A method for detecting heat transfer efficiency of hot air duct based on infrared images

Abstract

The present invention belongs to the field of image processing technology, and specifically relates to a method for detecting the heat transfer efficiency of a hot air duct based on an infrared image, the method comprising: decomposing an infrared image to obtain downsampled sub-images at different resolutions and detail sub-images in each direction, performing Hough line detection on foreground pixels in a binary image of the downsampled sub-image, calculating the probability that the obtained straight line belongs to stripe noise, recording a straight line with a probability greater than a preset threshold as a target straight line, obtaining the neighborhood range of each pixel on the target straight line according to the distance and probability between adjacent target straight lines, filtering the downsampled sub-images and detail sub-images in each direction according to the neighborhood range, performing wavelet reconstruction on the filtered downsampled sub-images and detail sub-images in each direction, obtaining a denoised infrared image, and then calculating the heat transfer efficiency of the hot air duct. The present invention effectively removes the stripe noise of the infrared image, and can more accurately analyze the heat transfer efficiency of the hot air duct.

Description

Technical Field

[0001] The present invention relates to the field of image processing technology, and more specifically, to a method for detecting heat transfer efficiency of a hot air duct based on infrared images. Background Art

[0002] When the hot air duct is conveying hot air, if the heat transfer efficiency is low, it indicates that the duct has leakage, insulation material damage or other faults; by monitoring the heat transfer efficiency of the hot air duct, problems can be discovered early and maintenance costs can be reduced.

[0003] Infrared imaging technology can measure temperature distribution in a non-contact manner without physical contact with the air duct. This method is not only safe, but also can effectively reduce interference during the measurement process. Among them, stripe noise interferes with the detection and extraction of useful information in infrared images. Therefore, it is necessary to remove the stripe noise in the infrared image to improve the reliability of subsequent processing and analysis results.

[0004] In the related technology, for example, the Chinese patent application document with application publication number CN118195943A discloses a method for removing stripe noise from infrared images, including: smoothing the original image with a low-pass filter to obtain a low-frequency component image; obtaining a weight matrix of filtered high-frequency weighted image data; subtracting the low-frequency component image from the original image to obtain a high-frequency component image: traversing each column of the high-frequency component image; dividing each column into several segments; screening each segment of each column for high-frequency values ​​that meet the stripe noise weight threshold, and accumulating and summing them respectively; calculating the stripe noise value of each segment of each column by averaging; selecting the minimum stripe noise value of a uniform scene area; obtaining the final stripe noise array of each column; subtracting the corresponding stripe noise from each column of the original image to obtain a denoised image.

[0005] The related technology uses a frequency domain method. When the frequency of the stripe noise overlaps with the frequency of the effective information, the stripe noise cannot be completely removed, and even causes the loss of effective information. The processing effect of the stripe noise is not ideal. Summary of the invention

[0006] In order to solve the technical problem that the frequency domain method is not ideal for processing stripe noise, the present invention provides a method for detecting heat transfer efficiency of a hot air duct based on infrared images, comprising: collecting an infrared image of the hot air duct by an infrared focal plane imaging device; decomposing the infrared image by wavelet transform to obtain down-sampled sub-images at different resolutions and detail sub-images in various directions; performing threshold segmentation on the edge image of the down-sampled sub-image to obtain a binary image; performing Hough line detection on all foreground pixels in the binary image to obtain multiple straight lines in a target direction, wherein the target direction refers to a direction perpendicular to the arrangement direction of the amplifiers in the infrared focal plane imaging device; and calculating the probability that the straight line belongs to stripe noise. , , is the number of all foreground pixels on the line, is the length of the edge of the infrared image along the target direction, is the difference in the gradient amplitude of all foreground pixels on the straight line, For the straight line The angle between the gradient direction of the foreground pixel and the target direction; the straight line with a probability greater than a preset threshold is recorded as the target line; the neighborhood range of each foreground pixel on the target line is obtained according to the distance between adjacent target lines and the probability that the target line belongs to stripe noise; according to the neighborhood range of each foreground pixel, the down-sampled sub-image and the detail sub-image in each direction are filtered; the denoised infrared image is obtained by wavelet reconstruction of the filtered down-sampled sub-image and the filtered detail sub-image in each direction; the heat transfer efficiency of the hot air duct is calculated according to the denoised infrared image.

