Method and device for predicting pollution failure of 3D camera lens with air curtain

Abstract

The invention discloses a pollution fault prediction method and device for a 3D camera lens with an air curtain, and the device is characterized in that a lens pollution fault prediction part, a main lens pollution recognition sample and an environmental pollution simulation generator are disposed in a closed box body; the lens pollution fault prediction part comprises a 3d video auxiliary camera, a 3d video main camera, a lens air curtain, an auxiliary lens pollution identification sample and a pollution concentration sensor, and the relation between the lens pollution degree and the pollution time under the condition that the lens pollution fault threshold value is determined is obtained through the environmental pollution simulation generator with the adjustable pollution degree; the problem that model training data is difficult to obtain is effectively solved; through a specific air curtain structure, the two same lenses are located at positions with different pollution degrees under the condition of the same pollution source, the correlation of the pollution degrees of the two lenses is obtained, amplification processing of a pollution model is achieved, and the prediction precision of lens pollution faults is improved; the main 3D camera lens and the auxiliary 3D camera lens are adopted, so that normal work of the main 3D camera lens is not affected by lens pollution fault prediction.

Description

Technical Field

[0001] This invention relates to the field of 3D camera equipment fault prediction technology, specifically to a high-precision prediction method and device for 3D camera lens contamination faults with an air curtain. Background Technology

[0002] 3D camera lenses are widely used in industrial inspection, autonomous driving, and security monitoring, and their imaging accuracy directly affects the operational performance of equipment. In complex environments such as dust and fumes, contaminants easily adhere to the lens surface, leading to decreased light transmittance, increased imaging deviation, and contamination malfunctions. To mitigate this problem, some 3D camera lenses are equipped with air curtain devices to block contaminants from contacting the lens surface through airflow barriers; however, contamination malfunctions are still unavoidable.

[0003] The detection method disclosed in the Chinese patent publication CN118799263A entitled "A Camera Lens Dirt Detection Method Based on Self-Supervised Motion Modeling" utilizes the characteristics of lens dirt being relatively static and without obvious light and shadow changes in video, and uses the motion information contained in adjacent time frame images and the abnormal motion patterns in some areas caused by lens dirt to achieve camera lens dirt detection in general scenarios. The detection method disclosed in the Chinese patent publication CN118781110B entitled "Precise Detection Method for Lens Dirt, Terminal Equipment and Computer-Readable Storage Medium" uses algorithms to automatically detect dirt on the lens surface and can identify dirt on the lens. However, when the above two patents are applied to 3D camera lens contamination fault detection, there are two problems: (1) It is difficult to obtain training data for the lens contamination fault prediction model. Because under normal circumstances, the pollution in the air is very slight, and if the lens contamination reaches a certain threshold in such slightly polluted air, it will take a long time, and it is almost impossible to obtain a large amount of training data for the lens contamination fault prediction model. (2) It is difficult to improve the accuracy of lens contamination fault prediction. Improving the accuracy of lens contamination fault prediction cannot be achieved solely through conventional testing. Summary of the Invention

[0004] The purpose of this invention is to solve the above-mentioned problems existing in the prior art by proposing a method and device for predicting 3D camera lens contamination faults with an air curtain, which effectively solves the problem of difficulty in obtaining model training data and improves the accuracy of lens contamination fault prediction.

[0005] To achieve the above objectives, the technical solution adopted by the 3D camera lens contamination fault prediction device with air curtain of the present invention is as follows: it includes a sealed box, and the sealed box is equipped with a lens contamination fault prediction component, a main lens contamination identification sample, and an environmental contamination simulation generator. The lens contamination fault prediction component includes a 3D video secondary camera, a 3D video main camera, a lens air curtain, a secondary lens contamination identification sample, and a contamination concentration sensor. The secondary lens contamination identification sample is placed above the lens air curtain, the 3D video secondary camera and the secondary lens contamination identification sample are arranged face to face, the lens air curtain is opposite the 3D video secondary camera, the 3D video main camera is on the side of the lens air curtain in the direction of airflow, and the contamination concentration sensor is placed between the 3D video secondary camera and the secondary lens contamination identification sample. Outside the lens air curtain, the main lens pollution identification sample is placed directly opposite the 3D video main camera. The environmental pollution simulation generator is surrounded by a generator shell, inside which are a quantitative dust generating component, a multi-stage mixing chamber, and a parallel air generating component. The quantitative dust generating component can output a quantitative amount of dust. An external air intake fan is set at the output port of the quantitative dust generating component. The external air intake fan inputs the air-dust mixture into the multi-stage mixing chamber and the parallel air generating component. The air-dust mixture is first fully mixed in the multi-stage mixing chamber, and then output as parallel air after passing through the parallel air generating component. The secondary lens pollution identification sample has the same structure as the main lens pollution identification sample, both of which are divided into a sample black area and a white area. The black area and the white area are connected side by side and are fixedly connected to the cleaning air curtain above them.

[0006] Furthermore, the quantitative dust generating component has a cylinder liner, a rubber piston, and a through-type linear stepper motor. The rubber piston is sealed inside the cylinder liner and can slide along the inner wall of the cylinder liner. One end of the motor threaded shaft is connected to the center of the rubber piston, and the other end passes through the top of the cylinder liner and is coaxially connected to the through-type linear stepper motor outside the cylinder liner. The bottom of the cylinder liner is provided with dense dust leakage holes, and dust is filled between the rubber piston and the bottom of the cylinder liner. Each time the through-type linear stepper motor drives a step, it drives the motor threaded shaft to rotate, pushing the rubber piston. The rubber piston pushes the dust to leak out through the dust leakage holes.

