Nozzle atomization angle robust measurement method

Abstract

The invention discloses a robust measurement method for a nozzle atomization angle. The method comprises the steps of collecting a fog field grayscale image; defining a sampling window; defining fog field intensity; constructing an angle-fog field intensity histogram; calculating an initial nozzle atomization angle; obtaining the edges of the two sides of the corrected fog field by using a Canny edge detection method; obtaining a corrected nozzle center point; and obtaining an accurate nozzle atomization angle and the like. The nozzle atomization angle robust measurement method provided by the invention has the following beneficial effects: 1, integral boundary detection is realized through an intensity histogram and a gradient extreme value, and robustness is provided for a fuzzy boundary; 2, introducing a near-field clear area and Hough straight line crossing to complete center self-correction, and eliminating artificial errors; 3, a two-stage window strategy gives consideration to robustness and precision, and the average relative error is 1.72% and is obviously superior to gt; 5% of a traditional method.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of industrial spray monitoring, and particularly relates to a robust measurement method for a spray angle of a nozzle. BACKGROUND

[0002] The spray angle of a nozzle is the included angle formed by the boundary of a spray field. As a key parameter for evaluating the quality of atomization, the spray angle directly affects the application effect of industries such as pesticide spraying, spray cooling and paint spraying. However, the gas-liquid two-phase flow phenomenon existing in the nozzle jet process can cause jet instability and the boundary of the spray field to be blurred, making it difficult for conventional image processing methods to accurately measure the spray angle of the nozzle. Therefore, it is urgent to research a robust visual measurement algorithm for the spray angle of the nozzle to improve its adaptability under different environmental conditions.

[0003] Currently, the spray angle of the nozzle is mainly measured by visual test and visual measurement method, see Run Chen, Keiya Nishida, Baolu Shi, Quantitative measurement of mixture formation in an impinging spray of ethanol-gasoline blend under cold-start condition via UV-Vis dual-wavelength laser absorptionscattering (LAS) technique, Fuel, Volume 262, 2020, 116685, ISSN 0016-2361, https: / / doi.org / 10.1016 / j.fuel.2019.116685. Among them, the visual measurement method based on visible light image, by extracting the edge of the fog field in the image, obtaining the boundary of the fog field, and then calculating the spray angle of the nozzle, is the most economical and rapid method. Fu Lei et al. (Research and design of industrial nozzle spray characteristics test system[D]. Tianjin University, 2007.) preprocessed the image by median filtering, then processed the image using iterative threshold segmentation method, combined with Canny operator for edge detection, and finally calculated the angle by least square method. Wang Xiumei et al. (Research on computer measurement and control system of single nozzle flow tester[D]. Shenyang University of Technology, 2009.) used maximum variance threshold segmentation method for image segmentation, and combined with Laplace mask sharpening technology to enhance the image clarity, and finally calculated the nozzle spray angle by using the least square method.ZhangChengl et al. (H. Xuezhang and S. Zhiqiang, "Application research on digital image technology in the measurement of nozzle spray cone angle," 2011 IEEE International Instrumentation and Measurement Technology Conference, Hangzhou, China, 2011, pp. 1-5, doi: 10.1109 / IMTC.2011.5944077. keywords: {Image edge detection; Transforms; Digital images; Fuels; Noise; Computers; Cameras; digital image technology; motor nozzle; spray cone angle; measurement} removed random noise by the average method, smoothed the image using the LOG operator, and then used Hough transform for straight line detection. The spray angle of the nozzle was calculated by analyzing the slope of the straight line. Fan Xiaodong et al. (Research on Several Parameter Test Methods of Nozzle Spray Field[D]. Hebei University of Technology, 2018. DOI: 10.27105 / d.cnki.ghbgu.2018.000381.) performed gray scale transformation on the image and used Otsu threshold segmentation and morphological opening and closing operations for denoising. Finally, the edge was detected by the Canny operator, and the nozzle spray angle was calculated by fitting a straight line using the least squares method. Wu Wenfeng et al. (Measurement of Aviation Engine Fuel Nozzle Spray Angle[J]. Aviation Engine, 2017, 43(05): 69-73. DOI: 10.13477 / j.cnki.aeroengine.2017.05.012.) preprocessed the image by denoising and binarization, extracted its edge image, and then fitted the boundary of the edge image based on the least squares method. Bingbing Liu et al. (Study on Diesel Spray Characteristics Based on Image Processing[C] / / Shandong Jiaotong University (China), 2023:) enhanced image information using contrast stretching, removed noise using median filtering, performed threshold segmentation using the maximum inter-class variance method (OTSU), and detected the spray edge using the Canny operator. Finally, the nozzle spray angle was calculated using the precise point coordinate method.However, under the combined action of gas-liquid two-phase flow, air shear stress and gravity, the liquid droplets sprayed by the nozzle inside the porous spray form a discontinuous fluid field, and the image shooting of the mist field is easily affected by the background light, causing the mist field edge to have strong uncertainty. See Liu Qi, et al. Macroscopic Spray Characteristics of Centrifugal Nozzle for Aeroengine [J]. Propulsion Technology, 2021, 42(02): 362-371. DOI: 10.13675 / j.cnki.tjjs.190592. Figure 1 Two typical mist field gray scale images are shown in (a) and (b), respectively. The mist field edge of the small-angle nozzle shown in (a) is relatively clear, while the mist field edge of the wide-angle nozzle shown in (b) is discontinuous. The traditional image processing method based on edge extraction and region segmentation will increase the measurement error of the nozzle atomization angle, or even cause the measurement to fail. See L. Yang, S. Ren and H. Duan, "Detection System of Fuel Spray Nozzle Based on Machine Vision," 2020 IEEE 9th Joint International Information Technology and Artificial Intelligence Conference (ITAIC), Chongqing, China, 2020, pp. 1152-1156, doi: 10.1109 / ITAIC49862.2020.9339155.