[0007] The present invention utilizes the multi-resolution characteristics of wavelet transform to process images at multiple scales and directions, so that details and edge information are clearer, thereby effectively identifying noise and signals of different frequency components; through Hough line detection, the straight line features related to stripe noise can be accurately identified, and combined with the evaluation result of the probability that the straight line belongs to stripe noise, the noise and valid information can be effectively distinguished; for the foreground pixel points on the target straight line, according to the distance between adjacent target straight lines and the probability that the target straight line belongs to stripe noise, the neighborhood range of the foreground pixel points is obtained, and the down-sampled sub-image and the detail sub-images in each direction are filtered according to the neighborhood range, so that the information of the surrounding pixel points can be comprehensively considered during denoising, so as to better retain useful information and remove noise; through wavelet reconstruction, the filtered sub-images are effectively merged into the final infrared image, important features are retained, and the influence of noise is reduced.

[0008] The present invention effectively removes stripe noise, makes infrared images clearer, and improves the readability of the images, thereby improving the reliability of subsequent processing and analysis results, and can more accurately analyze the heat transfer efficiency of hot air ducts, thereby improving accuracy and reliability in industrial applications.

[0009] Preferably, the binary image includes foreground pixels and background pixels, the foreground pixels refer to pixels whose gradient amplitude is greater than or equal to a gradient amplitude threshold, and the background pixels refer to pixels whose gradient amplitude is less than a gradient amplitude threshold, and the gradient amplitude threshold is obtained by the Otsu method.

[0010] Preferably, the neighborhood range of the foreground pixel point is composed of the foreground pixel point on the left side. pixels and the adjacent pixels on the right side of the foreground pixel pixels, where , are the first quantity and the second quantity respectively.

[0011] Preferably, the first number and the second number satisfy the expression: ; ; In the formula, is the target straight line where the foreground pixel is located, For the target straight line Left and in a straight line from the target The nearest target straight line, For the target straight line Right and in a straight line from the target The nearest target straight line, is the distance function, , are the first coefficient and the second coefficient respectively, To round up.

[0012] Preferably, the first coefficient and the second coefficient satisfy the expression: ; ; In the formula, The target straight line The probability of being stripe noise, The target straight line The probability of being stripe noise.

[0013] The present invention obtains the neighborhood range of each foreground pixel point on the target straight line according to the probability that the target straight line belongs to stripe noise, so that the pixel points included in the neighborhood range are less affected by the noise, thereby improving the effect of noise removal.

[0014] Preferably, filtering the downsampled sub-image and the detail sub-images in each direction includes: recording the foreground pixel points on the target straight line as the target pixel points; for the pixel point A in the downsampled sub-image with the same coordinates as the target pixel point, recording the pixel points adjacent to the left of the pixel point A pixels and the pixels adjacent to the right of pixel A Pixels constitute the neighborhood range of pixel A; for pixel B with the same coordinates as the target pixel in the detail sub-images in each direction, the neighboring pixels on the left side of pixel B are pixels and the pixels adjacent to the right of pixel B Pixels constitute the neighborhood range of pixel B; according to the neighborhood range of each pixel, the downsampled sub-image and the detail sub-images in each direction are filtered.

[0015] The present invention filters the down-sampled sub-images and detail sub-images in various directions according to the neighborhood range, comprehensively considers the information of surrounding pixels, thereby better retaining useful information and removing noise.

[0016] Preferably, the algorithm used for filtering is the median average filtering method, including: for the grayscale values ​​of all pixels within the neighborhood of the pixel point, removing the largest grayscale value and the smallest grayscale value, and then calculating the arithmetic mean of all remaining grayscale values ​​as the filtering result of the pixel point.

[0017] The present invention performs filtering through a median average filtering method, which can not only suppress random interference but also filter out obvious pulse interference.

[0018] Preferably, the infrared image of the hot air duct includes an infrared image of the air inlet and an infrared image of the air outlet of the hot air duct, and the denoised infrared image includes a denoised infrared image of the air inlet and an infrared image of the air outlet of the hot air duct.

