Abnormal body temperature detection method based on infrared thermal imaging

Abstract

The present invention relates to the field of image data processing technology, and more specifically, to a method for detecting abnormal body temperature based on infrared thermal imaging. The method comprises: first, using the CLAHE algorithm to enhance the initial image according to different block sizes; then, calculating the directional consistency of each pixel in the enhanced image at each block size, and quantifying the noise confidence weight of each pixel using the information entropy of the pixels within the neighborhood; then, calculating the average of the product of the noise confidence weight and the directional consistency of all pixels in the same enhanced image to obtain the feasibility of each block size; finally, based on the weight coefficient converted from the feasibility, performing weighted fusion on the grayscale values ​​at the same pixel position in the enhanced images at all block sizes to detect abnormal temperature based on the generated target enhanced image. The present invention can optimize the image enhancement process and achieve accurate detection of abnormal surface temperature of livestock and poultry.

Description

Technical Field

[0001] The present invention relates to the technical field of image data processing, and more particularly to a method for detecting abnormal body temperature based on infrared thermal imaging. Background Art

[0002] In large-scale livestock and poultry farming, body temperature is a key indicator of animal health. Traditional contact temperature measurement methods can easily trigger stress responses in animals and make real-time monitoring of flocks difficult. Infrared thermal imaging technology, with its non-contact and high efficiency, provides an important means for detecting abnormal body temperatures in livestock and poultry. By analyzing the temperature distribution characteristics in infrared images, suspected diseased individuals can be quickly located, enabling early warning and precise disease prevention and control, which is of great significance for safeguarding the economic benefits of the livestock industry and animal welfare.

[0003] However, while infrared thermal imaging technology enables non-contact, wide-area temperature monitoring for livestock and poultry, raw infrared images often suffer from low contrast, blurred details, and noise interference due to environmental noise, insufficient device resolution, and the complexity of surface thermal radiation. This makes subtle temperature anomalies (such as localized inflammation or early-stage infection) difficult to discern. Therefore, image enhancement of the captured raw infrared images is a key preprocessing step to improve detection accuracy.

[0004] Currently, infrared image enhancement mostly uses the Contrast-Limited Adaptive Histogram Equalization (CLAHE) algorithm, and usually uses a fixed block size to enhance the image. However, when the block size is too large, the algorithm cannot effectively enhance the contrast of local small temperature anomaly areas in the image, resulting in an inability to fully enhance the image. When the block size is too small, although local details can be improved, artifact noise is easily generated due to over-correction of the local histogram, which interferes with the true temperature distribution. It is impossible to take into account the needs of maintaining the global temperature distribution characteristics and enhancing local details, resulting in poor image enhancement effect and affecting subsequent analysis results. Summary of the Invention

[0005] In order to solve the problem that when using the conventional CLAHE algorithm to enhance infrared images, it is impossible to take into account the requirements of maintaining global temperature distribution characteristics and enhancing local details, resulting in poor image enhancement effect, which in turn affects the accuracy of subsequent abnormal temperature detection, the present invention provides a body temperature abnormality detection method based on infrared thermal imaging. The method includes:

[0006] Obtain the temperature distribution map of the livestock and poultry body surface and perform preprocessing to obtain the initial image;

[0007] Based on different block sizes, the CLAHE algorithm is used to enhance the initial image. The directional consistency of each pixel in the enhanced image at each block size is calculated. The directional consistency represents the consistency of the gradient direction of the pixel with the gradient direction of the corresponding pixel in different enhanced images.

[0008] The information entropy of all pixels in the neighborhood of the pixel is used to quantify the noise confidence weight of each pixel in the enhanced image at each block size. The average product of the noise confidence weight and the directional consistency of all pixels in the same enhanced image is calculated to obtain the feasibility of each block size.

[0009] The feasibility is converted into a weight coefficient whose sum is 1, and the grayscale value of the same pixel position in the enhanced image at all block sizes is weightedly summed according to the weight coefficient to obtain the target enhanced image, so as to detect abnormal surface temperature of livestock and poultry based on the target enhanced image.

