A dust identification and early warning method and a piston-type camera cleaning device

Abstract

The present invention belongs to the field of video image processing technology, and in particular to a dust recognition and early warning method and a piston camera cleaning device, comprising the following steps: step one, image preprocessing, wherein the preprocessing includes grayscale processing, contrast enhancement processing and Gaussian blur processing; step two, feature extraction, which includes edge detection, brightness detection and blur measurement; the dust recognition and early warning method and the piston camera cleaning device generate a dust coverage index DCI through feature extraction and regional analysis, and judge whether the camera cleaning device needs to be started based on the index. Detect the dust coverage on the surface of the camera lens and automatically generate a dust coverage index DCI. Judge whether the cleaning threshold is reached based on the DCI to trigger the automatic cleaning function. Provide an efficient and accurate dust recognition algorithm suitable for the long-term operation environment of industrial or outdoor cameras.

Description

Technical Field

[0001] The invention relates to the technical field of video image processing, and in particular to a fine dust recognition and early warning method and a piston-type camera cleaning device. Background Art

[0002] In outdoor or heavy industrial environments, camera equipment is often exposed to dust, particles and other impurities, which may block the lens and affect the image quality. The cleaning methods currently available on the market, such as brushes, air blowers and fans, have limited effects and are difficult to completely remove adherent dust, and most of them require manual operation. For example, in the prior art, electric fans are used to blow air to remove dust, because the wind force of electric fans is relatively uniform, continuous and relatively gentle. It can cover a large area, but its ability to remove stubborn dust is relatively weak. If it is light dust cleaning, electric fans can meet the needs, but they cannot generate strong airflow to clean gaps and adherent dust; therefore, a dust identification and warning method and a piston-type camera cleaning device are needed. Summary of the invention

[0003] Based on the existing technical problems, the present invention proposes a dust identification and early warning method and a piston-type camera cleaning device.

[0004] The present invention provides a dust identification and early warning method, comprising the following steps:

[0005] Step 1: Image preprocessing, which includes grayscale processing, contrast enhancement processing and Gaussian blur processing.

[0006] Step 2: Feature extraction, which includes edge detection, brightness detection and blur measurement.

[0007] Step 3: Dust area analysis: In the analysis, a dust mask needs to be generated first, and then the dust coverage index DCI is calculated.

[0008] Step 4: Dust coverage judgment: set the DCI threshold. When the DCI exceeds the threshold, it indicates that the dust coverage is serious and the cleaning device is triggered.

[0009] Step 5: When the DCI exceeds the threshold, the algorithm outputs a cleaning signal and starts the cleaning device. After the cleaning device is completed, the algorithm re-analyzes the image to verify the cleaning effect. If the DCI does not meet the cleaning standard, the cleaning is repeated.

[0010] Preferably, the grayscale processing in step 1: converting the collected color image I RGB Convert to grayscale image I gray ; The grayscale formula is as follows: I gray =0.2989*R+0.587*G+0.114*B;

[0011] Among them: R is the red channel; G is the green channel; B is the blue channel;

[0012] Contrast enhancement processing: Perform histogram equalization on the grayscale image to make the dust area clearer by adjusting the contrast of the image;

[0013] Assume the gray value of the original image is I gray , histogram equalization is achieved by the following steps:

[0014] A. Calculate the histogram h(t) of the grayscale image, where t represents the grayscale level;

[0015] B. Calculate the cumulative distribution function:

[0016] C. Map the original grayscale value to the balanced grayscale value I eq :

[0017]

[0018] in:

[0019] N is the total number of pixels;

[0020] S max is the maximum value of the gray level;

[0021] C min is the smallest non-zero value in the cumulative distribution function;

[0022] I eq (x,y) is the gray value after equalization;

[0023] I gray (x,y) is the original grayscale value;

[0024] Gaussian blur processing: In order to reduce noise and enhance the characteristics of dust areas, Gaussian blur filtering is applied to the image; the formula for Gaussian blur filtering is:

[0025] The application on the image is:

[0026] in:

[0027] (x,y) is the original pixel pair;

[0028] I smoothed (x,y) is the pixel value after Gaussian blur filtering;

[0029] (i,j) is a pair of adjacent pixels;

[0030] σ is the standard deviation of the Gaussian distribution, which determines the degree of blur;

[0031] k is the radius of the filter, which is usually chosen to be an integer within 3σ.