[0007] The technical solution of the 3D camera lens contamination fault prediction method with air curtain of the present invention includes the following steps:

[0008] Step 1): The host computer acquires the gray values ​​of the black area and white area of ​​the sample for contamination identification by the main lens in a pollution-free environment;

[0009] Step 2): The through-type linear stepper motor and the external air intake fan work to make the environmental pollution simulator discharge polluted air containing dust. The pollution concentration sensor detects the air pollution concentration N(n) and transmits it to the host computer.

[0010] Step 3): Using the main 3D video camera and the secondary 3D video camera, acquire the gray values ​​of the black and white areas of the main lens contamination identification sample at time i, and the gray values ​​of the black and white areas of the secondary lens contamination identification sample at time j, respectively, and calculate the main lens contamination failure coefficient k. a (i) and secondary lens contamination failure coefficient k b (j);

[0011] Step 4): Repeat steps 2)-3) to obtain a set of primary and secondary lens pollution failure coefficients k corresponding to each air pollution concentration N(n). a (i), k b (j) The working time t of the main and secondary lenses is obtained based on the sampling interval Δt and the acquisition time of the main and secondary lenses. a t b Thus, the first prediction model was obtained. The second prediction model and the third prediction model function ;

[0012] Step 5): Let the variable Val = INT(T / Δt) to obtain the threshold value of the secondary lens contamination fault coefficient. N v This is the minimum level for light air pollution;

[0013] Step 6): Remove the main camera from the pollution identification sample and environmental pollution simulator. Place the sample to be tested opposite the 3D video main camera and obtain the current pollution concentration N through the pollution concentration sensor. d The secondary lens contamination identification sample image was acquired by the 3D video secondary camera, and the secondary lens contamination failure coefficient k was calculated. b (d) Calculate the total time required for the secondary lens to experience a contamination failure based on the second prediction model. The total time required for the main lens to experience a contamination failure, calculated using the third prediction model. The time taken for contamination to form in the secondary lens was calculated based on the second prediction model. Then, based on the third prediction model, time t bd Converted to the time it takes for the main lens to form contamination Finally, the remaining time before the main lens malfunctions due to contamination was calculated. .

[0014] The advantages of this invention using the above technical solution are:

[0015] 1) To address the difficulty in obtaining model training data, this invention uses an environmental pollution simulation generator with adjustable pollution levels to obtain the relationship between the degree of lens pollution and pollution time when the lens pollution fault threshold is determined, effectively solving the problem of difficulty in obtaining model training data.

[0016] 2) To achieve amplified processing of the contamination model, this invention utilizes a specific air curtain structure to position two identical lenses at different levels of contamination under the same contamination source—one at a more heavily contaminated location and the other at a less contaminated location. Although the contamination levels of these two lenses are completely different, a certain correlation inevitably exists between them, and the correlation between the contamination levels of the two lenses is obtained. In effect, the more heavily contaminated lens represents an amplified contamination fault model. Based on this principle, the accuracy of lens contamination fault prediction can be effectively improved.

[0017] 3) The present invention adopts a lens contamination fault prediction method based on the correlation between the main and secondary 3D camera lenses, which effectively solves the problem of real-time synchronization between real-time detection of the sample under test and lens contamination fault prediction. That is, the lens contamination fault prediction does not affect the normal operation of the main 3D camera lens at all. Attached Figure Description

[0018] Figure 1 This is a schematic diagram of the 3D camera lens contamination fault prediction device with air curtain of the present invention;

[0019] Figure 2 yes Figure 1 A magnified schematic diagram of the lens contamination fault prediction component 9 in the image;

[0020] Figure 3 yes Figure 1 An enlarged schematic diagram of the structure of the environmental pollution simulation generator 11 in the diagram;

[0021] Figure 4 yes Figure 3 Enlarged schematic diagram of the structure of the quantitative dust generating component 37 in the middle;

[0022] Figure 5 yes Figure 1 A magnified schematic diagram of the structure of sample 10 for contamination identification in the main lens;

[0023] Figure 6 This is a block diagram of the control circuit for the 3D camera lens contamination fault prediction device with air curtain of the present invention.

[0024] Figure 7 This is a schematic diagram of a lens contamination fault prediction device in actual use.

[0025] In the diagram: 1. Secondary camera bracket; 2. 3D video secondary camera; 3. 3D video main camera; 4. Host computer; 5. Lens air curtain; 6. Secondary lens contamination identification sample; 7. Contamination concentration sensor; 8. Sealed enclosure; 9. Lens contamination fault prediction component; 10. Main lens contamination identification sample; 11. Environmental pollution simulator; 12. Sample to be tested; 13. Cylinder liner bottom; 14. Cylinder liner; 15. Dust leakage hole; 16. Dust; 17. Rubber piston; 18. Cylinder liner bracket; 19. Through-type linear stepper motor; 20. Motor threaded shaft; 21. Motor bracket. 22. Generator housing; 23. Primary mixing chamber; 24. Mixing baffle; 25. Primary mixing chamber outlet; 26. Secondary mixing chamber; 27. Mixing chamber wall panel; 28. Tertiary mixing chamber; 29. ​​Tertiary mixing chamber outlet; 30. Compression chamber wall panel; 31. Baffle vent; 32. Parallel air baffle; 33. Environmental pollution simulator outlet; 34. Air-dust mixture; 35. External air intake fan; 36. Primary mixing chamber inlet; 37. Quantitative dust generation component; 38. Air and dust mixing chamber; 39. Secondary mixing chamber outlet; 40. Polluted air compression chamber; 41. Parallel air storage chamber; 42. Lens contamination identification sample fixing bracket; 43. Cleaning air curtain; 44. Identification sample black area; 45. Identification sample white area. Detailed Implementation

[0026] See Figure 1 The present invention provides a 3D camera lens contamination fault prediction device with air curtain. The exterior of the device is a sealed box 8, and inside the sealed box 8 are a lens contamination fault prediction component 9, a main lens contamination identification sample 10, and an environmental contamination simulation generator 11.