[0010] The present invention first uses the CLAHE algorithm to perform image enhancement on the initial image based on different block sizes, and then calculates the directional consistency of each pixel in the enhanced image at each block size, so as to utilize the feature that the local structure of the pixel points in the enhanced image at the appropriate block size has a high consistency to evaluate the suitability of each block size. Then, the noise confidence weight is quantified to reduce the influence of noise on the evaluation of the suitability of each block size, so that the feasibility of each block size can be accurately measured through the calculated feasibility. Then, by converting the feasibility into a weight coefficient, the grayscale values ​​of the same pixel position in the enhanced image at all block sizes are fused, so as to avoid insufficient enhancement or artifact problems caused by improper block size selection, optimize the image enhancement process, and provide an accurate image basis for the subsequent detection of abnormal surface temperature of livestock and poultry.

[0011] Preferably, calculating the directional consistency of each pixel in the enhanced image at each block size includes:

[0012] For the enhanced images at each block size, the directional consistency of any pixel is calculated based on the difference between the gradient direction angle of any pixel in any enhanced image and the gradient direction angle of the corresponding pixel in all remaining enhanced images. The directional consistency is negatively correlated with the difference.

[0013] The present invention utilizes the local structure of pixels in the enhanced image at a suitable block size, especially the feature that the gradient direction has a high consistency, to calculate the directional consistency of each pixel point, and can evaluate the suitability of image enhancement for the initial image at each block size.

[0014] Preferably, the difference is a comprehensive difference, and the comprehensive difference satisfies the following relationship:

[0015] ;

[0016] Where, For the The enhanced image at the block size The gradient direction angle of the pixel point is The enhanced image at the block size The comprehensive difference between the gradient direction angles of the pixels; 、 Respectively The block size and The size of each block; is the size of the maximum block size; is the ordinal number of the maximum block size; 、 Respectively The block size and The enhanced image at the block size The gradient direction angle of each pixel; is the absolute value symbol; A function that returns the minimum value; is the natural exponential function.

[0017] When calculating the difference in gradient direction angles of corresponding pixels in enhanced images at different block sizes, the present invention takes into account the influence of different block sizes, and can consider image enhancement results at different block sizes differently.

[0018] Preferably, the direction consistency of any pixel point is calculated to satisfy the following relationship:

[0019] ;

[0020] Where, For the The enhanced image at the block size The direction consistency of each pixel; For the The enhanced image at the block size The gradient direction angle of the pixel point is The enhanced image at the block size The comprehensive difference between the gradient direction angles of the pixels; is the ordinal number of the maximum block size; is the natural exponential function.

[0021] Preferably, the method for obtaining the gradient direction angle includes:

[0022] The Sobel operator is used to obtain the horizontal gradient value and vertical gradient value of any pixel point, and the inverse tangent value of the ratio of the vertical gradient value to the horizontal gradient value is calculated to obtain the gradient direction angle of any pixel point.

[0023] Preferably, the size of the maximum block size is a maximum positive integer smaller than half of the width of the initial image, and the difference between the sizes of ordinal adjacent block sizes is a preset integer multiple of 2.

[0024] The present invention ensures that the initial image can be divided into at least two or more image blocks by limiting the maximum block size.

[0025] Preferably, the noise confidence weight of each pixel in the enhanced image at each block size is quantified by the information entropy of all pixels within the pixel's neighborhood, including:

[0026] For each pixel in the enhanced image at each block size, the information entropy of all pixels in the neighborhood of any pixel is used as the local entropy of any pixel, and the difference between 1 and the local entropy is used as the noise confidence weight of any pixel.

[0027] Preferably, the feasibility of each block size satisfies the following relationship:

[0028] ;

[0029] Where, For the The feasibility of the block size; For the The enhanced image at the block size The local entropy of each pixel; For the The enhanced image at the block size The direction consistency of each pixel; For the The number of pixels in the enhanced image at each block size.

