An Image Processing Method for a Capillary Radiation System

Abstract

The present invention relates to the technical field of image processing. More specifically, the present invention relates to an image processing method for a capillary radiation system. The method includes: obtaining a grayscale image of a thermal imaging image of the capillary radiation system, constructing feature points for each pixel, using the split hierarchical clustering algorithm to iteratively split all the feature points as initial clusters. During each split, if the preference degree of the split point is greater than a preset value, then split according to the split point; if it is less than or equal to the preset value, then screen out feature points whose preference degree meets the preset conditions from the cluster where the split point is located as new split points for splitting until a preset termination condition is met, obtaining a split result, and performing anomaly detection on the corresponding region according to the anomaly index of each cluster in the split result. The present invention can improve the efficiency and accuracy of anomaly detection for the area where the capillary radiation system is installed.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing. More specifically, the present invention relates to an image processing method for a capillary radiation system. Background Art

[0002] With the rise of green buildings and smart homes, capillary radiation systems have been widely used as efficient and comfortable heating systems. However, when the capillary radiation system is operating normally, it uses a water supply pipe to continuously transport flowing hot water to circulate throughout the system, which is usually laid inside the wall. Once problems such as local blockage or structural damage occur, it will cause uneven water temperature distribution and thus generate a temperature difference. This temperature difference not only affects the heating effect of the system but also reduces the user's comfort experience. Therefore, quickly and accurately detecting abnormal areas in the area where the capillary radiation system is installed is of crucial significance for ensuring the normal operation of the system, improving the user experience, and timely carrying out repair and maintenance.

[0003] In the related art, for example, the patent application document with the publication number CN114049336A discloses a GIS bushing temperature anomaly detection method, device, equipment, and readable storage medium. The method includes: obtaining an infrared image marked with the GIS bushing area, obtaining the temperatures corresponding to all pixel points in the GIS bushing area to form a temperature set; constructing a temperature spatial distribution model for the GIS bushing area, removing invalid data, and smoothing the single-point noise in the infrared image; constructing a temperature frequency distribution model for the GIS bushing area, setting a double baseline, dividing the interference area and the normal area, and extracting the main information; respectively marking the pixel points after removing the noise, marking the pixel points with temperatures exceeding the preset threshold as suspected abnormal points, and marking the pixel points not exceeding the preset threshold as the highest temperature points and the lowest temperature points.

[0004] However, when marking suspected abnormal points in the above solution, it only considers that the higher the temperature, the higher the probability of an anomaly in the GIS bushing area. However, when the pipeline ruptures, the temperature in the corresponding area may be lower than the surrounding temperature, and all pixel points with temperatures lower than the low temperature baseline will be excluded in the above solution, resulting in missed detections and affecting the accuracy of the detection results. Summary of the Invention

[0005] In order to solve the problem of being unable to accurately detect abnormal temperature areas, the present invention provides an image processing method for a capillary radiation system. The method includes:

[0006] Obtaining a grayscale image of the thermal imaging image of the capillary radiation system, and constructing a feature point for each pixel with the temperature value of each pixel in the grayscale image as the abscissa and the pixel position of the corresponding pixel as the ordinate;

[0007] Use all the feature points as the initial clusters and perform iterative splitting using the splitting hierarchical clustering algorithm. During each split, if the preference degree of the split point is greater than the preset value, split according to this split point; if it is less than or equal to the preset value, select the feature points whose preference degrees meet the preset conditions from the cluster where the split point is located as the new split point for splitting until the preset termination condition is met to obtain the split result;

[0008] Perform anomaly detection on the corresponding regions according to the anomaly indices of each cluster in the split result. The anomaly index is the normalized value of the range of the gradient direction angles of the feature points in the corresponding cluster;

[0009] The method for obtaining the preference degree is as follows: Obtain the degree of change of the temperature value of any feature point. The degree of change characterizes the difference between the temperature value of this feature point and the mode of the temperature values of all feature points, and calculate the credibility of the temperature value of this feature point. The credibility is negatively correlated with the standard deviation of the gradient direction angles of all feature points in the eight-neighborhood, and the sum of the temperature value differences between this feature point and each feature point in the eight-neighborhood;

[0010] Take the ratio of the degree of change to the credibility as the split index of this feature point, and calculate the ratio of the split index of this feature point to the maximum value of the split indices of all feature points to obtain the preference degree of this feature point.