[0019] Preferably, the heat transfer efficiency of the hot air duct is calculated according to the denoised infrared image, comprising: inputting the denoised infrared image into FLIR Tools software to generate a temperature value file of all pixels of the denoised infrared image; the heat transfer efficiency of the hot air duct satisfies the expression: ; In the formula, is the heat transfer efficiency of the hot air duct, is the mean temperature value of all pixels in the denoised infrared image of the air outlet, It is the mean temperature value of all pixels in the denoised infrared image of the air inlet.

[0020] The present invention effectively removes stripe noise, makes infrared images clearer, and improves the readability of the images, thereby improving the reliability of subsequent processing and analysis results, and can more accurately analyze the heat transfer efficiency of hot air ducts, thereby improving accuracy and reliability in industrial applications.

[0021] The beneficial effects of the present invention are:

[0022] The present invention utilizes the multi-resolution characteristics of wavelet transform to process images at multiple scales and directions, so that details and edge information are clearer, thereby effectively identifying noise and signals of different frequency components; through Hough line detection, the straight line features related to stripe noise can be accurately identified, and combined with the evaluation result of the probability that the straight line belongs to stripe noise, the noise and valid information can be effectively distinguished; for the foreground pixel points on the target straight line, according to the distance between adjacent target straight lines and the probability that the target straight line belongs to stripe noise, the neighborhood range of the foreground pixel points is obtained, and the down-sampled sub-image and the detail sub-images in each direction are filtered according to the neighborhood range, so that the information of the surrounding pixel points can be comprehensively considered during denoising, so as to better retain useful information and remove noise; through wavelet reconstruction, the filtered sub-images are effectively merged into the final infrared image, important features are retained, and the influence of noise is reduced.

[0023] Furthermore, the present invention effectively removes stripe noise, making infrared images clearer and improving image readability, thereby improving the reliability of subsequent processing and analysis results, and can more accurately analyze the heat transfer efficiency of hot air ducts, thereby improving accuracy and reliability in industrial applications. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] The above and other purposes, features and advantages of the exemplary embodiments of the present invention will become readily understood by reading the detailed description below through the accompanying drawings. In the accompanying drawings, several embodiments of the present invention are shown in an exemplary and non-limiting manner, and the same or corresponding reference numerals represent the same or corresponding parts, wherein:

[0025] Figure 1 is a flow chart schematically illustrating a method for detecting heat transfer efficiency of a hot air duct based on infrared images in the present invention;

[0026] Figure 2 It schematically shows the decomposition and reconstruction process of wavelet transform. DETAILED DESCRIPTION

[0027] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0028] The specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0029] The embodiment of the present invention discloses a method for detecting the heat transfer efficiency of a hot air duct based on infrared images, referring to Figure 1 , including steps S1 to S6:

[0030] S1. Collect infrared images of hot air ducts.

[0031] Specifically, infrared focal plane imaging devices are installed at the air inlet and outlet of the hot air duct to collect infrared images of the air inlet and outlet of the hot air duct. The infrared images are grayscale images, so pixels in the infrared images have grayscale values.

[0032] It should be noted that an infrared focal plane imaging device generally contains multiple amplifiers. When the multiple amplifiers are arranged in the horizontal direction, a column of outputs of the device shares one amplifier, and the bias voltage of the readout circuit is not completely consistent, so longitudinal non-uniform noise, namely, stripe noise, will be generated; when the multiple amplifiers are arranged in the vertical direction, a row of outputs of the device shares one amplifier, and the bias voltage of the readout circuit is not completely consistent, so transverse non-uniform noise, namely, stripe noise, will be generated; wherein, the direction of the generated stripe noise is perpendicular to the arrangement direction of the amplifiers in the infrared focal plane imaging device, and the stripe noise in the infrared image usually runs through the entire row or column of the image.

[0033] It should be further explained that the stripe noise interferes with the detection and extraction of useful information in the infrared image. Therefore, it is necessary to remove the stripe noise in the infrared image to improve the reliability of subsequent processing and analysis results.

[0034] S2. Decompose the infrared image to obtain downsampled sub-images at different resolutions, perform Hough line detection on all foreground pixels in the binary image of the downsampled sub-image, and obtain multiple straight lines in the target direction.