[0030] Preferably, detecting abnormal temperature of the livestock and poultry body surface based on the target enhanced image includes:

[0031] Obtain a preset grayscale threshold. If the grayscale value of any pixel in the target enhanced image is less than the grayscale threshold, the area where the pixel is located is determined to be a normal temperature area.

[0032] If it is greater than or equal to the grayscale threshold, the area where any pixel is located is marked as a temperature abnormality area.

[0033] Preferably, the method for obtaining the initial image includes:

[0034] The temperature distribution map is gray-scaled and filtered in turn, and the initial image is obtained after filtering.

[0035] The present invention can provide a clearer and more accurate initial image basis for subsequent image enhancement through a series of preprocessing operations.

[0036] The present invention has the following effects:

[0037] The present invention can effectively take into account the global temperature distribution characteristics of the image and the need for local detail enhancement through CLAHE image enhancement based on different block sizes, combined with a comprehensive evaluation of directional consistency and noise confidence weight, and realizes adaptive fusion of enhanced images with different block sizes, avoiding insufficient enhancement or artifact problems caused by a single fixed block size, and optimizing the image enhancement process, so that accurate detection of abnormal surface temperature of livestock and poultry can be achieved based on the obtained target enhanced image. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 The figure is a flowchart of the steps of a method for detecting abnormal body temperature based on infrared thermal imaging according to an embodiment of the present invention. DETAILED DESCRIPTION

[0039] The specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0040] Reference Figure 1 The method for detecting abnormal body temperature based on infrared thermal imaging includes steps S1 to S4, which are as follows:

[0041] S1: Obtain the temperature distribution map of the livestock and poultry body surface and perform preprocessing to obtain the initial image.

[0042] It should be noted that the surface temperature distribution of healthy livestock and poultry has specific patterns and ranges, while when livestock and poultry have diseases or physiological abnormalities, their surface temperature often changes, such as local fever or overall abnormal body temperature. Therefore, the temperature distribution of livestock and poultry can be used to monitor their body temperature in real time.

[0043] Specifically, infrared thermal imaging cameras or infrared cameras can be used to capture infrared thermal images of livestock and poultry, generating a temperature distribution map that intuitively reflects the temperature distribution of the livestock and poultry's body surface. It should be noted that since the captured images may be pseudo-color and contain noise, a series of preprocessing operations are required to improve the image quality of the captured temperature distribution map.

[0044] In an exemplary embodiment of the present invention, the determination of the initial image can be achieved by the following steps:

[0045] The temperature distribution map is gray-scaled and filtered in turn, and the initial image is obtained after filtering.

[0046] Optionally, a Gaussian filter, a median filter, a mean filter or other algorithms may be used to filter the collected temperature distribution map of the livestock and poultry body surface. This embodiment does not impose any particular limitation on the selected filtering method.

[0047] It should be noted that the process of grayscale processing and filtering of an image is a conventional technology and will not be described in detail in this embodiment.

[0048] S2: Based on different block sizes, the initial image is enhanced using the CLAHE algorithm. The directional consistency of each pixel in the enhanced image at each block size is calculated. The directional consistency represents the consistency of the gradient direction of the pixel with the gradient direction of the corresponding pixel in different enhanced images.

[0049] It should be noted that when using the conventional CLAHE (Contrast Limited Adaptive Histogram Equalization) algorithm for image enhancement, a fixed-size block size is usually used to obtain an enhanced image. However, when the block size is too large, the image area contained in each block becomes larger, and the local features are diluted, which will result in the inability to fully enhance the image; when the selected block size is too small, artifacts may be introduced. Therefore, in order to fully enhance the image while avoiding too many artifacts as much as possible, the present invention sets a plurality of different block sizes, and then evaluates the similarity of the local structure of the enhanced image at each block size with the enhanced image at other block sizes through the directional consistency of the pixel points, thereby providing a data basis for subsequent calculations.