[0011] When the present invention uses the splitting hierarchical clustering algorithm to split the initial clusters composed of all feature points, it can avoid over-segmentation of the initial clusters, improve the segmentation efficiency, and thus can achieve efficient detection of abnormal temperature regions where a capillary radiation system is installed; and when splitting the initial clusters composed of all feature points, it can determine the split point by integrating various data, ensuring the accuracy of the split result, so that while improving the detection efficiency, the accuracy of the detection result can be guaranteed.

[0012] Preferably, the degree of change satisfies the following relational expression:

[0013] ;

[0014] In the formula, is the degree of change of the temperature value of the th feature point; is the temperature value of the th feature point; is the mode of the temperature values of all feature points; is the standard deviation of the temperature values of all feature points; is the absolute value symbol.

[0015] The present invention performs normalization through the standard deviation, which can eliminate the influence of different unit dimensions.

[0016] Preferably, the confidence level of the temperature value of any feature point satisfies the following relational expression:

[0017] ;

[0018] In the formula, is the confidence level of the temperature value of the th feature point; is the gradient direction angle of the th feature point within the eight-neighborhood of the th feature point; is the average value of the gradient direction angles of all feature points within this eight-neighborhood; is the number of feature points within this eight-neighborhood; is the temperature value of the th feature point; is the temperature value of the th feature point within this eight-neighborhood; is the natural exponential function.

[0019] The present invention utilizes the feature that the gradient directions of the temperature values in the normal temperature change region are relatively consistent, while the gradient directions of the temperature values in the abnormal temperature change region are relatively dispersed, and can accurately evaluate the possibility that the temperature value of each feature point is a normal value, so as to eliminate the influence of the normal temperature change region on the detection of the abnormal temperature region.

[0020] Preferably, in any splitting, if the preference degree of the splitting point determined by the splitting hierarchical clustering algorithm is less than or equal to a preset value, then it is judged whether the preference degree of the feature point with the largest splitting index in the cluster where the splitting point is located is greater than the preset value;

[0021] If it is greater than the preset value, it is determined that the preference degree of this feature point meets the preset conditions, and this feature point is used as a new splitting point for splitting; if it is less than or equal to the preset value, it is determined that the splitting process meets the preset termination conditions, and the splitting process ends to obtain the splitting result.

[0022] The way of determining the splitting point in the present invention can avoid the over-segmentation phenomenon of the initial cluster composed of all feature points, improve the splitting efficiency, and thus can improve the efficiency of abnormal detection.

[0023] Preferably, in any splitting, the maximum value of the Euclidean distances between any two feature points in the cluster where the splitting point is located is greater than the maximum value of the Euclidean distances between any two feature points in the remaining clusters.

[0024] Preferably, the range of the gradient direction angles of the feature points in any cluster is the difference between the maximum value and the minimum value of the gradient direction angles of the feature points in this cluster.

[0025] Preferably, the abnormal index of any cluster in the splitting result satisfies the following relational expression:

[0026] ;

[0027] In the formula, is the anomaly index of the th cluster in the splitting result; is the maximum value of the gradient direction angles of the feature points in this cluster; is the minimum value of the gradient direction angles of the feature points in this cluster; is the normalization function.

[0028] Preferably, the method for obtaining the gradient direction angle of any feature point includes:

[0029] Using the Sobel operator to obtain the gradient direction angle of each feature point.

[0030] Preferably, the anomaly detection of the corresponding area according to the anomaly index of each cluster in the splitting result includes:

[0031] Obtaining a preset warning threshold. If the anomaly index of any cluster in the splitting result is greater than the anomaly index threshold, it is determined that there is an anomaly in the capillary network in the area corresponding to this cluster, and an alarm is issued;

[0032] If it is less than or equal to the warning threshold, it is determined that the capillary network in the area corresponding to this cluster is normal.

[0033] The present invention can efficiently and accurately locate the area where the capillary network is abnormal.

[0034] Preferably, during each splitting, if the preference degrees of the splitting point and the feature points in the cluster where the splitting point is located are both less than or equal to the preset value, it is determined that the splitting process meets the termination condition, and the splitting process is stopped.