[0004] In addition, the prior art generally uses Canny, LOG or Otsu edge detection algorithm to extract the mist field boundary, and then uses least squares method or Hough transform to calculate the nozzle atomization angle. However, under the conditions of high-frequency jitter, blurred boundary, changing light and wide-angle working condition, the above method has the following defects:

[0005] 1. Edge breakage or missed detection, resulting in failure of nozzle atomization angle calculation;

[0006] 2. Sensitive to the error of the center position of the nozzle, and a small shift of the center can cause angle deviation;

[0007] 3. Cannot fully utilize the overall gray scale distribution characteristics of the mist field, and the robustness is insufficient. SUMMARY

[0008] In order to solve the above problems, the purpose of the present application is to provide a robust nozzle atomization angle measurement method.

[0009] In order to achieve the above purpose, the robust nozzle atomization angle measurement method provided by the present application comprises the following steps in sequence:

[0010] 1) Use an industrial camera to collect a gray-scale image I(x, y) of the mist field along a direction perpendicular to the central axis of the mist field;

[0011] 2) Set an initial nozzle center point P as the rotation center, define two strip-shaped sampling windows W(α) with the initial nozzle center point P as the midpoint of the short side, a width of W, a length of L, and a long side direction angle of α, and the rotation angle range of the sampling window W(α) is 0°-180°;

[0012] 3) Based on the above sampling window W(α) and the gray-scale image I(x, y) of the mist field, define the mist field intensity S α ;

[0013] 4) Based on the above mist field intensity S α , construct an angle-mist field intensity histogram H;

[0014] 5) Use the central difference method to calculate the first-order gradient of the above angle-mist field intensity histogram H, find the long side direction angles corresponding to the maximum and minimum values of the first-order gradient, and use the two long side direction angles to calculate the initial nozzle atomization angle;

[0015] 6) In the nozzle vicinity area, use the Canny edge detection method to obtain the two side edges of the corrected mist field;

[0016] 7) Use Hough transform to perform straight line fitting on the two side edges of the corrected mist field to obtain left and right boundary straight lines L1, L2, then calculate the intersection point of the left and right boundary straight lines L1, L2 and take it as the corrected nozzle center point Pc;

[0017] 8) Take the corrected nozzle center point Pc as the center, reduce the length L' of the strip-shaped sampling window L' < L, repeat steps 2) to 5), and use the mist field intensity histogram edge detection method with more optimal parameters to obtain the corrected long side direction angles α1', α2' and take them as the two side edges of the corrected mist field. The difference between the two corrected long side direction angles α1', α2' is the accurate nozzle atomization angle.

