;
and 3, filtering noise in the image data by using median filtering: the median filtering algorithm is utilized to inhibit salt and pepper noise in the water quality image, eliminate isolated noise points in the image and simultaneously eliminate the problem of image motion blur caused by the motion of the unmanned aerial vehicle;
step 4, extracting the color characteristics of the water quality image signal: extracting an image color characteristic vector on the basis of the water quality image subjected to median filtering, and establishing a water quality image characteristic sample set;
and 5, training a series echo state network by using the sample set: taking the extracted water quality sample characteristic data as input, taking a corresponding water quality grade label as input, and training a series echo state network;
and 6, embedding the trained series echo state network model into software for practical application, and calculating the water quality grade of the reservoir to be detected.
2. The water quality monitoring method based on aerial images and series echo state network as claimed in claim 1, wherein: the process of simulating data acquisition in a noisy environment in step 2 can be expressed as:
when the unmanned aerial vehicle is used for water quality images of the drainage reservoir, the unmanned aerial vehicle is possibly greatly interfered, the interference of the water quality images under shooting is simulated through the added salt and pepper noise, and the formula of the added salt and pepper noise is as follows:
wherein, ImaxAnd IminIs the maximum value and the minimum value of the pixel points of the water quality image, p is the probability of noise occurrence of the image, the value range of p is 15 percent to 25 percent, i isxyThe actual value of the collected image pixel point (x, y) is obtained, and the f (x, y) is the image pixel point value after the salt and pepper noise is added.
3. The water quality monitoring method based on aerial images and series echo state network as claimed in claim 1, wherein: the process of filtering noise in the image data using median filtering in step 3 can be expressed as:
the median filter formula is:
y(n)＝med[x(n-N)...x(n)...x(n+N)] (2)
in the formula, x (N-N) · x (N + N) represents pixel points in a water quality image, y (N) is an image pixel point after median filtering, and med [ ] is a median value of all pixel points in a median filtering window, wherein the median filtering can eliminate salt and pepper noise in the image and can solve the problem of motion image fuzzification caused by unmanned aerial vehicle flight.
4. The water quality monitoring method based on aerial images and series echo state network as claimed in claim 1, wherein: the process of extracting the color features of the water quality image signal in the step 4 can be expressed as follows:
converting RGB color in the aerial image into HSV space through color space, wherein H represents hue H in the color space to be in the range of 0 DEG, 360 DEG, S represents saturation S to be in the range of 0,1, and V represents brightness V to be in the range of 0,1, then carrying out non-uniform quantization on HSV component, and carrying out H quantization 7 level, S quantization 2 level and V quantization 2 level and using the following formula to extract color characteristics of the aerial image:
G＝QsQvH+QvS+V (3)
wherein Q issAnd QvQuantization levels, Q, of S and V, respectivelys＝3、QvA color feature G of 72 dimensions can be obtained.
5. The water quality monitoring method based on aerial images and series echo state network as claimed in claim 1, wherein: the process of training the tandem echo state network using the sample set in step 5 can be expressed as:
step 5.1, establishing an aerial image training sample set: labeling labels E (i) of the color feature samples G (i) obtained in the step 4 to form a water quality aerial photography image sample training set D (i) ((G) (i), E (i));
step 5.2, designing a series echo state network, wherein the series echo state network consists of an input layer, two reserve pools and an output layer, the input layer consists of 72 neurons, and the first reserve pool consists of N1A second reservoir consisting of N2The output layer consists of 1 neuron;
step 5.3, initialize the network, connect the training sample characteristics G (i) with the weight matrix W through the inputinEntering a reserve pool, E (i) connecting the weights W through feedbackbackEntering a reserve pool, and acquiring the system state and the output state according to the following sequence:
x1(i+1)＝a1f1(WinG(i+1)+W1x1(i)) (4)
x2(i+1)＝a2f2(Woutx1(i+1)+W2x2(i)+Wbacky(i)) (5)
y(i)＝g(Wout[x2(i),G(i)]) (6)
wherein x is1(i) And x2(i) Is a system parameter of the first reserve pool and the second reserve pool, a1And a1Is the regulating factor of two storage tanks, f1(. and f)2(. g) is a stimulus function sigmoid of a reserve pool node, and is a stimulus function tanh, W of a reserve pool output unit1And W2A connection weight matrix, W, representing neurons inside the reservoiroutRepresenting a matrix of output values;