[0050] In an exemplary embodiment of the present invention, the size of the maximum block size is a maximum positive integer smaller than half of the width of the initial image, and the difference between the sizes of ordinal adjacent block sizes is a preset integer multiple of 2.

[0051] The width of the initial image refers to the number of pixels contained in the shorter side of the initial image.

[0052] Specifically, you can set the block size set , satisfying the relationship: , where For the The size of each block; is the ordinal number of the block size; The ordinal number of the maximum block size.

[0053] For example, when When , the set block sizes include 8×8, 16×16, 32×32 and 64×64, so that the initial image can be enhanced respectively based on the above four block sizes using the CLAHE algorithm to obtain enhanced images under the four block sizes.

[0054] The purpose of limiting the maximum block size is to ensure that the initial image can be divided into at least two blocks, and the purpose of limiting the difference in the sizes of ordinal adjacent blocks to an integer multiple of 2 is to allow the block sizes to cover a range from small to large, thereby adapting to abnormal body temperature areas of different scales in the initial image.

[0055] It should be noted that when using the CLAHE algorithm to enhance the initial image based on different block sizes, the contrast limit factor set at each block size is 2.0. This value is a conventional value selected in the CLAHE algorithm and is not specifically limited in this embodiment. The process of using the CLAHE algorithm for image enhancement when the block size is known is conventional technology and is not described in detail in this embodiment.

[0056] In an exemplary embodiment of the present invention, the directional consistency of each pixel in the enhanced image at each block size can be determined by the following steps:

[0057] For the enhanced images at each block size, the directional consistency of any pixel is calculated based on the difference between the gradient direction angle of any pixel in any enhanced image and the gradient direction angle of the corresponding pixel in all remaining enhanced images. The directional consistency is negatively correlated with the difference.

[0058] It should be noted that if the block size is appropriate, the local structure of the pixels in the enhanced image (especially the gradient direction) at that size is highly consistent with the results of other appropriate sizes. Conversely, an inappropriate block size can lead to insufficient image enhancement or the introduction of artifacts, resulting in significant differences from the CLAHE results obtained with the appropriate block size. Therefore, the present invention leverages this feature to assess the directional consistency of each pixel to determine whether the block size is appropriate.

[0059] In an exemplary embodiment of the present invention, the determination of the gradient direction angle of a pixel point can be achieved by the following steps:

[0060] The Sobel operator is used to obtain the horizontal gradient value and vertical gradient value of any pixel point, and the inverse tangent value of the ratio of the vertical gradient value to the horizontal gradient value is calculated to obtain the gradient direction angle of any pixel point.

[0061] Optionally, the horizontal and vertical gradient values ​​of the pixel point can be obtained using a Prewitt operator, a Roberts operator, or the like. This embodiment does not particularly limit the type of operator selected. It should be noted that the process of obtaining the horizontal and vertical gradient values ​​using the Sobel operator is a prior art technique and is not described in detail in this embodiment.

[0062] In an exemplary embodiment of the present invention, the difference between the gradient direction angles of pixels at the same pixel position in the enhanced images at different block sizes is a comprehensive difference, and the comprehensive difference satisfies the following relationship:

[0063] ;

[0064] Where, For the The enhanced image at the block size The gradient direction angle of the pixel point is The enhanced image at the block size The comprehensive difference between the gradient direction angles of the pixels; 、 Respectively The block size and The size of each block; is the size of the maximum block size; is the ordinal number of the maximum block size; 、 Respectively The block size and The enhanced image at the block size The gradient direction angle of each pixel; is the absolute value symbol; A function that returns the minimum value; is the natural exponential function; It is 180 degrees.

[0065] in, Reflects the The block size and The size difference weight of the two block sizes is set. When the difference between the two block sizes (the absolute value of the difference between the values) is small, if the two block sizes are both appropriate block sizes, the local features of the pixels in the enhanced images under the two block sizes have high similarity. At this time, setting a larger size difference weight can differently consider the image enhancement results under different block sizes.