[0035] The present invention has the following effects:

[0036] 1. When the present invention performs splitting operations on the initial clusters composed of all feature points using the splitting hierarchical clustering algorithm, it can ensure that the preference degree of the splitting point determined by each splitting operation is the largest and greater than the preset value. Thus, while being able to divide the feature points with similar temperature values into the same cluster, it avoids over-segmentation of the clusters composed of feature points in the normal temperature change area, improves the splitting efficiency, and thereby can improve the detection efficiency of the abnormal temperature area in the installed capillary radiation system.

[0037] 2. The present invention determines the preference degrees of each feature point by integrating various aspects of data, ensuring the accuracy of the preference degrees, so that the feature points in the abnormal temperature change region have relatively large preference degrees, and thus the feature points in the abnormal temperature change region can be segmented, ensuring the accuracy of the splitting result, and thereby improving the accuracy of abnormal detection for the region where the capillary radiation system is installed. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] By reading the following detailed description with reference to the accompanying drawings, the above and other objects, features, and advantages of the exemplary embodiments of the present invention will become readily understandable. In the drawings, several embodiments of the present invention are shown by way of example and not limitation, and the same or corresponding reference numerals represent the same or corresponding parts, wherein:

[0039] Figure 1 is a schematic flowchart of the steps of an image processing method for a capillary radiation system according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0040] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some but not all of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative efforts fall within the scope of protection of the present invention.

[0041] The following will describe the specific embodiments of the present invention in detail with reference to the accompanying drawings.

[0042] Referring to Figure 1 , an image processing method for a capillary radiation system includes steps S1 - S3, specifically as follows:

[0043] S1: Obtain a grayscale image of the thermal imaging image of the capillary radiation system, and construct a feature point for each pixel with the temperature value of each pixel in the grayscale image as the abscissa and the pixel position of the corresponding pixel as the ordinate.

[0044] Specifically, a thermal imaging image of the region where the capillary radiation system is installed can be obtained by using a thermal imager, and the thermal imaging image can be converted into a grayscale image to simplify the image data for subsequent processing and analysis. It should be noted that the process of converting the thermal imaging image into a grayscale image is a prior art, and this embodiment will not elaborate on it here.

[0045] Furthermore, the temperature value of each pixel in the grayscale image can be used as the abscissa, and the position information of the corresponding pixel in the grayscale image can be used as the ordinate, so as to construct a feature point for each pixel in the grayscale image to provide a data basis for subsequent analysis.

[0046] It should be noted that the present invention constructs feature points based on the temperature values and pixel positions of each pixel in the grayscale image, can utilize the temperature information of all pixels to identify abnormal temperature regions, and accurately locate the identified abnormal temperature regions through the pixel positions, so as to effectively detect abnormalities in the area where the capillary radiation system is installed.

[0047] S2: Use the splitting hierarchical clustering algorithm to iteratively split all feature points as initial clusters. In each split, if the preference degree of the split point is greater than the preset value, split according to this split point; if it is less than or equal to the preset value, select the feature points whose preference degrees meet the preset conditions from the cluster where this split point is located as the new split point for splitting until the preset termination condition is met, and obtain the split result.

[0048] It should be noted that since the temperature characteristics of the area where the capillary radiation system is installed are relatively obvious, when there are abnormalities in the capillary network, such as blockage or rupture, it usually leads to uneven temperature distribution in the local area, making the abnormal area significantly different from other normal areas in temperature distribution. And the splitting hierarchical clustering algorithm can effectively identify the area with uneven temperature distribution by gradually splitting the clusters, so the abnormal detection of the capillary network can be realized by using the splitting hierarchical clustering algorithm.

[0049] It should be further noted that the process of selecting the split point in the traditional splitting hierarchical clustering algorithm is: select the object with the largest average dissimilarity from the cluster with the largest diameter as the split point. Among them, the diameter of the cluster is the maximum value of the Euclidean distance between any two objects in the corresponding cluster; the average dissimilarity is the average distance between any object in the corresponding cluster and the remaining objects.