[0035] It should be noted that the present invention utilizes the multi-resolution characteristics of wavelet transform to process images at multiple scales and directions, making details and edge information clearer, thereby effectively identifying noise and signals of different frequency components.

[0036] Specifically, the infrared image is decomposed by wavelet transform to obtain down-sampled sub-images at different resolutions and detail sub-images in various directions, and the pixels in the down-sampled sub-images and detail sub-images in various directions have grayscale values; each level of decomposition is mainly achieved by low-pass filtering and high-pass filtering the infrared image in both horizontal and vertical directions; in the present embodiment, five levels of decomposition are performed, and the decomposition is performed in the order of primary decomposition, secondary decomposition, tertiary decomposition, quaternary decomposition and fifth level decomposition. In other embodiments, the decomposition level can be set according to the actual application scenario and requirements; wavelet transform is a well-known technology and will not be described in detail here.

[0037] Exemplarily, the infrared image is first-level decomposed to obtain a down-sampling sub-image LL1, a detail sub-image LH1 in the vertical direction, a detail sub-image HL1 in the horizontal direction, and a detail sub-image HH1 in the diagonal direction, as shown in FIG. Figure 2 As shown; the down-sampled sub-image LL1 is decomposed at two levels to obtain the down-sampled sub-image LL2, the detail sub-image LH2 in the vertical direction, the detail sub-image HL2 in the horizontal direction, and the detail sub-image HH2 in the diagonal direction, as shown in FIG. Figure 2 As shown; each level of decomposition decomposes the downsampled sub-image obtained by the previous level of decomposition, and as the decomposition level increases, the resolution of the obtained downsampled sub-image and the detail sub-image in each direction becomes lower.

[0038] Among them, the down-sampling sub-image contains the low-frequency components of the infrared image in the horizontal and vertical directions, so the down-sampling sub-image is the result of downsampling the infrared image; the detail sub-image in the vertical direction contains the low-frequency components of the infrared image in the horizontal direction and the high-frequency components in the vertical direction, so the detail sub-image in the vertical direction contains the detail features of the infrared image in the vertical direction; the detail sub-image in the horizontal direction contains the high-frequency components of the infrared image in the horizontal direction and the low-frequency components in the vertical direction, so the detail sub-image in the horizontal direction contains the detail features of the infrared image in the horizontal direction; the detail sub-image in the diagonal direction contains the high-frequency components of the infrared image in the horizontal direction and the high-frequency components in the vertical direction, so the detail sub-image in the diagonal direction contains the detail features of the infrared image in the diagonal direction.

[0039] The sub-images finally obtained include: vertical detail sub-image LH1, horizontal detail sub-image HL1, diagonal detail sub-image HH1, vertical detail sub-image LH2, horizontal detail sub-image HL2, diagonal detail sub-image HH2, vertical detail sub-image LH3, horizontal detail sub-image HL3, diagonal detail sub-image HH3, vertical detail sub-image LH4, horizontal detail sub-image HL4, diagonal detail sub-image HH4, vertical detail sub-image LH5, horizontal detail sub-image HL5, diagonal detail sub-image HH5 and down-sampled sub-image LL5.

[0040] Furthermore, edge detection is performed on the downsampled sub-image to obtain an edge image of the downsampled sub-image, wherein the pixel points in the edge image have a gradient amplitude; the algorithms used in edge detection include but are not limited to the Canny algorithm, the Sobel algorithm and the Laplace algorithm, which are all well-known technologies and will not be described in detail here.

[0041] Further, the edge image of the downsampled sub-image is threshold segmented to obtain a binary image; the binary image includes foreground pixels and background pixels, the foreground pixels refer to pixels whose gradient amplitude is greater than or equal to the gradient amplitude threshold, and the background pixels refer to pixels whose gradient amplitude is less than the gradient amplitude threshold, and the gradient amplitude threshold is obtained by Otsu's method, including: obtaining a histogram composed of the gradient amplitudes of all pixels in the edge image, the abscissa of the histogram is the gradient amplitude, and the ordinate is the frequency, the histogram is divided into two groups at a certain threshold, by traversing all possible thresholds, the variance between the two groups under each threshold is calculated, and the threshold that makes the variance maximum is selected as the gradient amplitude threshold; Otsu's method is a well-known technology and will not be described in detail here.