[0018] In step 3), the method for defining the mist field intensity S α is:

[0019] Calculate the gray-scale cumulative value in the sampling window W(α) corresponding to a long side direction angle α in the gray-scale image I(x, y) of the mist field and take it as the mist field intensity S α of the long side direction angle α, the formula is as follows:

[0020] S α =∫ W(α) I(x,y)dxdy.

[0021] In step 4), the fog field intensity S α The method for constructing the angle-fog field intensity histogram H is as follows:

[0022] In the range of 0°-180°, the gray value accumulated in the sampling window W(α) corresponding to each long-side direction angle α is calculated as the fog field intensity S α of the sampling window W(α) corresponding to each long-side direction angle α and the corresponding fog field intensity S α The angle-fog field intensity histogram H is constructed, and the formula is as follows:

[0023]

[0024] The angle-fog field intensity histogram H takes the long-side direction angle as the abscissa and the fog field intensity as the ordinate.

[0025] In step 5), the method for calculating the first-order gradient of the angle-fog field intensity histogram H by using the central difference method, finding the long-side direction angles corresponding to the maximum and minimum values of the first-order gradient as the two side edges of the fog field, and then calculating the initial nozzle atomization angle by using the two long-side direction angles is as follows:

[0026] The first-order gradient of the angle-fog field intensity histogram H is calculated by using the central difference method, the long-side direction angles α1 and α2 corresponding to the maximum and minimum values of the first-order gradient are found as the two side edges of the fog field, and the difference between the two long-side direction angles α1 and α2 is the initial nozzle atomization angle.

[0027] In step 6), the method for obtaining the two side edges of the corrected fog field in the region near the nozzle by using the Canny edge detection method is as follows:

[0028] The two side edges of the fog field near the nozzle are detected by using the Canny edge detection method, a 3x3 Sobel kernel is used for edge detection in the fog field near the nozzle to remove possible interference edges, and the two side edges of the corrected fog field are obtained.

[0029] The nozzle atomization angle robust measurement method provided by the application has the following beneficial effects:

[0030] 1. The integral boundary detection is realized by "intensity histogram + gradient extreme value", and the fuzzy boundary is robust;

[0031] 2. The "near-field clear area + Hough line intersection" is introduced to complete the center self-correction, and the artificial error is eliminated;

[0032] 3. The two-stage window strategy considers the robustness and accuracy, and the average relative error is 1.72%, which is significantly better than the traditional method of >5%. BRIEF DESCRIPTION OF DRAWINGS

[0033] Figure 1 Two typical fog field gray scale images, (a) small angle nozzle, (b) wide angle nozzle.

[0034] Figure 2 Flow chart of the robust measurement method of nozzle atomization angle provided by the present application.

[0035] Figure 3 Schematic diagram of fog field intensity defined in the present application.

[0036] Figure 4 Angle-fog field intensity histogram in the present application.

[0037] Figure 5 Contrastive diagram of nozzle center point in the present application.

[0038] Figure 6 Nozzle atomization angle detection result in the present application.

[0039] Figure 7 Test result contrast of two typical nozzles in the present application.

[0040] Figure 8 Atomization angle measurement error curve of 12 nozzles in the present application. DETAILED DESCRIPTION

[0041] The present application will be described in detail below in combination with the drawings and specific embodiments.