[0050] However, when the capillary network is working normally, there may be normal temperature changes. For example, the opening and closing operations of openings such as windows and doors will reduce the temperature of the corresponding area, so that the temperature of the corresponding area is significantly different from that of other areas. And the traditional splitting hierarchical clustering algorithm may split the area with normal temperature changes into multiple clusters, resulting in over-segmentation and reducing the detection efficiency. Therefore, the present invention improves this algorithm. The specific improvement content is: calculate the preference degree of the feature points based on the temperature characteristics of the feature points, as well as the difference characteristics between normal temperature changes and abnormal temperature changes, and use the preference degree to judge the split point determined by the traditional splitting hierarchical clustering algorithm. If the preference degree of this split point meets the conditions, split according to this split point; otherwise, re-select the feature points whose preference degrees meet the conditions from the cluster where this split point is located as the new split point to perform splitting based on the new split point.

[0051] Specifically, the determination of the preference degree of any feature point can be realized through the following steps:

[0052] Step 1: Obtain the degree of change in the temperature value of any feature point, where the degree of change characterizes the difference between the temperature value of this feature point and the mode of the temperature values of all feature points.

[0053] It should be noted that when the capillary radiation system operates normally, hot water is transported through the water supply pipe to the entire capillary network, and the temperature distribution is relatively uniform. However, when problems such as blockage or rupture occur in the capillary network, it will cause obvious heat accumulation or heat dispersion in local areas, thereby making the water temperature distribution uneven. Therefore, the mode of the temperature value is used as a measurement index in the present invention, so as to be able to analyze the change of the temperature value of each feature point compared with the normal temperature value.

[0054] In an exemplary embodiment of the present invention, the difference between the temperature value of any feature point and the mode of the temperature values of all feature points is a comprehensive difference, and the determination of the degree of change of any feature point can be achieved through the following steps. Specifically, the degree of change of any feature point satisfies the following relational expression:

[0055] ;

[0056] In the formula, is the degree of change in the temperature value of the th feature point; is the temperature value of the th feature point; is the mode of the temperature values of all feature points; is the standard deviation of the temperature values of all feature points, which is used to eliminate the influence of unit dimensions; is the absolute value symbol.

[0057] Among them, reflects the deviation of the temperature value of the th feature point from the normal temperature value. The larger this value is, the greater the possibility that the capillary network at the position corresponding to the th feature point is blocked or ruptured, resulting in a significant deviation of the temperature at this position from the normal temperature level, and the degree of change of the corresponding feature point is relatively large.

[0058] Step 2: Calculate the credibility of the temperature value of this feature point. The credibility is negatively correlated with the standard deviation of the gradient direction angles of all feature points in the eight-neighborhood, and the sum of the temperature value differences between this feature point and each feature point in the eight-neighborhood.

[0059] It should be noted that when the capillary network is working properly, the temperature changes caused by the opening and closing operations of openings such as windows and doors usually have a relatively consistent direction; while when problems such as blockage and rupture occur in the network, the direction of temperature change is usually divergent from the high-temperature area to the low-temperature area, showing a radial distribution centered on the high-temperature area. Therefore, the present invention uses this feature to calculate the credibility of the temperature value of the feature point, which can accurately measure the possibility that the temperature value of each feature point is a normal value, thereby effectively eliminating the influence of the feature points corresponding to the normal temperature change area on the abnormal detection result.

[0060] Specifically, the credibility of the temperature value of any feature point satisfies the following relational expression:

[0061] ;

[0062] In the formula, is the credibility of the temperature value of the th feature point; is the gradient direction angle of the th feature point in the eight-neighborhood of the th feature point; is the average value of the gradient direction angles of all feature points in this eight-neighborhood; is the number of feature points in this eight-neighborhood; is the temperature value of the th feature point; is the temperature value of the th feature point in this eight-neighborhood; is the natural exponential function.

[0063] Among them, reflects the standard deviation of the gradient direction angles of all feature points in the eight-neighborhood of the th feature point. The smaller this value is, the more consistent the gradient direction of the temperature values corresponding to the feature points in the eight-neighborhood of this feature point is, and further it indicates that this feature point is more likely to be in the normal area, and then the credibility of the temperature value of this feature point is relatively large.

[0064] reflects the sum of the temperature value differences between the th feature point and each feature point in the eight-neighborhood. The smaller this value is, the more consistent the temperature values in the neighborhood range of this feature point are, and further it indicates that this feature point is more likely to be in the normal area, and then the credibility of the temperature value of this feature point is relatively large.