[0042] Further, Hough line detection is performed on all foreground pixels in the binary image to obtain multiple straight lines, and the straight lines whose angles are the same as the target direction are retained. The target direction refers to the vertical direction of the arrangement direction of the amplifiers in the infrared focal plane imaging device. Since the direction of the stripe noise is perpendicular to the arrangement direction of the amplifiers in the infrared focal plane imaging device, the target direction is the direction of the stripe noise. Hough line detection is a well-known technology and will not be described in detail here.

[0043] S3. Calculate the probability that each straight line belongs to stripe noise.

[0044] It should be noted that the present invention can accurately identify the straight line features related to stripe noise through Hough line detection, and can effectively distinguish noise from valid information by combining the evaluation result of the probability that the straight line belongs to stripe noise.

[0045] Specifically, the probability that a straight line belongs to fringe noise is Satisfies the expression:

[0046] ;

[0047] In the formula, is the number of all foreground pixels on the line, is the length of the edge of the infrared image along the target direction, is the difference in the gradient amplitude of all foreground pixels on the straight line, For the straight line The angle between the gradient direction of the foreground pixel and the target direction, is a sine function.

[0048] In one embodiment, the degree of difference in the gradient amplitudes of all foreground pixels on a straight line refers to the standard deviation of the gradient amplitudes of all foreground pixels on the straight line; in another embodiment, the degree of difference in the gradient amplitudes of all foreground pixels on a straight line refers to the information entropy of the gradient amplitudes of all foreground pixels on the straight line.

[0049] It should be noted that stripe noise usually has obvious directionality, and in infrared images, the direction of stripe noise is perpendicular to the arrangement direction of the amplifiers in the infrared focal plane imaging device. Therefore, the target direction is the direction of stripe noise. When the gradient direction of all foreground pixels on a straight line is closer to parallel with the target direction, that is, the angle between the gradient direction of all foreground pixels on a straight line and the target direction is closer to 0° or 180°, the pixels on the straight line are more likely to be stripe noise. At this time, the sine function The smaller the The smaller it is, the probability that the corresponding straight line belongs to stripe noise. The larger the difference, the greater the difference in the gradient amplitude of all foreground pixels on the straight line. The smaller it is, the more consistent the gradient amplitudes of all foreground pixels on the line are. The gradients of these foreground pixels are likely to be caused by stripe noise. The pixels on the line are more likely to be stripe noise. At this time, the probability that the line belongs to stripe noise is The larger the value, the greater the ratio. Since the stripe noise in the infrared image usually runs through the entire row or column of the image, the ratio of the number of all foreground pixels on the straight line to the length of the edge of the infrared image along the target direction is The larger the value is, the greater the degree to which the straight line penetrates the entire row or column of the image, and the more likely the pixels on the straight line are stripe noise. At this time, the probability that the straight line is stripe noise is The bigger.

[0050] Furthermore, the straight line with a probability greater than a preset threshold is recorded as the target straight line; wherein the specific value of the preset threshold can be set according to the actual application scenario and requirements, and the value method of the preset threshold is [0.5, 0.8]. The present invention sets the preset threshold to 0.6.

[0051] S4. Filter the downsampled sub-image and the detail sub-images in each direction according to the neighborhood range of each foreground pixel.

[0052] It should be noted that for the foreground pixel points on the target line, the neighborhood range of the foreground pixel points is obtained according to the distance between adjacent target lines and the probability that the target line belongs to stripe noise, and the downsampled sub-image and the detail sub-images in each direction are filtered according to the neighborhood range, so that the information of the surrounding pixels can be comprehensively considered during denoising, thereby better retaining useful information and removing noise.