[0066] Reflects the enhanced image under the two block sizes The difference in gradient direction angles of pixels. The smaller the value, the The enhanced image at the block size The gradient direction of the pixel point is The enhanced image at the block size The gradient directions of the pixels are relatively consistent, and the corresponding The enhanced image at the block size The direction consistency of each pixel is relatively high.

[0067] In another embodiment, the relationship: ; Evaluate the difference in gradient direction angles of pixels at the same pixel position in enhanced images with different block sizes.

[0068] Furthermore, after determining the difference between the gradient direction angle of any pixel point in the enhanced image at any block size and the gradient direction angle of the corresponding pixel point in the enhanced images at other block sizes, the direction consistency of any pixel point can be calculated based on all the differences obtained.

[0069] Specifically, the directional consistency of any pixel satisfies the following relationship:

[0070] ;

[0071] Where, For the The enhanced image at the block size The direction consistency of each pixel; For the The enhanced image at the block size The gradient direction angle of the pixel point is The enhanced image at the block size The comprehensive difference between the gradient direction angles of the pixels; is the ordinal number of the maximum block size; is the natural exponential function.

[0072] In another embodiment, a Sigmoid function may also be used to characterize the negative correlation between the difference and the direction consistency.

[0073] Furthermore, the directional consistency calculation formula can be used to calculate the directional consistency of all pixels in the enhanced image at each block size, thereby providing a data basis for the subsequent analysis process.

[0074] S3: Quantify the noise confidence weight of each pixel in the enhanced image at each block size through the information entropy of all pixels in the pixel's neighborhood, and calculate the average product of the noise confidence weight and direction consistency of all pixels in the same enhanced image to obtain the feasibility of each block size.

[0075] It should be noted that although the image has been filtered in the preprocessing stage, there may still be high-noise areas in the initial image. Therefore, in order to reduce the interference of noise on the evaluation of image enhancement effect, the present invention further quantifies the noise confidence weight of each pixel point, so as to accurately evaluate whether the size of each block is appropriate, that is, the feasibility of each block size.

[0076] In an exemplary embodiment of the present invention, the noise confidence weight of each pixel may be determined by the following steps:

[0077] For each pixel in the enhanced image at each block size, the information entropy of all pixels in the neighborhood of any pixel is used as the local entropy of any pixel, and the difference between 1 and the local entropy is used as the noise confidence weight of any pixel.

[0078] Among them, information entropy reflects the complexity and uncertainty index of the grayscale value distribution in the range around any pixel point. The larger the value, the more chaotic the grayscale value distribution around any pixel point is, which further indicates that there may be noise around any pixel point, and the corresponding noise confidence weight of any pixel point is relatively low.

[0079] The neighborhood range refers to the local range centered on any pixel point, with a neighborhood size of an empirical value (such as 3×3). Therefore, the noise confidence weights of all pixels in the enhanced image at each block size can be determined based on the information entropy of the grayscale values ​​of all pixels within the neighborhood range of each pixel point.

[0080] Furthermore, for the enhanced image at each block size, the average value of the product of the noise confidence weight and the direction consistency of all pixels in the same enhanced image can be calculated to determine the reliability of each block size.

[0081] Specifically, the feasibility of each block size satisfies the following relationship:

[0082] ;

[0083] Where, For the The feasibility of the block size; For the The enhanced image at the block size The local entropy of each pixel; For the The enhanced image at the block size The direction consistency of each pixel; For the The number of pixels in the enhanced image at each block size; is the summation symbol.

[0084] in, Reflects the The enhanced image at the block size The noise confidence weight of the pixel. When the directional consistency of all pixels with higher noise confidence weights in the enhanced image at a certain block size is also high, it means that it is more appropriate to perform image enhancement according to the block size, and the feasibility of the corresponding block size is relatively high.

[0085] At this point, the feasibility of each block size can be determined.

