Water purification equipment state identification method based on image identification technology

Abstract

The invention discloses a water purification equipment state identification method based on an image identification technology. The method comprises the following steps: S1, acquiring and processing an image; s2, marking the image; s3, training the model; s4, verifying, adjusting and optimizing the deep learning model; and S5, identifying the abnormal state of the water purification equipment. After being processed by a water quality image algorithm, microorganisms with unobvious characteristics are enhanced and displayed, and background noise and interference terms are subjected to fuzzy processing, so that the contrast ratio of the microorganisms to the surrounding background is more obvious; meanwhile, by adopting the state identification method of the water purification equipment, the working efficiency and accuracy of identifying the running state of the water purification equipment can be greatly improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of image processing, and relates to a method for identifying the state of a water purification device, and specifically to a method for identifying the state of a water purification device based on water quality image recognition technology. Background Art

[0002] With the development of science and technology, people's production activities are also increasing. While the economy is developing and people's living standards are improving, human industrial production, daily life, and the machinery and facilities produced also lead to water pollution. In order to avoid the pollution of groundwater and seawater caused by arbitrary discharge, it is necessary to purify them.

[0003] Each standard state of the water purification equipment corresponds to a different indicator light. When an indicator light does not meet the current standard state, it means that the equipment is in an abnormal state. At present, manual inspection and identification are mainly used to determine whether the water purification equipment is in an abnormal state, or to manually check the monitoring screen and alarm lights to determine whether the water purification equipment is operating normally. This method has problems such as low work efficiency and possible missed detection and false detection due to manual operation. Summary of the invention

[0004] In order to solve the above problems existing in the background technology, the present invention provides a method for identifying the state of a water purification device based on image recognition technology.

[0005] The objective of the present invention is achieved through the following technical solutions:

[0006] A method for identifying the state of a water purification device based on image recognition technology comprises the following steps:

[0007] Step S1: Image acquisition and processing:

[0008] Step S101, water quality image and indicator light image acquisition: a water quality image acquisition system is used to acquire water quality images of the water purification equipment under different standard conditions, and a monitor is installed near the indicator light to acquire the indicator light images of the water purification equipment under different standard conditions;

[0009] Step S102, filtering and denoising: using a sharpening spatial filter to process the noise problem when collecting water quality images and indicator light images;

[0010] Step S103, image grayscale: converting the multi-channel color image into a single-channel grayscale image by using a weighted average grayscale processing method on the denoised image;

[0011] Step S104, histogram equalization: the water quality grayscale image is further subjected to the limited contrast histogram equalization process to enhance the image contrast;

[0012] Step S2: Image marking:

[0013] Step S201, all collected images are divided into two groups according to a ratio of 7:3, the first group of water quality images is a water quality image training set A, and the second group is a water quality image test set B; the first group of indicator light images is a indicator light image training set C, and the second group is an indicator light image test set D;

[0014] Step S202: Mark all images so that each group of images corresponds to different standard working states of the actual water purification equipment;

[0015] Step S3: training model:

[0016] Step S301: Based on deep learning technology, the water quality image training set A obtained in step S201 is used as model training data, and a VGG-16 convolutional neural network is used to construct a water quality image water purification equipment state recognition model W;

[0017] Step S302: Based on deep learning technology, the indicator light image training set C obtained in step S201 is used as model training data, and a VGG-16 convolutional neural network is used to construct an indicator light water purification equipment state recognition model L;

[0018] Step S4: Verification and tuning of deep learning model:

[0019] Step S401, taking the water quality image test group B obtained in step S201 as an input parameter and inputting it into the water quality image water purification equipment state recognition model W obtained in step S302;

[0020] Step S402: Based on the verification result, the training weight parameters and network topology parameters of the water quality image water purification equipment state recognition model W are optimized;

[0021] Step S403, taking the indicator light image test group D obtained in step S201 as an input parameter, and inputting it into the indicator light water purification equipment state recognition model L obtained in step S304;

[0022] Step S404: Based on the verification result, the training weight parameters and network topology parameters of the indicator light water purification equipment state recognition model L are optimized;

[0023] Step S5: Identification of abnormal state of water purification equipment:

[0024] Step S501, obtain a water quality image and an indicator light image through a water quality image acquisition system and a monitor respectively, use the water quality image water purification equipment state recognition model W optimized in step S402 to recognize the acquired water quality image, and obtain the water purification equipment state 1; use the indicator light water purification equipment state recognition model R optimized in step S404 to recognize the acquired water quality image, and obtain the water purification equipment state 2;

[0025] Step S502: Compare the water purification device status 1 with the water purification device status 2. If they are the same, it indicates that the water purification device is working normally. If they are different, it indicates that the water purification device is working abnormally.