[0032] Preferably, in the step 2, edge detection: using Canny edge detection to extract image edges, and calculating the number of edges to represent edge strength E; the edge strength is defined as:

[0033] Among them I edge (x, y) is a binary image, edge pixels are 1, and non-edge pixels are 0; this formula counts the number of all edge pixels; dust-covered areas can cause edges to be broken or lost, resulting in a decrease in the number of edges; the number of edges represents edge strength, and the degree of dust coverage is inversely proportional to the edge strength.

[0034] Preferably, in the blur detection in step 2: using the Laplace operator to calculate the image clarity L, the higher the clarity, the less dust coverage;

[0035] Blurred areas reduce the variance of the Laplacian operator:

[0036] Where: MN is the total number of pixels in the image; ΔI gray (i, j) is the pixel gradient, which indicates the magnitude of the change in the grayscale value of the image;

[0037] ΔI gray The calculation formula for (i,j) is as follows:

[0038] in: is the rate of change of gray value in the x direction;

[0039] is the rate of change of grayscale value in the y direction.

[0040] Preferably, the brightness detection in step 2 is performed by calculating the local brightness standard deviation σ of the image. local To evaluate brightness uniformity; dust coverage will cause local brightness uniformity to decrease and the standard deviation to increase, especially at the edge of the lens:

[0041]

[0042] Where μ is the average brightness of the image.

[0043] Preferably, in step 3, the dust coverage is calculated: a dust coverage index DCI is defined to quantitatively evaluate the dust coverage; the DCI is calculated by the following formula:

[0044]

[0045] Among them, E ref , L refand σ ref is the reference value collected in a clean state;

[0046] E, L, and σ local is the feature value of the current image;

[0047] ω1, ω2 and ω3 are weight coefficients of each feature, determined by experiments: ω1 = 0.4, ω2 = 0.3, ω3 = 0.3;

[0048] The higher the DCI, the more severe the dust coverage.

[0049] A piston-type camera cleaning device proposed in the present invention comprises a camera body, a protective cover is fixedly mounted on the surface of the camera body, both end surfaces of the protective cover are hinged with a U-shaped base via a pin shaft, a U-shaped connecting plate is fixedly mounted on the inner wall of the right end of the protective cover, a control module is fixedly mounted on the inner wall of the bottom end of the U-shaped connecting plate, a driving motor is fixedly mounted on the vertical inner wall of the U-shaped connecting plate, and a driving gear is fixedly mounted on the output end of the driving motor via a coupling.

[0050] Preferably, the top inner wall of the U-shaped connecting plate is rotatably connected to a camshaft via a bearing, a driving cover is fixedly installed on the top of the U-shaped connecting plate, a driven gear is fixedly installed on the bottom end of the camshaft, the teeth of the driven gear are meshed with the teeth of the driving gear, and a cam is fixedly installed on the top of the camshaft, and the cam is located inside the driving cover.

[0051] Preferably, a swing shaft is fixedly installed on the top of the eccentric end of the cam, and the arc surface of the swing shaft is rotatably connected to a connecting rod through a bearing. A sealed pressure tube is fixedly installed on one end of the drive cover, and one end of the connecting rod extends to the interior of the sealed pressure tube. A pneumatic piston is slidably inserted into the inner wall of the sealed pressure tube through a sealing ring, and one end of the pneumatic piston is hinged to the opposite end of the connecting rod through a pin shaft.

[0052] Preferably, a transition pipe is fixedly installed at one end of the sealed pressure pipe, an air filter is fixedly installed at the top of the transition pipe, an air storage tank is fixedly installed at one end of the transition pipe, a one-way valve is arranged between the air storage tank and the transition pipe and between the air filter and the transition pipe, a high-pressure nozzle is fixedly installed at one end of the air storage tank, and an electrically controlled one-way valve is also arranged between the high-pressure nozzle and the air storage tank.

[0053] The beneficial effects of the present invention are:

[0054] 1. Generate a dust coverage index DCI through feature extraction and regional analysis, and judge whether to start the camera cleaning device based on this index. Detect the dust coverage on the camera lens surface and automatically generate the dust coverage index DCI. Determine whether the cleaning threshold is reached based on DCI to trigger the automatic cleaning function. Provide an efficient and accurate dust recognition algorithm suitable for long-term operation environments of industrial or outdoor cameras.