[0027] See Figure 2The lens contamination fault prediction component 9 shown comprises a secondary camera bracket 1, a 3D video secondary camera 2, a 3D video main camera 3, a host computer 4, a lens air curtain 5, a secondary lens contamination identification sample 6, and a contamination concentration sensor 7. The 3D video secondary camera 2 is mounted on the secondary camera bracket 1, and the 3D video main camera 3 is mounted above the host computer 4, which is fixedly positioned below the 3D video main camera 3. The secondary lens contamination identification sample 6 is mounted above the lens air curtain 5. The 3D video main camera 3, the 3D video secondary camera 2, and the secondary lens contamination identification sample 6 are arranged in a triangular configuration, with the secondary camera 2 and the secondary lens contamination identification sample 6 facing each other. The lens air curtain 5, located opposite the 3D video secondary camera 2, has its own fan, called the lens air curtain fan, and is connected to and controlled by the host computer 4. The air generated is directed towards the 3D video secondary camera 2 at a wind speed V. The 3D video main camera 3 is located to the side of the lens air curtain 5, facing the direction of the airflow. The contamination concentration sensor 7 is installed between the 3D video secondary camera 2 and the secondary lens contamination identification sample 6, and is located outside the airflow of the lens air curtain 5, thus unaffected by the airflow force of the lens air curtain 5. This arrangement has three advantages: First, the 3D video secondary camera 2 is subjected to more environmental contamination than the 3D video main camera 3, causing lens contamination faults to occur earlier on the 3D video secondary camera 2, and improving the granularity of contamination fault resolution; second, the sampling and fault diagnosis of the secondary lens contamination identification sample 6 by the 3D video secondary camera 2 can be carried out simultaneously with the operation of the 3D video main camera 3, and fault diagnosis does not affect the operation of the 3D video main camera 3; third, due to the effect of the lens air curtain 5, the contamination of the 3D video main camera 3 is greatly reduced, and the contamination concentration sensor 7 is unaffected by the airflow force of the lens air curtain 5, effectively reflecting the environmental contamination concentration. Although the 3D video secondary camera 2 and the 3D video main camera 3 are subjected to different levels of contamination, their lens contamination levels both increase with the increase of environmental contamination concentration and the length of time since contamination, exhibiting synchronous correlation.

[0028] Inside the sealed enclosure 8, the main lens contamination identification sample 10 is placed directly opposite the 3D video main camera 3.

[0029] See Figure 3The environmental pollution simulator 11 shown has an outer casing 22. Inside the casing 22 are a metered dust generating component 37, a multi-stage mixing chamber, and a parallel air generating component. The metered dust generating component 37 outputs a metered amount of dust. An external air intake fan 35 is installed at the output port of the metered dust generating component 37. The external air intake fan 35 forms an air-dust mixture 34 from the dust output by the metered dust generating component 37. The air-dust mixture 34 is further fed into the multi-stage mixing chamber and the parallel air generating component. The air-dust mixture 34 is first thoroughly mixed in the multi-stage mixing chamber, and then output as parallel air after passing through the parallel air generating component.

[0030] The multi-stage mixing chamber and parallel air generating unit consists of a first-stage mixing chamber 23, a second-stage mixing chamber 26, a third-stage mixing chamber 28, a polluted air compression chamber 40, and a parallel air storage chamber 41 connected in sequence. The first-stage mixing chamber 23 is equipped with a first-stage mixing chamber inlet 36 and a first-stage mixing chamber outlet 25. The first-stage mixing chamber outlet 25 communicates with the second-stage mixing chamber 26. The first-stage mixing chamber inlet 36 faces the external air intake fan 35, and the air intake direction of the first-stage mixing chamber inlet 36 is perpendicular to the air outlet direction of the first-stage mixing chamber outlet 25. Multiple rows of mixing baffles 24 are arranged inside the first-stage mixing chamber 23, and the mixing baffles 24 are parallel to the air outlet direction of the first-stage mixing chamber outlet 25. The secondary mixing chamber 26 and the tertiary mixing chamber 28 are separated by a mixing chamber wall panel 27. A secondary mixing chamber air outlet 39 is provided on the mixing chamber wall panel 27, parallel to the primary mixing chamber air outlet 25 and communicating with the tertiary mixing chamber 28. Multiple rows of mixing baffles are provided inside both the secondary mixing chamber 26 and the tertiary mixing chamber 28. The multiple rows of mixing baffles in the secondary mixing chamber 26 are perpendicular to the air outlet direction of the secondary mixing chamber air outlet 39, while the multiple rows of mixing baffles in the tertiary mixing chamber 28 are parallel to the air outlet direction of the secondary mixing chamber air outlet 39. The three-stage mixing chamber 28 and the polluted air compression chamber 40 are separated by a compression chamber wall panel 30. The polluted air compression chamber 40 and the parallel air storage chamber 41 are separated by a parallel air baffle 32. The compression chamber wall panel 30 has a three-stage mixing chamber outlet 29 that runs through the three-stage mixing chamber 28 and the polluted air compression chamber 40. The parallel air baffle 32 has several parallel baffle air holes 31 that run through the polluted air compression chamber 40 and the parallel air storage chamber 41. The generator housing 22 has an environmental pollution simulation generator outlet 33 that runs through the parallel air storage chamber 41. The environmental pollution simulation generator outlet 33 is parallel to the baffle air holes 31.