[0053] Specifically, the target straight line is composed of foreground pixel points in the binary image; according to the distance between adjacent target straight lines and the probability that the target straight line belongs to stripe noise, the neighborhood range of each foreground pixel point on the target straight line is obtained, and the neighborhood range of the foreground pixel point is composed of the foreground pixel points adjacent to the left side of the foreground pixel point. pixels and the adjacent pixels on the right side of the foreground pixel pixels, where , are the first number and the second number respectively. It should be noted that the neighborhood range of the foreground pixel does not include the foreground pixel itself. Therefore, the neighborhood range of the foreground pixel has a total of Pixels.

[0054] Among them, the first number and the second quantity Satisfies the expression:

[0055] ;

[0056] ;

[0057] In the formula, is the target straight line where the pixel point is located, For the target straight line Left and in a straight line from the target The nearest target straight line, For the target straight line Right and in a straight line from the target The nearest target straight line, is the distance function, , are the first coefficient and the second coefficient respectively, To round up, The target straight line and target straight line The distance The target straight line and target straight line distance.

[0058] Among them, the first coefficient and the second coefficient Satisfies the expression:

[0059] ;

[0060] ;

[0061] In the formula, The target straight line The probability of being stripe noise, The target straight line The probability of being stripe noise.

[0062] It should be noted that the target straight line The probability of being stripe noise The larger the target line The greater the impact on nearby pixels, the more likely it is that the target straight line contained in the neighborhood will be affected. and target straight line There should be fewer pixels between them, so the first coefficient The smaller the first number The smaller the target line The probability of being stripe noise The larger the target line The greater the impact on nearby pixels, the more likely it is that the target straight line contained in the neighborhood will be affected. and target straight line There should be fewer pixels between them, so the second coefficient The smaller the corresponding second quantity The smaller.

[0063] It should be further explained that the present invention obtains the neighborhood range of each foreground pixel point on the target straight line according to the probability that the target straight line belongs to stripe noise, so that the pixel points included in the neighborhood range are less affected by the noise, thereby improving the effect of noise removal.

[0064] Furthermore, the foreground pixel points on the target straight line are recorded as target pixels; for pixel A in the downsampled sub-image with the same coordinates as the target pixel point, the pixel points adjacent to the left of pixel A are recorded as pixels and the pixels adjacent to the right of pixel A Pixels constitute the neighborhood range of pixel A; for pixel B with the same coordinates as the target pixel in the detail sub-images in each direction, the neighboring pixels on the left side of pixel B are pixels and the pixels adjacent to the right of pixel B Pixels constitute the neighborhood range of pixel B.

[0065] Furthermore, according to the neighborhood range of each pixel point, the downsampled sub-image and the detail sub-images in each direction are filtered. The algorithm used for filtering is the median average filtering method, specifically: for the neighborhood range of pixel point C, The grayscale values ​​of pixels, remove the largest grayscale value and the smallest grayscale value, and then calculate the remaining The arithmetic mean of the gray values ​​is taken as the filtering result of pixel C.

[0066] Among them, the median average filtering method is equivalent to the combination of the median filtering method and the arithmetic mean filtering method. Therefore, the median average filtering method combines the advantages of these two filtering methods, which can not only suppress random interference but also filter out obvious pulse interference; the median average filtering method is a well-known technology and will not be described here.

[0067] S5. Perform wavelet reconstruction on the filtered down-sampled sub-image and the filtered detail sub-images in each direction to obtain a denoised infrared image.

[0068] It should be noted that the present invention effectively merges the filtered sub-images into a final infrared image through wavelet reconstruction, retains important features, and reduces the impact of noise.

[0069] Specifically, wavelet reconstruction is performed on the filtered down-sampled sub-image and the filtered detail sub-images in each direction to obtain a denoised infrared image; the wavelet reconstruction is the inverse process of the wavelet transform, and the wavelet reconstruction process is performed in the order of five-level reconstruction, four-level reconstruction, three-level reconstruction, two-level reconstruction and one-level reconstruction, specifically:

[0070] 1. According to step S2 to step S4, obtain the denoised down-sampled sub-image LL5, the denoised detail sub-image LH5 in the vertical direction, the denoised detail sub-image HL5 in the horizontal direction, and the denoised detail sub-image HH5 in the angular line direction; perform five-level reconstruction on the denoised down-sampled sub-image LL5, the denoised detail sub-image LH5 in the vertical direction, the denoised detail sub-image HL5 in the horizontal direction, and the denoised detail sub-image HH5 in the angular line direction to obtain the down-sampled sub-image LL4.