[0086] S4: Convert the feasibility into a weight coefficient whose sum is 1, and perform weighted summation on the grayscale values ​​of the same pixel position in the enhanced images at all block sizes according to the weight coefficient to obtain a target enhanced image, so as to detect abnormal surface temperature of livestock and poultry based on the target enhanced image.

[0087] It should be noted that using a larger block size for image enhancement can effectively preserve the overall temperature distribution characteristics, while using a smaller block size for image enhancement can better highlight local details. Therefore, to fully utilize the advantages of both, the present invention dynamically assigns weights to the grayscale values ​​of the same pixel position in enhanced images with different block sizes based on the feasibility of the corresponding block size. A weighted fusion method is then used to generate a comprehensive enhanced image, namely the target enhanced image. This target enhanced image can then be used to accurately detect abnormal surface temperatures of livestock and poultry.

[0088] It should be further explained that in order to eliminate the influence of the absolute numerical difference of feasibility, the feasibility of each block size needs to be converted into a weight coefficient first.

[0089] Specifically, the feasibility of each block size is converted into a weight coefficient, satisfying the following relationship:

[0090] ;

[0091] Where, For the The weight coefficient of each block size; For the The feasibility of the block size; is an activation function used to convert the feasibility of all block sizes into weight coefficients whose sum is 1.

[0092] In another embodiment, a normalization function may be used, such as Function that converts the feasibility of all block sizes into weight coefficients that sum to 1.

[0093] Furthermore, after the feasibility of each block size is converted into a weight coefficient, the weighted cumulative sum of the grayscale values ​​at the same pixel position in the enhanced image under all block sizes can be calculated to obtain the fused grayscale value of each pixel position, thereby determining the target enhanced image.

[0094] Specifically, the fused grayscale value at any pixel position satisfies the following relationship:

[0095] ;

[0096] Where, For the The fused gray value of the pixel position; For the The weight coefficient of each block size; For the The enhanced image at the block size Gray value of pixel position; is the ordinal number of the maximum block size; is the summation symbol.

[0097] At this point, the fused grayscale value of each pixel position can be obtained, thereby forming a target enhanced image, which provides an accurate temperature distribution basis for subsequent abnormal detection of livestock and poultry surface temperature.

[0098] In an exemplary embodiment of the present invention, the abnormal surface temperature of livestock and poultry can be identified by the following steps:

[0099] Obtain a preset grayscale threshold. If the grayscale value of any pixel in the target enhanced image is less than the grayscale threshold, the area where the pixel is located is determined to be a normal temperature area.

[0100] If it is greater than or equal to the grayscale threshold, the area where any pixel is located is marked as a temperature abnormality area.

[0101] Optionally, the grayscale threshold can be set to 200 (grayscale value). If the grayscale value of a certain position in the target enhanced image is greater than the grayscale threshold, the position is determined to be beyond the normal temperature range and the position is marked. By analyzing and judging the grayscale values ​​of all pixels in the entire target enhanced image, the abnormal temperature areas on the surface of livestock and poultry can be accurately identified, providing a reliable basis for livestock and poultry health monitoring and disease prevention and control.

[0102] It should be understood that various alternatives to the embodiments of the invention described herein may be employed in practicing the invention.

Claims

1. A method for detecting abnormal body temperature based on infrared thermal imaging, characterized in that: include: Obtain the temperature distribution map of the livestock and poultry body surface and perform preprocessing to obtain the initial image; Based on different block sizes, the CLAHE algorithm is used to enhance the initial image. The directional consistency of each pixel in the enhanced image at each block size is calculated. The directional consistency represents the consistency of the gradient direction of the pixel with the gradient direction of the corresponding pixel in different enhanced images. The information entropy of all pixels in the neighborhood of the pixel is used to quantify the noise confidence weight of each pixel in the enhanced image at each block size. The average product of the noise confidence weight and the directional consistency of all pixels in the same enhanced image is calculated to obtain the feasibility of each block size. The feasibility is converted into a weight coefficient whose sum is 1, and according to the weight coefficient, the grayscale value of the same pixel position in the enhanced image at all block sizes is weightedly summed to obtain a target enhanced image, so as to detect abnormal temperature of the livestock and poultry body surface based on the target enhanced image.