[0031] The external air intake fan 35 draws the dust output from the quantitative dust generator 37 into an air-dust mixture 34, which enters the primary mixing chamber 23 through the primary mixing chamber inlet 36. Under the action of the mixing baffle 24, the mixture is fully mixed and then enters the secondary mixing chamber 26 through the primary mixing chamber outlet 25. After being fully mixed in the secondary mixing chamber 26, the mixture enters the tertiary mixing chamber 28 through the secondary mixing chamber outlet 39. After being fully mixed again, the mixture enters the polluted air compression chamber 40 through the tertiary mixing chamber outlet 29. The mixture is then converted into parallel air through the baffle air holes 31 of the parallel air baffle 32 and enters the parallel air storage chamber 41. Finally, the parallel air containing dust pollution is discharged from the environmental pollution simulator outlet 33.

[0032] See Figure 4 The quantitative dust generating component 37 shown has a cylinder liner 14, a rubber piston 17, and a through-type linear stepper motor 19. The cylinder liner 14 is fixedly supported on a cylinder liner bracket 18. The through-type linear stepper motor 19 and the cylinder liner bracket 18 are fixedly mounted on a motor bracket 21. The rubber piston 17 is inside the cylinder liner 14, in sealed contact with the cylinder liner 14, and can slide along the inner wall of the cylinder liner 14. One end of the motor threaded shaft 20 is connected to the center of the rubber piston 17, and the other end passes through the top of the cylinder liner 14 and is coaxially connected to the through-type linear stepper motor 19 outside the cylinder liner 14. The cylinder liner bottom 13 is fixed to the cylinder liner 14, and the cylinder liner bottom 13 has dense dust leakage holes 15. Dust 16 is installed between the rubber piston 17 and the cylinder liner bottom 13. When the through-type linear stepper motor 19 drives one step, it drives the motor threaded shaft 20 to rotate, thereby pushing the rubber piston 17 forward h millimeters towards the cylinder liner bottom 13. If the pitch of the motor thread shaft 20 is S millimeters, and the through-type linear stepper motor 19 requires W steps to rotate one revolution, then h = S / W. For every h millimeters that the rubber piston 17 advances towards the bottom of the cylinder liner 13, the rubber piston 17 pushes the dust 16 to leak out through the dust hole 15. The dust 16 mixes with the air to form an air-dust mixture 34, which is then blown into the primary mixing chamber 23 by the external air intake fan 35.

[0033] See Figure 5The secondary lens contamination identification sample 6 or the primary lens contamination identification sample 10 shown are identical in structure, each divided into a black area 44 and a white area 45. The black area 44 and white area 45 are connected side-by-side and fixedly connected above each other by a cleaning air curtain 43. One side of the cleaning air curtain 43 is fixedly connected to a sample fixing bracket 42. The cleaning air curtain 43 has a built-in fan that blows away dust from the surface of the lens contamination identification sample, achieving a cleaning effect. The fan inside the cleaning air curtain 43 in the secondary lens contamination identification sample 6 is called the secondary sample cleaning air curtain fan, and the fan inside the cleaning air curtain 43 in the primary lens contamination identification sample 10 is called the primary sample cleaning air curtain fan. Both the primary and secondary sample cleaning air curtain fans are connected to and controlled by the host computer 4.

[0034] See Figure 6 The host computer 4 connects to and sends commands via communication cables to control the lens air curtain fan, the main sample cleaning air curtain fan, the secondary sample cleaning air curtain fan, the through-type linear stepper motor 19, and the external air intake fan 35. The host computer 4 also connects to the 3D video main camera 3, the 3D video secondary camera 2, and the contamination concentration sensor 7 via communication cables, receiving images from the 3D video main camera 3 and the 3D video secondary camera 2, and receiving contamination concentration information from the contamination concentration sensor 7. The lens air curtain fan is installed inside the lens air curtain 5, the main sample cleaning air curtain fan is installed inside the cleaning air curtain 43 of the main lens contamination identification sample 10, and the secondary sample cleaning air curtain fan is installed inside the cleaning air curtain 43 of the secondary lens contamination identification sample 6.

[0035] The high-precision prediction method for 3D camera lens contamination faults consists of two processes: data acquisition and prediction model establishment, and lens contamination fault prediction, as detailed below:

[0036] I. Data Acquisition and Predictive Model Establishment

[0037] 1. Acquire the initial grayscale values ​​of the main and secondary lens contamination identification sample image data in a pollution-free environment.

[0038] Referring to Figure 6, the host computer 4 sends commands via the communication cable to shut down the lens air curtain fan, the main sample cleaning air curtain fan, the auxiliary sample cleaning air curtain fan, and the external air intake fan 35, while the through-type linear stepper motor 19 is in a standby state. In a pollution-free environment (relatively free of pollution), the host computer 4 controls the 3D video main camera 3 to acquire image data of the opposite main lens contamination identification sample 10. This data is then transmitted to the host computer 4 for processing. Based on the image data, the host computer 4 calculates the grayscale value G of the black area 44 and the white area 45 of the main lens contamination identification sample 10. a10 and the gray value G of the white area a20The data is stored in memory. Simultaneously, the 3D video secondary camera acquires and processes image data of sample 6 (for secondary lens contamination identification), calculating the grayscale value G of the black area 44 and white area 45 of the sample 6 image. b10 and the gray value G of the white area b20 The grayscale value of a black and white image is calculated using the following formula (1):

[0039] (1)

[0040] R (red), G (green), and B (blue) are the brightness values ​​of the three channels of the color pixel (the value range is usually 0-255). The calculated Gray value is the gray value of the corresponding black and white pixel (also between 0-255, where 0 is pure black and 255 is pure white).