[0055] 2. The piston airflow technology can generate high-pressure pulse airflow in a short time, which can effectively remove stubborn dust and is suitable for such scenarios. Compared with electric fan blowing and dust removal, piston blowing has the advantages of high wind strength, high directionality, strong periodicity and high degree of automation. The airflow generated by the piston is pulsed, which can form a high-intensity directional airflow in a short time. It is suitable for tasks that require centralized cleaning, especially when removing stubborn dust particles. The piston system is more suitable for use in conjunction with an automated control system. It can be driven by a motor or pneumatic drive and cooperate with sensors to achieve a fully automatic dust removal process. For unattended heavy industrial camera equipment, the piston system is easier to integrate automated cleaning. Although the piston blowing has multiple conversion losses in energy consumption, the energy efficiency can be significantly improved by selecting high-efficiency motors, optimizing piston design, heat recovery, and gas storage utilization. In particular, intelligent control and on-demand cleaning technology can play a big role in reducing unnecessary energy consumption. The piston-type air-blowing dust removal camera is suitable for cleaning and maintenance tasks in heavy industry, outdoor monitoring, and highly polluted areas, including but not limited to ports, construction sites, mining areas, chemical plants, and outdoor monitoring stations, effectively extending the service life of the camera equipment and improving image clarity. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 It is an algorithm flow chart of a dust particle identification and early warning method;

[0057] Figure 2 A three-dimensional diagram of the protective cover structure of a piston-type camera cleaning device;

[0058] Figure 3 A three-dimensional diagram of the high-pressure nozzle structure of a piston-type camera cleaning device;

[0059] Figure 4 A three-dimensional diagram of the drive cover structure of a piston-type camera cleaning device;

[0060] Figure 5 The present invention is a stereoscopic diagram of the cam structure of a piston-type camera cleaning device.

[0061] In the figure: 1. Camera body; 2. Protective cover; 3. U-shaped base; 4. U-shaped connecting plate; 5. Control module; 6. Drive motor; 7. Drive gear; 8. Camshaft; 9. Drive cover; 10. Driven gear; 11. Cam; 12. Swing shaft; 13. Connecting rod; 14. Sealed pressure pipe; 15. Pneumatic piston; 16. Transition pipe; 17. Air filter; 18. Air tank; 19. One-way valve; 20. High-pressure nozzle; 21. Electric one-way valve. DETAILED DESCRIPTION

[0062] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0063] Embodiment 1

[0064] Reference Figure 1 , a dust recognition and early warning method comprises the following steps: Step 1, image preprocessing, wherein the preprocessing comprises grayscale processing, contrast enhancement processing and Gaussian blur processing.

[0065] Grayscale processing in step 1: The collected color image I RGB Convert to grayscale image I gray ; The grayscale formula is as follows: I gray =0.2989*R+0.587*G+0.114*B;

[0066] Among them: R is the red channel; G is the green channel; B is the blue channel;

[0067] Contrast enhancement processing: Perform histogram equalization on the grayscale image to make the dust area clearer by adjusting the contrast of the image;

[0068] Assume the gray value of the original image is I gray , histogram equalization is achieved by the following steps:

[0069] A. Calculate the histogram h(t) of the grayscale image, where t represents the grayscale level;

[0070] B. Calculate the cumulative distribution function:

[0071] C. Map the original grayscale value to the balanced grayscale value I eq :

[0072]

[0073] in:

[0074] N is the total number of pixels;

[0075] S max is the maximum value of the gray level;

[0076] C min is the smallest non-zero value in the cumulative distribution function;

[0077] I eq (x,y) is the grayscale value after equalization;

[0078] I gray (x,y) is the original grayscale value;

[0079] Gaussian blur processing: In order to reduce noise and enhance the characteristics of dust areas, Gaussian blur filtering is applied to the image; the formula for Gaussian blur filtering is:

[0080] The application on the image is:

[0081] in:

[0082] (x,y) is the original pixel pair;

[0083] I smoothed (x,y) is the pixel value after Gaussian blur filtering;

[0084] (i,j) is a pair of adjacent pixels;

[0085] σ is the standard deviation of the Gaussian distribution, which determines the degree of blur;

[0086] k is the radius of the filter, which is usually chosen to be an integer within 3σ.