2. The method for detecting abnormal body temperature based on infrared thermal imaging according to claim 1, characterized in that: The calculating of the directional consistency of each pixel in the enhanced image at each block size includes: For the enhanced image at each block size, the directional consistency of any pixel point in any enhanced image is calculated based on the difference between the gradient direction angle of the pixel point in any enhanced image and the gradient direction angle of the corresponding pixel point in all remaining enhanced images. The directional consistency is negatively correlated with the difference.

3. The method for detecting abnormal body temperature based on infrared thermal imaging according to claim 2, characterized in that: The difference is a comprehensive difference, and the comprehensive difference satisfies the following relationship: ； Where, For the The enhanced image at the block size The gradient direction angle of the pixel point is The enhanced image at the block size The comprehensive difference between the gradient direction angles of the pixels; 、 Respectively The block size and The size of each block; is the size of the maximum block size; is the ordinal number of the maximum block size; 、 Respectively The block size and The enhanced image at the block size The gradient direction angle of each pixel; is the absolute value symbol; A function that returns the minimum value; is the natural exponential function.

4. The method for detecting abnormal body temperature based on infrared thermal imaging according to claim 3, characterized in that: The calculation of the direction consistency of any pixel point satisfies the following relationship: ； Where, For the The enhanced image at the block size The direction consistency of each pixel; For the The enhanced image at the block size The gradient direction angle of the pixel point is The enhanced image at the block size The comprehensive difference between the gradient direction angles of the pixels; is the ordinal number of the maximum block size; is the natural exponential function.

5. The method for detecting abnormal body temperature based on infrared thermal imaging according to claim 4, characterized in that: The method for obtaining the gradient direction angle includes: The Sobel operator is used to obtain the horizontal gradient value and the vertical gradient value of any pixel point, and the inverse tangent value of the ratio of the vertical gradient value to the horizontal gradient value is calculated to obtain the gradient direction angle of the any pixel point.

6. The method for detecting abnormal body temperature based on infrared thermal imaging according to claim 4, characterized in that: The size of the maximum block size is a maximum positive integer that is smaller than half of the width of the initial image, and the difference between the sizes of ordinal adjacent block sizes is a preset integer multiple of 2.

7. The method for detecting abnormal body temperature based on infrared thermal imaging according to claim 1, characterized in that: The method of quantifying the noise confidence weight of each pixel in the enhanced image at each block size by using the information entropy of all pixels in the neighborhood of the pixel includes: For each pixel in the enhanced image at each block size, the information entropy of all pixels in the neighborhood of any pixel is used as the local entropy of the pixel, and the difference between 1 and the local entropy is used as the noise confidence weight of the pixel.

8. The method for detecting abnormal body temperature based on infrared thermal imaging according to claim 7, characterized in that: The feasibility of each block size satisfies the following relationship: ； Where, For the The feasibility of the block size; For the The enhanced image at the block size The local entropy of each pixel; For the The enhanced image at the block size The direction consistency of each pixel; For the The number of pixels in the enhanced image at each block size.

9. The method for detecting abnormal body temperature based on infrared thermal imaging according to claim 1, characterized in that: The detecting of abnormal temperature on the surface of livestock and poultry based on the target enhanced image includes: Obtaining a preset grayscale threshold, and if the grayscale value of any pixel in the target enhanced image is less than the grayscale threshold, determining that the area where the pixel is located is a normal temperature area; If it is greater than or equal to the grayscale threshold, the area where any pixel point is located is marked as a temperature abnormality area.

10. The method for detecting abnormal body temperature based on infrared thermal imaging according to claim 1, characterized in that: The method for obtaining the initial image includes: Grayscale processing and filtering processing are performed on the temperature distribution map in sequence, and the initial image is obtained after filtering is completed.

Images