[0071] 2. According to step S2 to step S4, the denoised down-sampled sub-image LL4, the denoised detail sub-image LH4 in the vertical direction, the denoised detail sub-image HL4 in the horizontal direction, and the denoised detail sub-image HH4 in the angular line direction are obtained; and four-level reconstruction is performed on the denoised down-sampled sub-image LL4, the denoised detail sub-image LH4 in the vertical direction, the denoised detail sub-image HL4 in the horizontal direction, and the denoised detail sub-image HH4 in the angular line direction to obtain the down-sampled sub-image LL3.

[0072] 3. According to step S2 to step S4, the denoised down-sampled sub-image LL3, the denoised vertical detail sub-image LH3, the denoised horizontal detail sub-image HL3 and the denoised angular line detail sub-image HH3 are obtained; and three-level reconstruction is performed on the denoised down-sampled sub-image LL3, the denoised vertical detail sub-image LH3, the denoised horizontal detail sub-image HL3 and the denoised angular line detail sub-image HH3 to obtain the down-sampled sub-image LL2.

[0073] 4. According to step S2 to step S4, the denoised down-sampled sub-image LL2, the denoised vertical detail sub-image LH2, the denoised horizontal detail sub-image HL2 and the denoised angular line detail sub-image HH2 are obtained; and secondary reconstruction is performed on the denoised down-sampled sub-image LL2, the denoised vertical detail sub-image LH2, the denoised horizontal detail sub-image HL2 and the denoised angular line detail sub-image HH2 to obtain the down-sampled sub-image LL1.

[0074] 5. According to step S2 to step S4, the denoised down-sampling sub-image LL1, the denoised detail sub-image LH1 in the vertical direction, the denoised detail sub-image HL1 in the horizontal direction, and the denoised detail sub-image HH1 in the angular line direction are obtained; and a first-level reconstruction is performed based on the denoised down-sampling sub-image LL1, the denoised detail sub-image LH1 in the vertical direction, the denoised detail sub-image HL1 in the horizontal direction, and the denoised detail sub-image HH1 in the angular line direction to obtain a denoised infrared image.

[0075] For example, Figure 2 Shown is the process of secondary reconstruction and primary reconstruction.

[0076] It should be noted that the present invention filters the downsampled sub-images and detail sub-images in various directions according to the neighborhood range, comprehensively considering the information of surrounding pixels, so as to better retain useful information and remove noise.

[0077] S6. Calculate the heat transfer efficiency of the hot air duct according to the denoised infrared image.

[0078] It should be noted that the present invention effectively removes stripe noise, making infrared images clearer and improving image readability, thereby improving the reliability of subsequent processing and analysis results, and can more accurately analyze the heat transfer efficiency of hot air ducts, thereby improving accuracy and reliability in industrial applications.

[0079] Specifically, through step S2 to step S5, a denoised infrared image of the air inlet and an infrared image of the air outlet of the hot air duct are obtained; the denoised infrared image is input into the FLIR Tools software to generate a temperature value file of all pixel points of the denoised infrared image, which is used to measure the surface temperature of the hot air duct.

[0080] Further, the heat transfer efficiency of the hot air duct is calculated, and the heat transfer efficiency of the hot air duct satisfies the expression:

[0081] ;

[0082] In the formula, is the heat transfer efficiency of the hot air duct, is the mean temperature value of all pixels in the denoised infrared image of the air outlet, It is the mean temperature value of all pixels in the denoised infrared image of the air inlet.

[0083] In the description of this specification, "plurality" or "several" means at least two, such as two, three or more, etc., unless otherwise clearly and specifically defined.

[0084] Although this specification has shown and described a number of embodiments of the present invention, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Those skilled in the art will conceive of many modifications, changes and alternatives without departing from the ideas and spirit of the present invention. It should be understood that in the practice of the present invention, alternatives to the embodiments of the present invention described herein may be employed.