[0041] 2. Obtain pollution concentration N(n) data and main lens pollution failure coefficient k under simulated pollution environment. a (i) Data and secondary lens contamination failure coefficient k b (j) data

[0042] Step 1: Obtain pollution concentration N(n) data

[0043] The host computer 4 controls the through-type linear stepper motor 19 in the quantitative dust generating component 37 to operate, and simultaneously controls the external air intake fan 35 to operate, causing the environmental pollution simulator 11 to discharge polluted air containing trace amounts of dust, quantitatively generating a pollution concentration increment ΔN. The pollution concentration sensor 7 detects the air pollution concentration N(n) data at this time and transmits it to the host computer 4, which stores the air pollution concentration N(n) data in its memory. After each quantitative generation of a pollution concentration increment ΔN, the through-type linear stepper motor 19 and the external air intake fan 35 stop operating for one cycle.

[0044] Step 2: Obtain the main lens contamination failure coefficient k a (i) Data

[0045] The host computer 4 acquires the image data of the main lens contamination identification sample 10 at time i through the 3D video main camera 3, and calculates the gray value G of the black area of ​​the main lens contamination identification sample 10 image at time i using equation (1). a1 (i) and the gray value G of the white area a2 (i) The main lens contamination failure coefficient k is calculated from equation (2). a (i) and store it in memory:

[0046] (2)

[0047] Obtain the main lens contamination failure coefficient ka (i) After that, the host computer 4 sends a command through the communication cable to control the main sample cleaning air curtain fan to remove the dust from the surface of the main lens to identify the sample 10, and then turns it off after a delay in preparation for the next collection.

[0048] Based on the main lens contamination failure coefficient k at time i a (i) At time i>=3, the host computer 4 calculates the curvature q of the main lens contamination fault coefficient at time i using the following formula (3). a (i):

[0049] (3)

[0050] Where Δt is the time step, and k a (i-2Δt), k a (i-Δt) represent time i-2Δt and k, respectively. a The main lens contamination failure coefficient at time (i-Δt).

[0051] Determine the main lens contamination failure coefficient k a (i) Is it within a linear interval? If it is within a linear interval, continue collecting k data under different pollution concentrations. a (i) data, otherwise k a (i) Data acquisition complete. Identify the main lens contamination fault coefficient k. a (i) The method for determining whether it is in a linear interval is as follows:

[0052] Because the relationship between the lens contamination failure coefficient and time is not perfectly linear, the curvature cannot always be zero. If the curvature exceeds a set threshold, it is considered to have entered the non-linear range; this threshold is called the linear fluctuation margin. The linear fluctuation margin of the main lens contamination failure coefficient is set to v. a When the pollution failure coefficient curvature q a (i)>v a When the curvature of the lens contamination failure coefficient increases, it indicates that the coefficient has entered the nonlinear range; conversely, it remains in the linear range. When the concentration of environmental pollutants is constant, the lens contamination failure coefficient increases with time, exhibiting an approximately linear relationship within a certain time range. However, beyond a certain time range, the relationship between the lens contamination failure coefficient and time becomes nonlinear.

[0053] Step 3: Obtain the secondary lens contamination failure coefficient k b (j) data

[0054] The host computer 4 acquires image data of sample 6 for secondary lens contamination identification via the 3D video secondary camera 2, and uses this data to obtain the contamination failure coefficient k of the main lens. a (i) Using the same method, the gray value G of the black area of ​​the image of sample 6 for secondary lens contamination identification is calculated by equation (1).b1 (j) and grayscale value G b2 (j), and then the secondary lens contamination failure coefficient k is calculated by the following formula (4). b (j), and store it in memory:

[0055] (4)

[0056] Obtaining the secondary lens contamination failure coefficient k b (j) After that, the host computer 4 sends a command through the communication cable to control the secondary sample cleaning air curtain fan to remove the dust on the surface of the secondary lens and identify the sample 6. After a delay, it turns off to prepare for the next collection.

[0057] When j>=3, based on the secondary lens contamination fault coefficient k b (j), the host computer 4 calculates the curvature q of the secondary lens contamination fault coefficient at time j using the following formula (5). b (j):

[0058] (5)

[0059] Where Δt is the time step, and k b (j-2Δt), k b (j-Δt) represents time j-2Δt, k a The contamination failure coefficient of the secondary lens at time (j-Δt).

[0060] Determine the contamination failure coefficient k of the secondary lens b (j) Is it within the linear interval? If it is within the linear interval, continue collecting the secondary lens contamination failure coefficient k under different contamination concentrations. b (j) data, otherwise for k b (j) Data collection ends. The method for determining whether the secondary lens contamination fault coefficient is within the linear range is as follows: Let the linear fluctuation margin of the secondary lens contamination fault coefficient be v. b When q b (j)>v b When the curvature of the secondary lens contamination fault coefficient increases, it indicates that the lens has entered a nonlinear interval; otherwise, it is in a linear interval.

[0061] Repeat steps 1 to 3 above until the main lens contamination failure coefficient k is reached. a (i) and secondary lens contamination failure coefficient k b (j) All data collection is complete. Thus, for each air pollution concentration N(n), there is a corresponding main lens pollution fault coefficient k. a (i) and secondary lens contamination failure coefficient k b (j). The data obtained from the host computer's memory include: air pollution concentration N(n), and main lens pollution failure coefficient k.a (i) and secondary lens contamination failure coefficient k b (j), where i=1,2,…,I(n); j=1,2,…,J(n); n=1,2,…,N m The sampling interval is Δt, and I(n) and N m These represent the number of times the main and secondary lenses were sampled, respectively. The corresponding main lens working time t can be obtained from the sampling interval Δt and the main and secondary lens sampling times. a Working time of secondary lens t b .