[0065] Optionally, the absolute value of the difference between the temperature value of any feature point and the temperature values of each feature point in the eight-neighborhood can also be used as the temperature value difference between this feature point and the corresponding feature points in the eight-neighborhood. In this embodiment, the determination method of the difference is not particularly limited.

[0066] In another embodiment, the credibility of the temperature value of any feature point can also be calculated through other calculation formulas, such as: ; in the formula, is the credibility of the temperature value of the th feature point; is the standard deviation of the gradient direction angles of all feature points within the eight-neighborhood of the th feature point; is the temperature value of the th feature point; is the temperature value of the th feature point within this eight-neighborhood; is the number of feature points within this eight-neighborhood; is the absolute value symbol.

[0067] In an exemplary embodiment of the present invention, the determination of the gradient direction angle of any feature point can be achieved through the following steps:

[0068] Use the Sobel operator to obtain the gradient direction angle of each feature point.

[0069] It should be noted that the process of obtaining the gradient direction angle using the Sobel operator is a prior art, and this embodiment will not elaborate on it here.

[0070] Optionally, other operators, such as the Prewitt operator, can also be used to calculate the gradient direction angle of each feature point.

[0071] Step 3: Take the ratio of the degree of change to the credibility as the splitting index of this feature point, and calculate the ratio of the splitting index of this feature point to the maximum value of the splitting indices of all feature points to obtain the preference degree of this feature point.

[0072] Among them, the splitting index refers to the possibility of any feature point being a splitting point. It should be noted that when the credibility of the temperature value of any feature point is small and the degree of change of the temperature value of this feature point is large, it indicates that there is a relatively high possibility that the temperature value of this feature point is abnormal, and correspondingly, the possibility of this feature point being a splitting point is relatively large, that is, the splitting index of this feature point is relatively large. Therefore, the present invention takes the ratio of the degree of change of the temperature value of any feature point to the credibility of the temperature value of this feature point as the splitting index of this feature point, so as to accurately measure the possibility of each feature point being a splitting point.

[0073] It should be further noted that when the splitting index of any feature point is closer to the maximum value of the splitting indices of all feature points, then this feature point is more likely to be a splitting point, and the corresponding preference degree of this feature point is greater. Therefore, in the present invention, the ratio of the splitting index of any feature point to the maximum value of the splitting indices of all feature points is used as the preference degree of this feature point, so that the splitting point can be determined based on the preference degree each time a splitting operation is performed, avoiding over-segmentation phenomena, improving the segmentation efficiency, and thus improving the efficiency of anomaly detection.

[0074] Next, the process of iteratively splitting the initial cluster composed of all feature points using the improved splitting hierarchical clustering algorithm will be described in detail:

[0075] Step 1: In the first splitting process, use the traditional splitting hierarchical clustering algorithm to determine the splitting point in this splitting process.

[0076] In an exemplary embodiment of the present invention, in any splitting, the maximum value of the Euclidean distances between any two feature points in the cluster where the splitting point is located is greater than the maximum value of the Euclidean distances between any two feature points in the remaining clusters. That is, the diameter of the cluster where the splitting point is located is the largest, which further indicates that the probability that the cluster where the splitting point is located needs to be split is relatively large compared to the remaining clusters.

[0077] It should be noted that the process of using the traditional splitting hierarchical clustering algorithm to determine the splitting point is a prior art, and this embodiment will not elaborate on it here.

[0078] Step 2: Determine whether the preference degree of this splitting point is greater than a preset value. If it is greater than the preset value, perform a splitting operation based on this splitting point to obtain two clusters; if it is less than or equal to the preset value, select a feature point whose preference degree meets the preset conditions from the initial cluster as a new splitting point for splitting operation to obtain two clusters; and repeat the splitting process until the preset termination condition is met to obtain the splitting result. In this embodiment, the preset value is 0.6. Of course, an appropriate preset value can also be selected according to specific situations, and the present embodiment does not make a special limitation on the value of the preset value.

[0079] In an exemplary embodiment of the present invention, in any splitting, if the preference degree of the splitting point determined by the splitting hierarchical clustering algorithm is less than or equal to the preset value, then determine whether the preference degree of the feature point with the largest splitting index in the cluster where the splitting point is located is greater than the preset value; if it is greater than the preset value, determine that the preference degree of this feature point meets the preset conditions, and use this feature point as a new splitting point for splitting; if it is less than or equal to the preset value, determine that the splitting process meets the preset termination condition, and end the splitting process to obtain the splitting result.