[0042] As shown in Figure 2 , the robust measurement method of nozzle atomization angle provided by the present application comprises the following steps performed in sequence:

[0043] 1) Collecting fog field gray scale image I(x, y) along the direction perpendicular to the central axis of the fog field by using an industrial camera;

[0044] 2) As shown in Figure 3 , setting an initial nozzle center point P as the rotation center, defining two strip-shaped sampling windows W(α) with the initial nozzle center point P as the midpoint of the short side, width W and length L, and the long side direction angle α, and the rotation angle range of the sampling window W(α) is 0°-180°;

[0045] 3) Defining fog field intensity S α based on the above sampling window W(α) and fog field gray scale image I(x, y);

[0046] Calculating the gray scale cumulative value in the sampling window W(α) corresponding to a long side direction angle α in the fog field gray scale image I(x, y) and taking it as the fog field intensity S α of the long side direction angle α, and the formula is as follows:

[0047] Sα = ∫ W(α) I(x,y)dxdy

[0048] By definition, the mist field intensity S α is the total mist field intensity in the strip sampling window W(α) with the long side direction angle α, which can amplify the mist field region signal. The length L determines the range of the mist field intensity calculation. The wider the width W, the more reliable the mist field intensity evaluation, but the weaker the direction indication. In the initial edge detection, a larger width W can be selected to enhance the robustness of the detection. Conversely, in the accurate measurement of the nozzle atomization angle, the width W is smaller.

[0049] 4) Based on the above mist field intensity S α , the angle-mist field intensity histogram H is constructed:

[0050] The position of the mist field edge can be determined by the difference in the mist field intensity distribution in each direction. Rotate the sampling window W(α) in the range of 0°-180°, calculate the gray value accumulation in the sampling window W(α) corresponding to each long side direction angle α and take it as the mist field intensity S α of the long side direction angle α. Use all long side direction angles α and the corresponding mist field intensity S α to construct the angle-mist field intensity histogram H as shown in Figure 4 , and the formula is as follows:

[0051]

[0052] The angle-mist field intensity histogram H takes the long side direction angle as the horizontal coordinate and the mist field intensity as the vertical coordinate.

[0053] 5) Use the central difference method to calculate the first-order gradient of the above angle-mist field intensity histogram H, find the long side direction angles corresponding to the maximum and minimum values of the first-order gradient and take them as the two side edges of the mist field, and then calculate the initial nozzle atomization angle using the two long side direction angles;

[0054] As shown in Figure 4 , on the distribution curve of the angle-mist field intensity histogram H, the mist field intensity changes sharply at the two side edges of the mist field. Therefore, the present application uses the central difference method to calculate the first-order gradient of the angle-mist field intensity histogram H, finds the long side direction angles α1, α2 corresponding to the maximum and minimum values of the first-order gradient and takes them as the two side edges of the mist field, and the difference between the two long side direction angles α1, α2 is the initial nozzle atomization angle.

[0055] 6) In the nozzle vicinity region, use the Canny edge detection method to obtain the corrected two side edges of the mist field;

[0056] As shown in Figure 5As shown in the figure, if there is a deviation in the manually set initial nozzle center point P, when the initial nozzle center point P is biased upward, the calculated nozzle atomization angle will become smaller; when the initial nozzle center point P is biased downward, the calculated nozzle atomization angle will become larger. It can be seen that the deviation of the initial nozzle center point P will cause deviations in the two side edges of the fog field, and finally cause errors in the nozzle atomization angle. Therefore, in actual operation, it is necessary to correct the initial nozzle center point P.

[0057] As the fog field spreads, under the action of air resistance, turbulent perturbation and gravity, its movement trajectory will deviate from the central axis of the fog field and form a Gaussian distribution, as Figure 1 shown. At the same time, the characteristics of the two side edges of the fog field may have a large fuzzy transition area, which directly makes it difficult for traditional edge extraction methods to detect the two side edges of the fog field.

[0058] However, observing Figure 1 the distribution of the fog field, it can be found that although the transition area of the two side edges of the fog field is large, due to the large injection intensity near the nozzle, the two side edges of the fog field at this place are relatively clear.

[0059] Therefore, the Canny edge detection method can be used to detect the two side edges of the fog field near the nozzle. Since the Canny edge detection method can effectively identify the boundaries of the rectangular area, in this invention, a 3x3 Sobel kernel is used for edge detection in the fog field near the nozzle to remove possible interfering edges and obtain the two side edges of the corrected fog field.