[0062] 3. Establish a prediction model

[0063] The host computer 4 calculates the air pollution concentration N(n) and the main lens pollution failure coefficient k. a (i) and secondary lens contamination failure coefficient k b (j) Main camera working time t a Working time of secondary lens t b Establish relevant models based on the existing relationships.

[0064] For each air pollution concentration N(n), there is a corresponding main lens pollution failure coefficient k. a (i) and secondary lens contamination failure coefficient k b (j), the corresponding main shot working time t a Working time of secondary lens t b We obtain the results from equations (6) and (7) respectively:

[0065] (6)

[0066] (7)

[0067] From this, we can obtain two sets of prediction model training data: the first set is the training data F for the main lens contamination fault prediction model. a (N(n),k a (i),t a The second group is the secondary lens contamination fault prediction model F. b (N(n),k b (j),t b ). F a () represents a fully correlated data table relationship, indicating that for each air pollution concentration N(n), there is a corresponding pollution fault coefficient k. a (i), k b (j) and lens working time t b t b .

[0068] Establish a neural network prediction model: using air pollution concentration N(n) and main lens pollution failure coefficient k a (i) As input to the neural network prediction model, the main shot working time t a As the output of the neural network prediction model, the first set of data F a (N(n),k a (i),t a The neural network prediction model is trained to obtain the first prediction model function expression:

[0069] (8)

[0070] Similarly, the air pollution concentration N(n) and the secondary lens pollution failure coefficient k are used as the basis for this calculation. b (j) is used as input to the neural network prediction model, with the secondary shot working time t. b As the output of the neural network prediction model, the second set of data F b (N(n),k b (j),t b The neural network prediction model is trained to obtain the second prediction model function relationship:

[0071] (9)

[0072] Depend on Figure 1 and Figure 2 It can be seen that, since the secondary lens of the second prediction model is directly facing the wind direction of the air curtain when collecting data, while the main lens of the first prediction model is installed to the side of the wind direction when collecting data, the pollution failure coefficient k of the main lens of the first prediction model is obviously higher. a (i) The secondary lens failure coefficient k compared to the second prediction model b (j) is much smaller. However, since both prediction models were obtained under the same air pollution concentration N(n), as the air pollution concentration N(n) increases, the lens contamination failure coefficient k... a (i) and k b (j) also increases accordingly, showing a monotonically increasing function relationship. Therefore, the lens working time t b t b There is a necessary correlation between them. A set of data F corresponding to each point of air pollution concentration N(n) can be obtained through the two prediction models. ab (t a ,t b ), with data F ab (t a ,t b The neural network prediction model is trained to obtain the working time t of the secondary camera. b As input, the main camera working time ta The third predictive model function relationship is output.

[0073] (10)

[0074] 4. Determine the threshold K for the secondary lens contamination failure coefficient. valb

[0075] Under conditions of light air pollution, the lens cleaning cycle is set to T, based on a light air pollution level of 75 μg / m³. 3 -115μg / m 3 The minimum value satisfying this condition is found in the air pollution concentration N(n) data and taken as the minimum value N for light air pollution. v ,Right now:

[0076] (11)

[0077] When N is satisfied v The method for finding the corresponding fault coefficient under the given conditions is as follows:

[0078] Let the variable Val = INT(T / Δt), that is, calculate T divided by Δt, round the result down (INT), and assign the integer value to the variable Val. The second group is the secondary lens contamination fault prediction model F. b (N(n),k b (j),t b The data can be used to determine the threshold K for secondary lens contamination failure. valb :

[0079] (12)

[0080] Expression (12) expresses the result of a table lookup, meaning that in F b (N(n),k b (j),t b In the training data, when the air pollution concentration N(n) = the minimum value of light air pollution N... v Secondary lens working time t b =Val×Δt, then the threshold K for secondary lens contamination failure coefficient. valb =k b (Val). This indicates that the lens contamination failure coefficient threshold K has been determined. valb When describing the lens contamination fault prediction method later, we will directly use k. b (Val), because the second prediction model is modeled through a neural network, f b The parentheses function has a generalization effect; the input can be continuous values, while the lookup table can only use discrete integer values.

[0081] II. Lens Contamination Fault Prediction

[0082] pass Figure 1 To acquire model training data and establish a prediction model, in actual use, the main lens contamination identification sample 10 and the environmental contamination simulation generator 11 are not needed; the space where these two components are placed is used to place the test sample 12. The real-time prediction process of lens contamination faults mainly solves the problem of real-time synchronization between real-time detection of the test sample and real-time prediction of lens contamination faults. Therefore, after completing the model training data acquisition and prediction model establishment in step one, in actual use, the main lens contamination identification sample 10 and the environmental contamination simulation generator 11 are not needed; the space where these two components are placed is used to place the test sample 12. Figure 1 The main lens contamination identification sample 10 and the environmental contamination simulation generator 11, which were placed in the sealed box 8, were removed and replaced with the sample to be tested 12, while the lens contamination fault prediction component 9 remained unchanged. The sample to be tested 12 was then placed opposite the 3D video main camera 3, becoming... Figure 7 The image shows a lens contamination fault prediction device. The lens contamination fault prediction method involves collecting the current secondary lens contamination fault coefficient and the environmental contamination concentration, and then using a second prediction model. And the third prediction model To predict the timing of main lens contamination malfunctions, the specific method is as follows:

[0083] 1. Collection of secondary lens contamination failure coefficient and environmental pollution concentration.