[0026] Compared with the prior art, the present invention has the following advantages:

[0027] After being processed by the water quality image algorithm, microorganisms whose characteristics are not obvious are enhanced, while background noise and interference items are blurred, and the contrast between the microorganisms and the surrounding background is more significant; at the same time, the use of this water purification equipment status identification method can greatly improve the work efficiency and accuracy of identifying the operating status of water purification equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 This is a flow chart for water purification equipment status identification based on image recognition technology;

[0029] Figure 2 This is the flow chart of water quality image processing;

[0030] Figure 3 It is a water quality image acquisition system. DETAILED DESCRIPTION

[0031] The technical solution of the present invention is further described below in conjunction with the accompanying drawings, but is not limited thereto. Any modification or equivalent replacement of the technical solution of the present invention without departing from the spirit and scope of the technical solution of the present invention should be included in the protection scope of the present invention.

[0032] The present invention provides a method for identifying the state of a water purification device based on image recognition technology. Figure 1 As shown, the method comprises the following steps:

[0033] Step S1: Image acquisition and processing:

[0034] Step S101, water quality image and indicator light image acquisition: a water quality image acquisition system is used to acquire water quality images of the water purification equipment under different standard conditions, and a monitor is installed near the indicator light to acquire the indicator light images of the water purification equipment under different standard conditions.

[0035] like Figure 2 As shown, the water quality image acquisition system includes an underwater fill light, a microscope, an optical CCD digital camera and an image acquisition card. When collecting water quality images, the underwater fill light is turned on, the optical CCD digital camera is aimed at the microscope to take pictures, and then the data is transmitted to the computer through the image acquisition card.

[0036] Step S102, filtering and denoising: Due to the intensity of the light source, camera heat, and microbial flow during the image acquisition process, the image will generate noise; at the same time, general filters do not pay much attention to the edges of the image, while the edges of the microorganisms are important information extraction points of the image, and have high requirements for noise processing. Therefore, a sharpening spatial filter is used here to deal with the noise problem during image acquisition.

[0037] Step S103, image grayscale: a weighted average grayscale processing method is used to convert the multi-channel color image into a single-channel grayscale image.

[0038] In this step, the weighted average grayscale processing method takes the weights of the three-channel components R component, G component and B component as 0.299, 0.587 and 0.144 respectively.

[0039] Step S104, histogram equalization: the water quality grayscale image is further subjected to limited contrast histogram equalization processing to enhance the image contrast, further highlight the water impurities while suppressing the enhancement effect of background noise as much as possible.

[0040] In this step, the threshold for limiting contrast histogram equalization is based on 2, and can be adaptively adjusted according to different situations.

[0041] Step S2: Image marking:

[0042] Step S201, divide all collected images into two groups according to the ratio of 7:3, the first group of water quality images is water quality image training set A, and the second group is water quality image test group B; the first group of indicator light images is indicator light image training set C, and the second group is indicator light image test group D.

[0043] Step S202: Mark all the images so that each group of images corresponds to a different standard working state of the actual water purification equipment.

[0044] Step S3: training model:

[0045] Step S301: Based on deep learning technology, the water quality image training set A obtained in step S201 is used as model training data, and a water quality image water purification equipment state recognition model W is constructed using a VGG-16 convolutional neural network;

[0046] In this step, the VGG-16 convolutional neural network is mainly composed of 13 convolutional layers and 3 fully connected layers. The water quality image training set A of different standard states of water purification equipment is input. The features of water quality image training set A are extracted in the first 13 convolutional layers, and the last 3 fully connected layers map the extracted features to category probabilities. The last fully connected layer outputs the prediction results of the water quality image water purification equipment state recognition model W.

[0047] Step S302: Based on deep learning technology, the indicator light image training set C obtained in step S201 is used as model training data, and a VGG-16 convolutional neural network is used to construct an indicator light water purification equipment state recognition model L;

[0048] In step S302, the VGG-16 convolutional neural network is mainly composed of 13 convolutional layers and 3 fully connected layers. The indicator light image training set C of different standard states of the water purification equipment is input. The features of the indicator light image training set C are extracted in the first 13 convolutional layers, and the last three fully connected layers map the extracted features to the category probability. The last fully connected layer outputs the prediction result of the indicator light water purification equipment state recognition model L.

[0049] Step S4: Verification and tuning of deep learning model:

[0050] Step S401, taking the water quality image test group B obtained in step S201 as an input parameter and inputting it into the water quality image water purification equipment state recognition model W obtained in step S301;

[0051] Step S402: Based on the verification result obtained in step S401, the training weight parameters and network topology parameters of the water quality image water purification equipment state recognition model R are optimized;

[0052] Step S403, using the indicator light image test group D obtained in step S201 as an input parameter, and inputting it into the indicator light water purification equipment state recognition model L obtained in step S302;

[0053] Step S404: Based on the verification result obtained in step S403, the training weight parameters and network topology parameters of the indicator light water purification equipment state recognition model L are optimized.