[0087] Through these formulas, image enhancement improves contrast, while Gaussian smoothing reduces noise and makes the features of dust areas more distinct.

[0088] In order to more clearly understand the effects of image enhancement and Gaussian smoothing, a simple example is used below to illustrate their application process and effects.

[0089] Suppose a grayscale image has an uneven distribution of pixel values ​​within a certain range, such as mainly concentrated between 50 and 100, and the image looks gray as a whole. Histogram equalization can increase the contrast of the image and make the details of the gray area clearer. Suppose the pixel values ​​of the original image are [60,70,80,90,100,110,120,130]. ​​After histogram equalization, each pixel value will be redistributed to the full range (0 to 255) to increase the contrast. Suppose the pixel value distribution after equalization is: [0,32,64,96,128,160,192,224]. After histogram equalization, the pixel value distribution range of the image is wider, the image contrast is improved, and the originally unclear details will be clearer. This is helpful for dust recognition because the edges and structures of dust will be more prominent.

[0090] Gaussian blur is usually used to remove noise and make the image smooth while still retaining important structural features. Assume that the original image has noise and the grayscale value of the local area of ​​the image is:

[0091]

[0092] Assume that the Gaussian kernel (3×3) is as follows, and the standard deviation σ=1:

[0093]

[0094] Taking the central pixel as an example, we use the Gaussian kernel to perform a convolution operation on its neighborhood and calculate the smoothed pixel value:

[0095] The smoothed central pixel value is calculated. Through this smoothing operation, the pixel values ​​with large noise will be smoothly transitioned, reducing the interference of noise on dust features.

[0096] Histogram equalization enhances the contrast of the image, making the dust area more obvious. Gaussian blur smoothes the image, reduces noise, and retains the edge features of a larger area, making the dust boundary clearer. In this way, the accuracy of the dust recognition algorithm is improved through image enhancement and Gaussian smoothing.

[0097] Step 2: Feature extraction, which includes edge detection, brightness detection and blur measurement.

[0098] In step 2, edge detection: use Canny edge detection to extract image edges, and calculate the number of edges to represent edge strength E; edge strength is defined as:

[0099] Among them I edge(x, y) is a binary image, edge pixels are 1, and non-edge pixels are 0; this formula counts the number of all edge pixels; dust-covered areas can cause edges to be broken or lost, resulting in a decrease in the number of edges; the number of edges represents edge strength, and the degree of dust coverage is inversely proportional to the edge strength.

[0100] Blur detection in step 2: Use the Laplace operator to calculate the image clarity L. The higher the clarity, the less dust coverage;

[0101] Blurred areas reduce the variance of the Laplacian operator:

[0102] Where: MN is the total number of pixels in the image; ΔI gray (i, j) is the pixel gradient, which indicates the magnitude of the change in the grayscale value of the image;

[0103] ΔI gray The calculation formula for (i,j) is as follows:

[0104] in: is the rate of change of gray value in the x direction;

[0105] is the rate of change of grayscale value in the y direction.

[0106] Brightness detection in step 2: by calculating the local brightness standard deviation σ of the image local To evaluate brightness uniformity; dust coverage will cause local brightness uniformity to decrease and the standard deviation to increase, especially at the edge of the lens:

[0107]

[0108] Where μ is the average brightness of the image.

[0109] To understand these feature extraction methods more clearly, we use a simple image example to illustrate the calculation process of edge number, sharpness, and brightness distribution detection.

[0110] Suppose we have a grayscale image that is covered by dust, with some areas blurred or uneven brightness distribution.

[0111] (1). Calculate the edge strength E

[0112] Steps: First, use the Canny edge detection algorithm to extract the edge of the image.

[0113] In the generated binary edge image, edge pixels are marked as 1 and non-edge pixels are marked as 0.

[0114] Example: Assume that our detected edge image matrix is:

[0115]

[0116] Calculate the edge strength E:

[0117]

[0118] Edge strength E = 7. If the image is heavily covered with dust, the edge strength may be further reduced.

[0119] (2) Clarity L

[0120] Steps: Use the Laplacian operator to calculate the sharpness of each pixel position.