[0084] While the 3D video main camera 3 is performing normal detection of the sample 12 under test, the host computer 4 obtains the current pollution concentration N through the pollution concentration sensor 7. d Meanwhile, the host computer 4 collects image data of the secondary lens contamination identification sample 6 through the 3D video secondary camera 2, and calculates the secondary lens contamination failure coefficient k by equation (4). b (d). The host computer 4 sends a command via the communication cable to control the secondary sample cleaning air curtain fan to remove surface dust from the secondary lens and identify sample 2. The fan then shuts off after a short delay in preparation for the next collection.

[0085] 2. The total time t required for the main lens to experience a pollution malfunction under the current environmental pollution concentration. av

[0086] Based on the previously determined threshold K for secondary lens contamination failure coefficient valb and the current environmental pollution concentration N collected d The second prediction model The total time t required for the secondary lens to experience a contamination malfunction can be calculated. bv :

[0087] (13)

[0088] Where, kb (Val)=K valb .

[0089] Then, based on the third prediction model The time t required for the secondary lens to develop a contamination malfunction bv Converted to the total time t required for the main lens to experience a contamination malfunction. av :

[0090] (14)

[0091] 3. The time t taken for the main lens to form pollution under the current environmental pollution concentration. ad

[0092] Based on the current secondary lens contamination failure coefficient k b (d) and current environmental pollution concentration N d The second prediction model The time t taken for the secondary lens to form contamination can be calculated. bd :

[0093] (15)

[0094] Then, based on the third prediction model The time t consumed in causing contamination in the secondary lens bd Converted to the time t taken for the main lens to form contamination ad :

[0095] (16)

[0096] 4. Remaining time t before the main lens experiences a contamination malfunction cl

[0097] The total time t required for the main lens to experience a contamination malfunction av Subtract the time t taken for the main lens to form contamination. ad The remaining time t for the main lens to experience a contamination malfunction is obtained. cL :

[0098] (17)

[0099] Based on the remaining time t when the main lens malfunctions. cL The timing of 3D camera lens contamination failure with air curtain was predicted.

Claims

1. A 3D camera lens contamination fault prediction device with an air curtain, comprising a sealed housing, characterized in that: The sealed box contains a lens contamination fault prediction component (9), a main lens contamination identification sample (10), and an environmental pollution simulation generator (11). The lens contamination fault prediction component (9) includes a 3D video secondary camera (2), a 3D video main camera (3), a lens air curtain (5), a secondary lens contamination identification sample (6), and a pollution concentration sensor (7). The secondary lens contamination identification sample (6) is located above the lens air curtain (5). The 3D video secondary camera (2) and the secondary lens contamination identification sample (6) are arranged face to face. The lens air curtain (5) is opposite to the 3D video secondary camera (2). The 3D video main camera (3) is located on the side of the lens air curtain (5) in the wind direction. The pollution concentration sensor (7) is located between the 3D video secondary camera (2) and the secondary lens contamination identification sample (6) and on the outside of the wind force of the lens air curtain (5). The main lens contamination identification sample (10) is placed directly opposite the (3D) video main camera (3). The environmental pollution simulator (11) is surrounded by a generator housing (22). Inside the generator housing (22) are a metered dust generating component (37), a multi-stage mixing chamber, and a parallel air generating component. The metered dust generating component (37) can output a metered amount of dust. An external air intake fan (35) is installed at the output port of the metered dust generating component (37). The external air intake fan (35) feeds the air-dust mixture into the multi-stage mixing chamber and the parallel air generating component. The air-dust mixture is first fully mixed in the multi-stage mixing chamber, and then outputs parallel air after passing through the parallel air generating component. The secondary lens contamination identification sample (6) and the main lens contamination identification sample (10) have the same structure, both divided into a sample black area and a white area. The black area and the white area are connected side by side and are fixedly connected above each other by a cleaning air curtain (43).

2. The 3D camera lens contamination fault prediction device with air curtain according to claim 1, characterized in that: The quantitative dust generating component (37) has a cylinder liner (14), a rubber piston (17), and a through-type linear stepper motor (19). The rubber piston (17) is sealed inside the cylinder liner (14) and can slide along the inner wall of the cylinder liner (14). One end of the motor threaded shaft (20) is connected to the center of the rubber piston (17), and the other end passes through the top of the cylinder liner (14) and is coaxially connected to the through-type linear stepper motor (19) outside the cylinder liner (14). The bottom of the cylinder liner is provided with dense dust leakage holes (15). Dust is filled between the rubber piston (17) and the bottom of the cylinder liner. Each time the through-type linear stepper motor (19) drives the motor threaded shaft (20) to rotate, it pushes the rubber piston (17), and the rubber piston (17) pushes the dust to leak out from the dust leakage holes (15) to the outside.

3. The 3D camera lens contamination fault prediction device with air curtain according to claim 1, characterized in that: The multi-stage mixing chamber and parallel wind generating component consists of a first-stage mixing chamber, a second-stage mixing chamber, a third-stage mixing chamber, a polluted air compression chamber (40), and a parallel wind storage chamber (41) connected in sequence. The air inlet of the first-stage mixing chamber faces the external air intake fan (35). The third-stage mixing chamber and the polluted air compression chamber (40) are separated by a compression chamber wall panel (30). The polluted air compression chamber (40) and the parallel wind storage chamber (41) are separated by a parallel wind baffle (32). The compression chamber wall panel (30) has a third-stage mixing chamber air outlet that runs through the third-stage mixing chamber and the polluted air compression chamber (40). The parallel wind baffle (32) has several parallel baffle air holes (31). The generator housing (22) has an environmental pollution simulation generator air outlet (33) that communicates with the parallel wind storage chamber (41). The environmental pollution simulation generator air outlet (33) is parallel to the baffle air holes (31).