[0080] It should be noted that if there is no feature point with a preference greater than the preset value in the cluster where the splitting point is located, it indicates that the temperature values of all feature points in the cluster where the splitting point is located are likely to be the temperature values of the normal region, and there is no need to perform the splitting operation again, avoiding over-segmentation, improving the splitting efficiency, and thus improving the efficiency of anomaly detection for the regions corresponding to each cluster in the splitting result.

[0081] S3: Perform anomaly detection on the corresponding regions according to the anomaly indices of each cluster in the splitting result, where the anomaly index is the normalized value of the range of the gradient direction angles of the feature points in the corresponding cluster.

[0082] It should be noted that when the pipe network is blocked, the water flow slows down, heat accumulates to form a local high-temperature area, and the gradient direction of the temperature diverges from the high-temperature area to the surrounding low-temperature areas, forming a radial distribution; while when the pipe network ruptures and water leaks, heat is dissipated to form a local low-temperature area, and the gradient direction of the temperature converges from the surrounding high-temperature areas to the low-temperature area, forming a convergent distribution; making the gradient direction of the temperature in the abnormal region usually show the characteristics of dispersion or convergence, while the gradient direction of the temperature in the normal region is often more consistent. Therefore, the present invention uses this feature to calculate the anomaly index of each cluster in the splitting result, which can accurately evaluate the anomaly situation of the regions corresponding to each cluster, and thus can accurately locate the abnormal region.

[0083] Specifically, the anomaly index of any cluster in the splitting result satisfies the following relational expression:

[0084] ;

[0085] In the formula, is the anomaly index of the th cluster in the splitting result; is the maximum value of the gradient direction angles of the feature points in this cluster; is the minimum value of the gradient direction angles of the feature points in this cluster; is the normalization function.

[0086] Among them, reflects the range of the gradient direction angles of the feature points in the th cluster in the splitting result. The larger this value is, the more dispersed the gradient direction angles of the temperature values of the feature points in this cluster are, and further indicates that the possibility of the region corresponding to this cluster being an abnormal region is relatively large, and the corresponding anomaly index of this cluster is large.

[0087] In another embodiment, it is also possible to limit the value range of the hyperbolic tangent function to 0-1, so as to normalize the range of the gradient direction angles of the feature points in each cluster using the hyperbolic tangent function with a value range of 0-1.

[0088] In an exemplary embodiment of the present invention, the detection of an abnormal area can be achieved through the following steps:

[0089] Obtain a preset warning threshold. If the abnormal index of any class cluster in the splitting result is greater than the abnormal index threshold, it is determined that there is an abnormality in the capillary network in the area corresponding to this class cluster, and an alarm is issued; if it is less than or equal to the warning threshold, it is determined that the capillary network in the area corresponding to this class cluster is normal.

[0090] Optionally, the warning threshold can be set to 0.8. When the abnormal index of any class cluster in the splitting result is less than or equal to 0.8, it is determined that the capillary network at the positions corresponding to all feature points in this class cluster is normal; when the abnormal index of any class cluster in the splitting result exceeds 0.8, it is determined that there are defects in the capillary network at the positions corresponding to all feature points in this class cluster, such as blockage or rupture, and an alarm is triggered. Then, the abnormal area can be located through the positions of the feature points in this class cluster for subsequent maintenance work, realizing the abnormal detection of the area where the capillary radiation system is installed. There is no specific limitation on the size of the warning threshold in this embodiment.

[0091] In the description of this specification, the meanings of "a plurality of" and "several" are at least two, such as two, three or more, etc., unless otherwise specifically and clearly defined.

[0092] Although this specification has shown and described multiple embodiments of the present invention, it is obvious to those skilled in the art that such embodiments are provided only by way of example. Those skilled in the art will think of many changes, alterations and alternative ways without departing from the spirit and idea of the present invention. It should be understood that various alternative solutions to the embodiments of the present invention described herein can be adopted in the process of practicing the present invention.