[0060] 7) Use the Hough transform to perform linear fitting on the two side edges of the corrected fog field to obtain the left and right boundary lines L1 and L2, and then calculate the intersection point of the left and right boundary lines L1 and L2 and use it as the corrected nozzle center point Pc;

[0061] 8) Take the corrected nozzle center point Pc as the center, reduce the length L' of the strip sampling window to L' < L, and repeat steps 2) to step 5), iteratively use the fog field intensity histogram edge detection method with better parameters to obtain the corrected long side direction angles α1' and α2' and use them as the two side edges of the corrected fog field. The difference between the two corrected long side direction angles α1' and α2' is the accurate nozzle atomization angle.

[0062] Experiment

[0063] In order to verify the effectiveness of the method of the present application, the present inventors carried out the following nozzle atomization angle detection experiment. The algorithm test was carried out using python 3.12, and the algorithm was implemented using VC++ programming tool and OpenCV library. The nozzle to be tested was installed downward, and the center axis of the mist field was basically perpendicular to the ground. An industrial camera was used to obtain a mist field gray scale image. According to the test requirements, the industrial camera was placed horizontally, located 1 meter below the nozzle, about 2 meters away from the center axis of the mist field, and the resolution of the mist field gray scale image was 1600x1200.

[0064] Typical nozzle test

[0065] The present inventors carried out nozzle atomization angle measurement experiments on three typical types of nozzles X, nozzle Y and nozzle Z. According to the comparison of multiple experiments, the optimized values of the length L and width W of the sampling window are 380 and 22 respectively. Therefore, in the experiment, the width W1, W2 of the sampling window is 22, the length L1, L2 is 380, and the long side direction angle a1, a2 is 1°. The mist field gray scale image in the experiment is shown in Figure 6 As shown in the figure, the droplets of the mist field of nozzle X are larger, the intensity of the mist field of nozzle Y is lower, and the edges of the mist field of nozzle Z are relatively blurred.

[0066] The experimental results are shown in Figure 6 The first row shows the mist field intensity histogram, the second row shows the labeled mist field edges and nozzle center, and the third row shows the corrected mist field intensity histogram and mist field edge position marker. As can be seen from the figure, the method of the present application can correctly process the mist field gray scale images of the three nozzles, successfully extract the edges of the mist field, correct the nozzle center point, and finally obtain the accurate nozzle atomization angle. The experimental results show that the method of the present application can correctly measure the atomization angle of the typical nozzle.

[0067] Robustness test

[0068] In order to verify the robustness of the method of the present application, the present inventors selected 12 standard nozzles, including 10°, 40°, 45°, 50°, 60°, 65°, 75°, 80°, 90°, 100°, 120° and 150°, for nozzle atomization angle measurement experiments of the present application, and compared and analyzed the method of the present application with four typical measurement methods: algorithm 1 (Fan Xiaodong method), algorithm 2 (Fu Lei method), algorithm 3 (Wang Xiumei method) and algorithm 4 (Zhang Changli method).

[0069] Figure 7The test results of two typical nozzles (45° and 150°) are shown. For the gray-scale image of the spray field of the 45° nozzle (the boundary is clearer), the two-side edges of the spray field can be effectively extracted by the method of the present application and the four comparative algorithms, but the edge precision of the method of the present application is higher. In the 150° nozzle test, the two-side edges of the spray field can be correctly detected by the algorithms 1 and 2, the detection result of the algorithm 3 is completely wrong, and the algorithm 4 only correctly identifies the left-side edge, and the measurement deviation of the nozzle atomization angle is 4.12°, 1.20°, 0.66° and 27.43° respectively, while the detection of the two-side edges by the method of the present application is accurate, and the measurement result is 149.85°, and the deviation is only 0.15°.

[0070] Figure 8 The absolute distance error measurement results of the method of the present application and the four comparative algorithms on 12 nozzles are shown. The experimental results show that the method of the present application realizes the maximum absolute error of 2°, the maximum relative error of 4.17% and the average relative error of 1.72% under different nozzle angles, which is significantly better than the 5% error threshold specified in GB5135.3-2003 standard. In contrast, the errors of the four comparative algorithms are significantly higher and fluctuate greatly, indicating that their robustness is poor. Taking 5% error as the failure standard, the four comparative algorithms fail 4 times, 4 times, 3 times and 3 times respectively, while the method of the present application successfully measures the atomization angle of all nozzles, which fully verifies the robustness of the method of the present application.