4. The 3D camera lens contamination fault prediction device with air curtain according to claim 1, characterized in that: The secondary lens contamination identification sample (6) and the main lens contamination identification sample (10) have the same structure. The cleaning air curtain (43) has a built-in fan that can blow away the dust on the surface of the lens contamination identification sample.

5. The 3D camera lens contamination fault prediction device with air curtain according to claim 2, characterized in that: The sealed box is also equipped with a host computer (4), which is fixedly installed below the 3D video main camera (3). The host computer (4) is connected to and controls the fan of the lens air curtain, the fan of the cleaning air curtain of the main and auxiliary samples, the through linear stepper motor (19) and the external air intake fan (35) through the communication cable. The host computer (4) is also connected to the 3D video main camera (3), the 3D video auxiliary camera (2) and the pollution concentration sensor (7) through the communication cable.

6. A prediction method for a 3D camera lens contamination fault prediction device with air curtain as described in claim 5, characterized in that... Includes the following steps: Step 1): The host computer (4) acquires the gray values ​​of the black area and the white area of ​​the main lens contamination identification sample (10) in a pollution-free environment; Step 2): The through-type linear stepper motor (19) and the external air intake fan (35) work to make the environmental pollution simulation generator (11) discharge polluted air containing dust. The pollution concentration sensor (7) detects the air pollution concentration N(n) and transmits it to the host computer (4). Step 3): The black area grayscale value and white area grayscale value of the main lens contamination identification sample (10) at time i and the black area grayscale value and white area grayscale value of the secondary lens contamination identification sample (6) at time j are obtained by the 3D video main camera (3) and the 3D video secondary camera (2), respectively, and the main lens contamination fault coefficient k is calculated. a (i) and secondary lens contamination failure coefficient k b (j); Step 4): Repeat steps 2)-3) to obtain a set of main and secondary lens pollution failure coefficients k corresponding to each air pollution concentration N(n). a (i), k b (j) The working time t of the main and secondary lenses is obtained based on the sampling interval Δt and the acquisition time of the main and secondary lenses. a t b Thus, the first prediction model t is obtained. a =f a (N(n),k a (i)), the second prediction model t b =f b (N(n), k) b (j) and the third prediction model function t a =f ab (t b ); Step 5): Let the variable Val = INT(T / Δt) to obtain the threshold k of the secondary lens contamination fault coefficient. Valb =F b (N V k b (Val), Val×Δt), N v This is the minimum level for light air pollution; Step 6): Remove the main lens pollution identification sample (10) and environmental pollution simulator (11), place the sample to be tested opposite the 3D video main camera (3), and obtain the current pollution concentration N through the pollution concentration sensor (7). d The image of the secondary lens contamination identification sample (6) was acquired by the 3D video secondary camera (2), and the secondary lens contamination failure coefficient k was calculated. b (d) Calculate the total time t required for the secondary lens to experience a contamination failure based on the second prediction model. bV =f b (N d k b (Val)) is converted into the total time t required for the main lens to experience a contamination failure based on the third prediction model. aV =f ab (t bV The time t taken for contamination to form in the secondary lens is calculated based on the second prediction model. bd =f b (N d k b (d)), and then based on the third prediction model, time t bd Converted to the time t taken for the main lens to form contamination ad =f ab (t bd Finally, the remaining time t before the main lens malfunctions due to contamination is calculated. cL =t aV -t ad .

7. The prediction method according to claim 6, characterized in that: The contamination failure coefficients of the main and secondary lenses are respectively G a10 and G a20 G b10 and G b20 These are the gray values ​​of the black area and the white area of ​​the main and secondary lens contamination identification samples (10, 6) under a pollution-free environment, respectively. a1 (i) and G a2 (i), G b1 (j) and G b2 (j) are the gray values ​​of the black area and the gray value of the white area of ​​the main and secondary lens contamination identification samples (10, 6) at time i.

8. The prediction method according to claim 7, characterized in that: When i>=3, calculate the curvature of the main lens contamination fault coefficient at time i. When j>=3, calculate the curvature of the secondary lens contamination fault coefficient at time j. When q a When (i)>va, the main lens contamination failure coefficient k a (i) If it enters the nonlinear interval, otherwise it is in the linear interval. If it is in the linear interval, continue to collect k under different pollution concentration conditions. a (i), otherwise for k a (i) Data collection complete; When q b (j)>v b Secondary lens contamination failure coefficient k b (j) Entering the nonlinear interval, otherwise it is in the linear interval. If it is in the linear interval, continue to collect k under different pollution concentrations. b (j), otherwise for k b (j) Data collection ends; Δt is the time step; k a (i-2Δt), k a (i-Δt) represent time i-2Δt and k, respectively. a The main lens contamination failure coefficient at time (i-Δt); k b (j-2Δt), k b (j-Δt) represents time j-2Δt, k a The secondary lens contamination failure coefficient at time (j-Δt); Va, v b These are the linear fluctuation margins for the pollution fault coefficients of the main and secondary lenses, respectively.

9. The prediction method according to claim 6, characterized in that: In step 2), after the environmental pollution simulator (11) generates a quantitative increase in pollution concentration each time, the through-type linear stepper motor (19) and the external air intake fan (35) stop working once.

10. The prediction method according to claim 6, characterized in that: Obtaining the pollution and fault coefficients k of the main and secondary lenses a (i), k b After (j), the host computer (4) controls the fans of the main and secondary sample cleaning curtains to remove the surface dust of the corresponding main and secondary lens contamination identification samples (10, 6), and then shuts them off after a delay in preparation for the next collection.

Images