Claims

1. An image processing method for a capillary radiation system, characterized in that Including: Obtain the grayscale image of the thermal imaging image of the capillary radiation system, and use the temperature value of each pixel in the grayscale image as the abscissa and the pixel position of the corresponding pixel as the ordinate to construct the feature points of each pixel; Use the splitting hierarchical clustering algorithm to iteratively split all the feature points as the initial clusters. In each split, if the preference degree of the split point is greater than the preset value, split according to the split point; if it is less than or equal to the preset value, screen out the feature points whose preference degrees meet the preset conditions from the cluster where the split point is located as the new split point for splitting until the preset termination condition is met to obtain the split result; Perform anomaly detection on the corresponding area according to the anomaly index of each cluster in the split result, where the anomaly index is the normalized value of the range of the gradient direction angles of the feature points in the corresponding cluster; The method for obtaining the preference degree is as follows: obtain the degree of change of the temperature value of any feature point, where the degree of change characterizes the difference between the temperature value of the feature point and the mode of the temperature values of all feature points, calculate the credibility of the temperature value of the feature point, and the credibility is negatively correlated with the standard deviation of the gradient direction angles of all feature points in the eight-neighborhood and the sum of the temperature value differences between the feature point and each feature point in the eight-neighborhood; Take the ratio of the degree of change to the credibility as the splitting index of the feature point, and calculate the ratio of the splitting index of the feature point to the maximum value of the splitting indices of all feature points to obtain the preference degree of the feature point.

2. The image processing method of a capillary radiation system according to claim 1, wherein The degree of change satisfies the following relational expression: ； Wherein, is the degree of change of the temperature value of the th feature point; is the temperature value of the th feature point; is the mode of the temperature values of all feature points; is the standard deviation of the temperature values of all feature points; is the absolute value symbol.

3. A method for processing capillary radiation system image, as described in claim 1, wherein The credibility of the temperature value of any feature point satisfies the following relational expression: ； Wherein, is the credibility of the temperature value of the th feature point; is the gradient direction angle of the th feature point in the eight-neighborhood of the th feature point; is the mean value of the gradient direction angles of all feature points in the eight-neighborhood; is the number of feature points in the eight-neighborhood; is the temperature value of the th feature point; is the temperature value of the th feature point in the eight-neighborhood; is the natural exponential function.

4. A method for image processing of a capillary radiation system according to claim 1, characterized in that, In any split, if the preference degree of the split point determined by the splitting hierarchical clustering algorithm is less than or equal to the preset value, determine whether the preference degree of the feature point with the largest splitting index in the cluster where the split point is located is greater than the preset value; If it is greater than the preset value, determine that the preference degree of the feature point meets the preset conditions and use the feature point as the new split point for splitting; if it is less than or equal to the preset value, determine that the splitting process meets the preset termination condition and end the splitting process to obtain the split result.

5. A method for image processing of a capillary radiation system according to claim 4, characterized in that In any split, the maximum value of the Euclidean distances between any two feature points in the cluster where the split point is located is greater than the maximum value of the Euclidean distances between any two feature points in the remaining clusters.

6. A method for image processing of a capillary radiation system according to claim 1, characterized in that, The range of the gradient direction angles of the feature points in any cluster is the difference between the maximum value and the minimum value of the gradient direction angles of the feature points in the cluster.

7. A method for image processing of a capillary radiation system according to claim 1, characterized in that The anomaly index of any cluster in the split result satisfies the following relational expression: ； In the formula, is the anomaly index of the -th cluster in the splitting result; is the maximum value of the gradient direction angles of the feature points in this cluster; is the minimum value of the gradient direction angles of the feature points in this cluster; is the normalization function.

8. A method for image processing of a capillary radiation system according to claim 7, characterized in that, The method for obtaining the gradient direction angle of any feature point includes: Use the Sobel operator to obtain the gradient direction angle of each feature point.

9. A method for image processing of a capillary radiation system according to claim 1, characterized in that, The performing anomaly detection on the corresponding area according to the anomaly index of each cluster in the split result includes: Obtain a preset warning threshold. If the anomaly index of any cluster in the split result is greater than the warning threshold, determine that there is an anomaly in the capillary network in the area corresponding to the cluster and issue an alarm; If it is less than or equal to the warning threshold, determine that the capillary network in the area corresponding to the cluster is normal.

10. A method for image processing of a capillary radiation system according to claim 1, characterized in that, In each split, if the preference degrees of the split point and the feature points in the cluster where the split point is located are both less than or equal to the preset value, determine that the splitting process meets the termination condition and stop the splitting process.

Images