[0121] After taking the absolute value of the Laplace value of each pixel, the average value is calculated as the clarity.

[0122] Example: Assume that the image matrix (absolute value) after Laplacian operator calculation is:

[0123]

[0124] Calculate the clarity L:

[0125]

[0126]

[0127] The clarity is L = 7.25. If the dust coverage is serious, the clarity value will decrease.

[0128] (3) Brightness distribution standard deviation σ local

[0129] Steps: Calculate the average brightness μ of the image.

[0130] Calculate the square of the difference between each pixel and the average brightness, find the average, and finally take the square root to get the standard deviation of the brightness distribution.

[0131] Example: Assume the pixel values ​​of the grayscale image are:

[0132]

[0133] Calculate the average brightness μ:

[0134] Calculate the brightness distribution standard deviation σ local :

[0135]

[0136] Calculate the square of the difference between each pixel's brightness and the average brightness and sum them:

[0137]

[0138] The final brightness distribution standard deviation can be obtained through calculation.

[0139] Edge Strength E: Dust coverage will reduce edge strength, especially fine edges.

[0140] Sharpness L: Dust coverage reduces the sharpness of the image, making the average value of the Laplacian operator smaller.

[0141] Brightness distribution standard deviation σ local : Dust coverage will cause the image brightness non-uniformity to increase and the standard deviation to become larger.

[0142] Step 3: Dust area analysis: In the analysis, a dust mask needs to be generated first, and then the dust coverage index DCI is calculated.

[0143] Calculate dust coverage in step 3: Define the dust coverage index DCI, which is used to quantitatively evaluate the degree of dust coverage; DCI is calculated using the following formula:

[0144]

[0145] Among them, E ref , L ref and σ ref is the reference value collected in a clean state;

[0146] E, L, and σ local is the feature value of the current image;

[0147] ω1, ω2 and ω3 are weight coefficients of each feature, which are determined according to experiments: ω1 = 0.4, ω2 = 0.3, ω3 = 0.3.

[0148] The higher the DCI, the more severe the dust coverage.

[0149] According to the image content, the dust coverage index DCI is defined here to evaluate the dust coverage of the image. In order to explain its calculation process more clearly, we use an example to demonstrate how to calculate the DCI value.

[0150] Example

[0151] Assume we have a current image and a reference image of a clean state. Here are the values ​​for each assumption:

[0152] Feature values ​​of the clean state (reference image):

[0153] E ref =1000 (reference edge strength);

[0154] L ref=8.5 (reference clarity);

[0155] σ ref =2.0 (reference brightness standard deviation);

[0156] The eigenvalues ​​of the current image:

[0157] E = 750 (current image edge strength);

[0158] L = 6.0 (current image clarity);

[0159] σ local =3.0 (current image brightness standard deviation);

[0160] Weight coefficient:

[0161] ω1=0.4; ω2=0.3; ω3=0.3;

[0162] Calculation steps

[0163] According to the calculation formula of DCI:

[0164]

[0165] a. Calculate the difference ratio of each item:

[0166] Edge strength term:

[0167] Clarity item:

[0168] Brightness standard deviation term:

[0169] b. Weighted summation:

[0170] Substituting these results into the formula and weighting them:

[0171] DCI=(0.4×0.25)+(0.3×0.294)+(0.3×0.333)=0.2881;

[0172] The DCI value is 0.2881, and the DCI threshold is set to DCI threshold =0.1, it is relatively high, which indicates that the dust coverage of the current image is high.

[0173] If the DCI value is higher, it means that the dust coverage is more serious.

[0174] Step 4: Dust coverage judgment: set the DCI threshold. When the DCI exceeds the threshold, it indicates that the dust coverage is serious and the cleaning device is triggered.

[0175] Step 5: When the DCI exceeds the threshold, the algorithm outputs a cleaning signal and starts the cleaning device; after the cleaning device is completed, the algorithm re-analyzes the image to verify the cleaning effect; if the DCI does not meet the cleaning standard, the cleaning is repeated.

[0176] Generate a dust coverage index DCI through feature extraction and regional analysis, and judge whether to start the camera cleaning device based on this index. Detect the dust coverage on the camera lens surface and automatically generate the dust coverage index DCI. Determine whether the cleaning threshold is reached based on DCI to trigger the automatic cleaning function. Provides efficient and accurate dust recognition algorithm, suitable for long-term operation environment of industrial or outdoor cameras.