Claims

1. A robust method for measuring the nozzle atomization angle, characterized in that: The robust measurement method for the nozzle atomization angle includes the following steps carried out in sequence: 1) Use an industrial camera to collect the gray-scale image I(x, y) of the fog field along the direction perpendicular to the central axis of the fog field; 2) Set the initial nozzle center point P as the rotation center, and define two strip sampling windows W(α) with the initial nozzle center point P as the midpoint of the short side, width W, length L, and long side direction angle α. The rotation angle range of the sampling window W(α) is 0°–180°; 3) Based on the above sampling window W(α) and the grayscale image of the fog field I(x, y), the fog field intensity S is defined. α ; 4) Based on the above fog field intensity S α Construct an angle-fog field intensity histogram H; 5) Use the central difference method to calculate the first-order gradient of the above-mentioned angle-fog field intensity histogram H, find the long side direction angles corresponding to the maximum and minimum values of its first-order gradient and use them as the two side edges of the fog field. Then calculate the initial nozzle atomization angle using the two long side direction angles; 6) In the area near the nozzle, use the Canny edge detection method to obtain the two side edges of the corrected fog field; 7) Use the Hough transform to perform linear fitting on the two side edges of the corrected fog field to obtain the left and right boundary lines L1, L2, and then calculate the intersection point of the left and right boundary lines L1, L2 and use it as the corrected nozzle center point Pc; 8) With the corrected nozzle center point Pc as the center, reduce the length L′<L of the strip sampling window, and repeat steps 2) to step 5), iteratively use the fog field intensity histogram edge detection method with better parameters to obtain the corrected long side direction angles α1′, α2′ and use them as the two side edges of the corrected fog field. The difference between the two corrected long side direction angles α1′, α2′ is the accurate nozzle atomization angle.

2. The robust measurement method for nozzle atomization angle according to claim 1, characterized in that: In step 3), the fog field intensity S is defined. α The method is: Calculate the cumulative grayscale value within the sampling window W(α) corresponding to a certain long-side direction angle α in the grayscale image I(x,y) of the fog field, and use it as the fog field intensity S at that long-side direction angle α. α The formula is as follows: S α =∫ W(α) I(x,y)dxdy。 3. The robust measurement method for nozzle atomization angle according to claim 1, characterized in that: In step 4), the fog field intensity S mentioned above... α The method for constructing the angle-fog field intensity histogram H is as follows: Rotate the sampling window W(α) within the range of 0°–180°, calculate the cumulative gray value within the sampling window W(α) for each long side direction angle α, and use this as the fog field intensity S for that long side direction angle α. α Using all the long side direction angles α and the corresponding fog field intensity S α The angle-fog field intensity histogram H is constructed using the following formula: The angle-fog field intensity histogram H uses the long side direction angle as the abscissa and the fog field intensity as the ordinate.

4. The robust measurement method for nozzle atomization angle according to claim 1, characterized in that: In step 5), the method of using the central difference method to calculate the first-order gradient of the above-mentioned angle-fog field intensity histogram H, finding the long side direction angles corresponding to the maximum and minimum values of its first-order gradient and using them as the two side edges of the fog field, and then calculating the initial nozzle atomization angle using the two long side direction angles is: Use the central difference method to calculate the first-order gradient of the angle-fog field intensity histogram H, find the long side direction angles α1, α2 corresponding to the maximum and minimum values of its first-order gradient and use them as the two side edges of the fog field. The difference between the two long side direction angles α1, α2 is the initial nozzle atomization angle.

5. The robust measurement method for nozzle atomization angle according to claim 1, characterized in that: In step 6), the method of using the Canny edge detection method to obtain the two side edges of the corrected fog field in the area near the nozzle is: Use the Canny edge detection method to detect the two side edges of the fog field near the nozzle. Use a 3x3 Sobel kernel for edge detection in the fog field near the nozzle to remove possible interfering edges and obtain the two side edges of the corrected fog field.

Images