and 5.4, training the echo state network through the training sample set, and adjusting parameters in the network to obtain a trained series echo state network model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110007265.4A CN112686801B (en) | 2021-01-05 | 2021-01-05 | Water quality monitoring method based on aerial image and serial echo state network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110007265.4A CN112686801B (en) | 2021-01-05 | 2021-01-05 | Water quality monitoring method based on aerial image and serial echo state network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112686801A true CN112686801A (en) | 2021-04-20 |
| CN112686801B CN112686801B (en) | 2023-06-20 |
Family
ID=75457259
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110007265.4A Active CN112686801B (en) | 2021-01-05 | 2021-01-05 | Water quality monitoring method based on aerial image and serial echo state network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112686801B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113901965A (en) * | 2021-12-07 | 2022-01-07 | 广东省科学院智能制造研究所 | Liquid state identification method in liquid separation and liquid separation system |
| CN114140362A (en) * | 2022-01-29 | 2022-03-04 | 杭州微影软件有限公司 | Thermal imaging image correction method and device |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109035142B (en) * | 2018-07-16 | 2020-06-19 | 西安交通大学 | Satellite image super-resolution method combining countermeasure network with aerial image prior |
-
2021
- 2021-01-05 CN CN202110007265.4A patent/CN112686801B/en active Active
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113901965A (en) * | 2021-12-07 | 2022-01-07 | 广东省科学院智能制造研究所 | Liquid state identification method in liquid separation and liquid separation system |
| CN113901965B (en) * | 2021-12-07 | 2022-05-24 | 广东省科学院智能制造研究所 | Liquid state identification method in liquid separation and liquid separation system |
| CN114140362A (en) * | 2022-01-29 | 2022-03-04 | 杭州微影软件有限公司 | Thermal imaging image correction method and device |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112686801B (en) | 2023-06-20 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Jin et al. | Deep learning for underwater image recognition in small sample size situations | |
| CN107665355B (en) | Agricultural pest detection method based on regional convolutional neural network | |
| CN111046967A (en) | Underwater image classification method based on convolutional neural network and attention mechanism | |
| CN109376589B (en) | ROV deformation small target identification method based on convolution kernel screening SSD network | |
| CN104899866B (en) | A kind of intelligentized infrared small target detection method | |
| CN112686801A (en) | Water quality monitoring method based on aerial image and series echo state network | |
| CN112597864B (en) | Monitoring video anomaly detection method and device | |
| CN105841961A (en) | Bearing fault diagnosis method based on Morlet wavelet transformation and convolutional neural network | |
| CN106156765A (en) | safety detection method based on computer vision | |
| Qu et al. | A pedestrian detection method based on yolov3 model and image enhanced by retinex | |
| CN110503613A (en) | Based on the empty convolutional neural networks of cascade towards removing rain based on single image method | |
| CN108734717B (en) | Single-frame star map background dark and weak target extraction method based on deep learning | |
| CN113111758A (en) | SAR image ship target identification method based on pulse neural network | |
| CN115358952B (en) | Image enhancement method, system, equipment and storage medium based on meta-learning | |
| CN114896887B (en) | Frequency-using equipment radio frequency fingerprint identification method based on deep learning | |
| CN105894507B (en) | Image quality evaluating method based on amount of image information natural scene statistical nature | |
| CN111340758A (en) | Novel efficient iris image quality evaluation method based on deep neural network | |
| CN111695413A (en) | Signal first arrival pickup method and device combining U-Net and Temporal encoding | |
| CN116206214A (en) | Automatic landslide recognition method, system, equipment and medium based on lightweight convolutional neural network and double attention | |
| CN106023250A (en) | Target occlusion intensity evaluation method in image identification and tracking | |
| CN112712008A (en) | Water environment early warning judgment method based on 3D convolutional neural network | |
| Wu et al. | SlimRGBD: A geographic information photography noise reduction system for aerial remote sensing | |
| CN109145810A (en) | Details in fingerprint point detecting method, device, equipment, system and storage medium | |
| CN109948571B (en) | Optical remote sensing image ship detection method | |
| CN116152288A (en) | Intelligent object approaching motion detection method suitable for electric power operation and maintenance |
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 |