[0177] Embodiment 2

[0178] Reference Figure 2-Figure 5 A piston-type camera cleaning device comprises a camera body 1, a protective cover 2 is fixedly mounted on the surface of the camera body 1, both end surfaces of the protective cover 2 are hinged with a U-shaped base 3 through a pin shaft, a U-shaped connecting plate 4 is fixedly mounted on the inner wall of the right end of the protective cover 2, a control module 5 is fixedly mounted on the inner wall of the bottom end of the U-shaped connecting plate 4, a driving motor 6 is fixedly mounted on the vertical inner wall of the U-shaped connecting plate 4, a driving gear 7 is fixedly mounted on the output end of the driving motor 6 through a coupling, and the driving motor 6 adopts a small brushless motor.

[0179] The camera regularly collects images and transmits them to the control module 5. The image analysis algorithm of the control module 5 identifies the image clarity and dust coverage. The algorithm determines the dust accumulation by analyzing the image's blur, halo or color difference and other features. When it is detected that the dust coverage of the image exceeds a preset threshold, the control module 5 determines that cleaning is required and automatically sends a signal to activate the pneumatic piston 15 to blow air for cleaning.

[0180] The top inner wall of the U-shaped connecting plate 4 is rotatably connected to a camshaft 8 through a bearing, a driving cover 9 is fixedly installed on the top of the U-shaped connecting plate 4, a driven gear 10 is fixedly installed on the bottom end of the camshaft 8, the gear teeth of the driven gear 10 are meshed with the tooth grooves of the driving gear 7, and a cam 11 is fixedly installed on the top of the camshaft 8, and the cam 11 is located inside the driving cover 9.

[0181] A swing shaft 12 is fixedly installed on the top of the eccentric end of the cam 11, and a connecting rod 13 is rotatably connected to the arc surface of the swing shaft 12 through a bearing. A sealed pressure tube 14 is fixedly installed on one end of the driving cover 9, and one end of the connecting rod 13 extends to the inside of the sealed pressure tube 14. A pneumatic piston 15 is slidably inserted into the inner wall of the sealed pressure tube 14 through a sealing ring, and one end of the pneumatic piston 15 is hinged to the opposite end of the connecting rod 13 through a pin shaft.

[0182] The driving motor 6 drives the camshaft 8 to rotate, and the camshaft 8 controls the rotation of the cam 11. Under the condition that the connecting rod 13 is connected to the pneumatic piston 15, the rotational motion of the driving motor 6 is converted into the linear reciprocating motion of the pneumatic piston 15. The sealed pressure tube 14 provides a working chamber for the pneumatic piston 15, and the material is aluminum alloy to reduce weight and friction. The pneumatic piston 15 reciprocates in the sealed pressure tube 14, compressing the air and forming a high-pressure airflow.

[0183] A transition pipe 16 is fixedly installed at one end of the sealed pressure pipe 14, an air filter 17 is fixedly installed at the top of the transition pipe 16, an air storage tank 18 is fixedly installed at one end of the transition pipe 16, a one-way valve 19 is arranged between the air storage tank 18 and the transition pipe 16 and between the air filter 17 and the transition pipe 16, a high-pressure nozzle 20 is fixedly installed at one end of the air storage tank 18, and an electrically controlled one-way valve 21 is also arranged between the high-pressure nozzle 20 and the air storage tank 18.

[0184] When the pneumatic piston 15 moves, air is sucked into the sealed pressure pipe 14 and compressed into the air tank 18. The air tank 18 is connected to the sealed pressure pipe 14 through a one-way valve 19. When the pneumatic piston 15 moves forward, compressed air enters the air tank 18. The volume of the air tank 18 is designed to be large enough to accumulate enough high-pressure airflow in one movement of the pneumatic piston 15 for the next cleaning step.

[0185] The high-pressure nozzle 20 is connected from the air tank 18 to the surface of the camera lens, and adopts a curved conical pipe design to concentrate the airflow and improve the directionality and impact of the airflow. It can be designed according to the length of the camera. The high-pressure nozzle 20 can also adjust the airflow angle according to the cleaning needs to ensure that the high-pressure airflow effectively covers the entire lens surface. The row group of high-pressure nozzles 20 can be circular hole-shaped, duckbill-shaped, or straight-line-shaped. Set up an air filter 17: Install a micro air filter 17 at the air inlet of the transition pipe 16 to prevent dust from entering the system and ensure that the blowing airflow is clean.