[0054] Step S5: Identification of abnormal state of water purification equipment:

[0055] Step S501, obtain a water quality image and an indicator light image through a water quality image acquisition system and a monitor respectively, use the water quality image water purification equipment state recognition model W optimized in step S402 to recognize the acquired water quality image, and obtain the water purification equipment state 1; use the indicator light water purification equipment state recognition model L optimized in step S404 to recognize the acquired water quality image, and obtain the water purification equipment state 2;

[0056] Step S502: Compare the water purification device status 1 with the water purification device status 2. If they are the same, it indicates that the water purification device is working normally. If they are different, it indicates that the water purification device is working abnormally.

Claims

1. A method for identifying the state of a water purification device based on image recognition technology, characterized in that The method comprises the following steps: Step S1: Image acquisition and processing: Step S101, water quality image and indicator light image acquisition: a water quality image acquisition system is used to acquire water quality images of the water purification equipment under different standard conditions, and a monitor is installed near the indicator light to acquire the indicator light images of the water purification equipment under different standard conditions; Step S102, filtering and denoising: using a sharpening spatial filter to process the noise problem when collecting water quality images and indicator light images; Step S103, image grayscale: converting the multi-channel color image into a single-channel grayscale image by using a weighted average grayscale processing method on the denoised image; Step S104, histogram equalization: the water quality grayscale image is further subjected to the limited contrast histogram equalization process to enhance the image contrast; Step S2: Image marking: Step S201, all collected images are divided into two groups according to a ratio of 7:3, the first group of water quality images is a water quality image training set A, and the second group is a water quality image test set B; the first group of indicator light images is a indicator light image training set C, and the second group is an indicator light image test set D; Step S202: Mark all images so that each group of images corresponds to different standard working states of the actual water purification equipment; Step S3: training model: Step S301: Based on deep learning technology, the water quality image training set A obtained in step S201 is used as model training data, and a VGG-16 convolutional neural network is used to construct a water quality image water purification equipment state recognition model W; Step S302: Based on deep learning technology, the indicator light image training set C obtained in step S201 is used as model training data, and a VGG-16 convolutional neural network is used to construct an indicator light water purification equipment state recognition model L; Step S4: Verification and tuning of deep learning model: Step S401, taking the water quality image test group B obtained in step S201 as an input parameter and inputting it into the water quality image water purification equipment state recognition model W obtained in step S302; Step S402: Based on the verification result, the training weight parameters and network topology parameters of the water quality image water purification equipment state recognition model W are optimized; Step S403, taking the indicator light image test group D obtained in step S201 as an input parameter, and inputting it into the indicator light water purification equipment state recognition model L obtained in step S304; Step S404: Based on the verification result, the training weight parameters and network topology parameters of the indicator light water purification equipment state recognition model L are optimized; Step S5: Identification of abnormal state of water purification equipment: Step S501, obtain a water quality image and an indicator light image through a water quality image acquisition system and a monitor respectively, use the water quality image water purification equipment state recognition model W optimized in step S402 to recognize the acquired water quality image, and obtain the water purification equipment state 1; use the indicator light water purification equipment state recognition model R optimized in step S404 to recognize the acquired water quality image, and obtain the water purification equipment state 2; Step S502: Compare the water purification device status 1 with the water purification device status 2. If they are the same, it indicates that the water purification device is working normally. If they are different, it indicates that the water purification device is working abnormally.

2. The water purification equipment state identification method based on image recognition technology according to claim 1 is characterized in that In step S101, the water quality image acquisition system includes an underwater fill light, a microscope, an optical CCD digital camera and an image acquisition card. When acquiring water quality images, the underwater fill light is turned on, the optical CCD digital camera is aimed at the microscope to take pictures, and then the data is transmitted to the computer through the image acquisition card.

3. The water purification equipment state identification method based on image recognition technology according to claim 1 is characterized in that In the step S103, the weights of the three-channel components R, G and B are taken as 0.299, 0.587 and 0.144 respectively by the weighted average grayscale processing method.

4. The water purification equipment state identification method based on image recognition technology according to claim 1 is characterized in that In step S301, the VGG-16 convolutional neural network is mainly composed of 13 convolutional layers and 3 fully connected layers. The water quality image training set A of different standard states of the water purification equipment is input. The features of the water quality image training set A are extracted in the first 13 convolutional layers, and the last 3 fully connected layers map the extracted features to the category probability. The last fully connected layer outputs the prediction result of the water quality image water purification equipment state recognition model W.

5. The water purification equipment state identification method based on image recognition technology according to claim 1 is characterized in that In S302, the VGG-16 convolutional neural network is mainly composed of 13 convolutional layers and 3 fully connected layers. The indicator light image training set C of different standard states of the water purification equipment is input. The features of the indicator light image training set C are extracted in the first 13 convolutional layers, and the last three fully connected layers map the extracted features to the category probability. The last fully connected layer outputs the prediction result of the indicator light water purification equipment state recognition model L.

Images