Claims

1. A method for detecting heat transfer efficiency of hot air ducts based on infrared images, characterized in that: include: The infrared image of the hot air duct is collected by an infrared focal plane imaging device; the infrared image is decomposed by wavelet transform to obtain down-sampled sub-images at different resolutions and detail sub-images in various directions; Perform threshold segmentation on the edge image of the downsampled sub-image to obtain a binary image; Performing Hough line detection on all foreground pixels in the binary image to obtain multiple straight lines in a target direction, where the target direction refers to a direction perpendicular to the arrangement direction of the amplifiers in the infrared focal plane imaging device; Calculate the probability that a line belongs to fringe noise , , is the number of all foreground pixels on the line, is the length of the edge of the infrared image along the target direction, is the difference in the gradient amplitude of all foreground pixels on the straight line, For the straight line The angle between the gradient direction of a foreground pixel and the target direction; The straight line with a probability greater than a preset threshold is recorded as the target straight line; According to the distance between adjacent target lines and the probability that the target line belongs to stripe noise, the neighborhood range of each foreground pixel point on the target line is obtained. The neighborhood range of the foreground pixel point has pixels, the neighborhood range of the foreground pixel is composed of the foreground pixel on the left side. pixels and the adjacent pixels on the right side of the foreground pixel pixels, where , are the first and second quantities respectively, , , is the target straight line where the foreground pixel is located, For the target straight line Left and in a straight line from the target The nearest target straight line, For the target straight line Right and in a straight line from the target The nearest target straight line, is the distance function, To round up, , are the first coefficient and the second coefficient respectively, , , The target straight line The probability of being stripe noise, The target straight line The probability of being stripe noise; according to the neighborhood range of each foreground pixel point, the downsampled sub-image and the detail sub-image in each direction are filtered, including: recording the foreground pixel point on the target straight line as the target pixel point, for the pixel point A in the downsampled sub-image with the same coordinates as the target pixel point, the pixel point A adjacent to the left side is filtered. pixels and the pixels adjacent to the right of pixel A Pixels constitute the neighborhood range of pixel A. For pixel B with the same coordinates as the target pixel in the detail sub-images in each direction, the neighboring pixels on the left side of pixel B are pixels and the pixels adjacent to the right of pixel B The neighborhood range of pixel B is composed of pixels; according to the neighborhood range of each pixel, the down-sampled sub-image and the detail sub-image in each direction are filtered; the denoised infrared image is obtained by wavelet reconstruction of the filtered down-sampled sub-image and the filtered detail sub-image in each direction; the heat transfer efficiency of the hot air duct is calculated according to the denoised infrared image.

2. The method for detecting heat transfer efficiency of hot air ducts based on infrared images according to claim 1 is characterized in that: The binary image includes foreground pixels and background pixels. The foreground pixels refer to pixels whose gradient amplitude is greater than or equal to a gradient amplitude threshold, and the background pixels refer to pixels whose gradient amplitude is less than a gradient amplitude threshold. The gradient amplitude threshold is obtained by the Otsu method.

3. The method for detecting heat transfer efficiency of hot air ducts based on infrared images according to claim 1 is characterized in that: The algorithm used for filtering is the median average filtering method, which includes: for the grayscale values ​​of all pixels within the neighborhood of the pixel point, removing the largest grayscale value and the smallest grayscale value, and then calculating the arithmetic mean of all remaining grayscale values ​​as the filtering result of the pixel point.

4. The method for detecting heat transfer efficiency of hot air ducts based on infrared images according to claim 1 is characterized in that: The infrared image of the hot air duct includes an infrared image of the air inlet and an infrared image of the air outlet of the hot air duct, and the denoised infrared image includes a denoised infrared image of the air inlet and an infrared image of the air outlet of the hot air duct.

5. The method for detecting heat transfer efficiency of hot air ducts based on infrared images according to claim 1 is characterized in that: The step of calculating the heat transfer efficiency of the hot air duct according to the de-noised infrared image includes: Input the denoised infrared image into FLIR Tools software to generate a temperature value file for all pixels of the denoised infrared image; The heat transfer efficiency of the hot air duct satisfies the expression: ； In the formula, is the heat transfer efficiency of the hot air duct, is the mean temperature value of all pixels in the denoised infrared image of the air outlet, It is the mean temperature value of all pixels in the denoised infrared image of the air inlet.

Images