[0186] Intelligent dust detection: The image recognition algorithm is used to detect dust coverage, avoiding the use of additional sensors and improving the stability and reliability of the device. Directional high-pressure cleaning: Through the design of the guide tube and the high-pressure nozzle 20, the high-pressure airflow is concentrated on the lens surface, which can effectively remove stubborn dust. Automation and energy saving: The control module 5 automatically determines the cleaning needs and triggers the movement of the pneumatic piston 15 as needed, reducing energy waste and extending the service life of the equipment.

[0187] The piston airflow technology can generate high-pressure pulse airflow in a short time, which can effectively remove stubborn dust and is suitable for such scenarios. Compared with electric fan blowing and dust removal, piston blowing has the advantages of high wind strength, high directionality, strong periodicity and high degree of automation. The airflow generated by the piston is pulsed, which can form a high-intensity directional airflow in a short time. It is suitable for tasks that require centralized cleaning, especially when removing stubborn dust particles. The piston system is more suitable for use in conjunction with an automated control system. It can be driven by a motor or pneumatic drive and cooperate with sensors to achieve a fully automatic dust removal process. For unattended heavy industrial camera equipment, the piston system is easier to integrate automated cleaning. Although the piston blowing has multiple conversion losses in energy consumption, the energy efficiency can be significantly improved by selecting high-efficiency motors, optimizing piston design, heat recovery, and gas storage utilization. In particular, intelligent control and on-demand cleaning technology can play a big role in reducing unnecessary energy consumption. The piston-type air-blowing dust removal camera is suitable for cleaning and maintenance tasks in heavy industry, outdoor monitoring, and highly polluted areas, including but not limited to ports, construction sites, mining areas, chemical plants, and outdoor monitoring stations, effectively extending the service life of the camera equipment and improving image clarity.

[0188] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A dust identification and early warning method, characterized in that: The method comprises the following steps: step 1, image preprocessing, wherein the preprocessing comprises grayscale processing, contrast enhancement processing and Gaussian blur processing; Step 2: Feature extraction, which includes edge detection, brightness detection and blur measurement; Step 3: Dust area analysis: dust mask needs to be generated first, and then the dust coverage index DCI is calculated; In the step 3, the dust coverage is calculated: a dust coverage index DCI is defined to quantitatively evaluate the dust coverage; the DCI is calculated by the following formula: Among them, E ref , L ref and σ ref is the reference value collected in a clean state; E, L, and σ local is the feature value of the current image; ω1, ω2 and ω3 are weight coefficients of each feature, determined by experiments: ω1 = 0.4, ω2 = 0.3, ω3 = 0.3; The higher the DCI, the more severe the dust coverage; Step 4: Dust coverage judgment, set the DCI threshold, when the DCI exceeds the threshold, it indicates that the dust coverage is serious, and the cleaning device is triggered; Step 5: When the DCI exceeds the threshold, the algorithm outputs a cleaning signal and starts the cleaning device. After the cleaning device is completed, the algorithm re-analyzes the image to verify the cleaning effect. If the DCI does not meet the cleaning standard, the cleaning is repeated.

2. A dust identification and early warning method according to claim 1, characterized in that: In the step 1, grayscale processing is performed by converting the collected color image I RGB Convert to grayscale image I gray ; The grayscale formula is as follows: I gray =0.2989*R+0.587*G+0.114*B; Among them: R is the red channel; G is the green channel; B is the blue channel; Contrast enhancement processing: Perform histogram equalization on the grayscale image to make the dust area clearer by adjusting the contrast of the image; Assume the gray value of the original image is I gray , histogram equalization is achieved by the following steps: A. Calculate the histogram h(t) of the grayscale image, where t represents the grayscale level; B. Calculate the cumulative distribution function: C. Map the original grayscale value to the balanced grayscale value I eq : in: N is the total number of pixels; S max is the maximum value of the gray level; C min is the smallest non-zero value in the cumulative distribution function; I eq (x,y) is the gray value after equalization; I gray (x,y) is the original grayscale value; Gaussian blur processing: In order to reduce noise and enhance the characteristics of dust areas, Gaussian blur filtering is applied to the image; the formula for Gaussian blur filtering is: The application on the image is: in: (x,y) is the original pixel pair; I smoothed (x,y) is the pixel value after Gaussian blur filtering; (i,j) is a pair of adjacent pixels; σ is the standard deviation of the Gaussian distribution, which determines the degree of blur; k is the radius of the filter, which is usually chosen to be an integer within 3σ.

3. A dust identification and early warning method according to claim 1, characterized in that: In the step 2, edge detection is performed by using Canny edge detection to extract image edges, and the number of edges is calculated to represent edge strength E; the edge strength is defined as: Among them I edge (x, y) is a binary image, edge pixels are 1, and non-edge pixels are 0; the formula counts the number of all edge pixels; dust-covered areas can cause edges to be broken or lost, resulting in a decrease in the number of edges; the number of edges represents edge strength, and the degree of dust coverage is inversely proportional to the edge strength.

4. A dust identification and early warning method according to claim 1, characterized in that: Blur detection in step 2: using the Laplace operator to calculate the image clarity L, the higher the clarity, the less dust coverage; Blurred areas reduce the variance of the Laplacian operator: Where: MN is the total number of pixels in the image; ΔI gray (i, j) is the pixel gradient, which indicates the magnitude of the change in the grayscale value of the image; ΔI gray The calculation formula for (i,j) is as follows: in: is the rate of change of gray value in the x direction; is the rate of change of grayscale value in the y direction.

5. The dust identification and early warning method according to claim 1, characterized in that: Brightness detection in step 2: by calculating the local brightness standard deviation σ of the image local To evaluate brightness uniformity; dust coverage can cause local brightness uniformity to decrease and the standard deviation to increase, such as at the edge of the lens: Where MN is the total number of pixels in the image; I gray (i, j) is the pixel value of the image in the i-th row and j-th column; μ is the average brightness of the image.

6. A piston camera cleaning device, using a dust identification and early warning method as claimed in any one of claims 1 to 5, characterized in that: The camera comprises a camera body (1), a protective cover (2) is fixedly mounted on the surface of the camera body (1), both end surfaces of the protective cover (2) are hingedly connected to a U-shaped base (3) via a pin, a U-shaped connecting plate (4) is fixedly mounted on the inner wall of the right end of the protective cover (2), a control module (5) is fixedly mounted on the inner wall of the bottom end of the U-shaped connecting plate (4), a driving motor (6) is fixedly mounted on the vertical inner wall of the U-shaped connecting plate (4), and a driving gear (7) is fixedly mounted on the output end of the driving motor (6) via a coupling; The top inner wall of the U-shaped connecting plate (4) is rotatably connected to a camshaft (8) via a bearing, a driving cover (9) is fixedly mounted on the top of the U-shaped connecting plate (4), a driven gear (10) is fixedly mounted on the bottom end of the camshaft (8), the gear teeth of the driven gear (10) are meshed with the tooth grooves of the driving gear (7), a cam (11) is fixedly mounted on the top of the camshaft (8), and the cam (11) is located inside the driving cover (9); A swing shaft (12) is fixedly mounted on the top of the eccentric end of the cam (11); a connecting rod (13) is rotatably connected to the circular arc surface of the swing shaft (12) via a bearing; a sealing pressure tube (14) is fixedly mounted on one end of the driving cover (9); one end of the connecting rod (13) extends into the interior of the sealing pressure tube (14); a pneumatic piston (15) is slidably inserted into the inner wall of the sealing pressure tube (14) via a sealing ring; one end of the pneumatic piston (15) is hinged to the opposite end of the connecting rod (13) via a pin shaft; A transition pipe (16) is fixedly mounted on one end of the sealed pressure pipe (14), an air filter (17) is fixedly mounted on the top end of the transition pipe (16), an air storage tank (18) is fixedly mounted on one end of the transition pipe (16), a one-way valve (19) is arranged between the air storage tank (18) and the transition pipe (16) and between the air filter (17) and the transition pipe (16), a high-pressure nozzle (20) is fixedly mounted on one end of the air storage tank (18), and an electrically controlled one-way valve (21) is also arranged between the high-pressure nozzle (20) and the air storage tank (18).

Images