Real-time battery health monitoring system based on infrared thermal imager

Abstract

This invention relates to the field of image processing technology, specifically to a real-time battery health monitoring system based on an infrared thermal imager. The system includes: acquiring continuous thermal imaging images of the battery and obtaining the battery region for each image; acquiring a grayscale scatter plot of any pixel within the battery region; analyzing the characteristics of all data points in the grayscale scatter plot, the temporal differences in grayscale changes between the pixel and its neighboring pixels, and obtaining normal data points in the grayscale scatter plot; based on all normal data points in the grayscale scatter plot, obtaining the temporal grayscale distribution of the pixel; and based on the grayscale distribution of all pixels within the battery region, obtaining the temperature change index of the battery region; and obtaining the thermal change parameters of the battery region through the temperature change index, thereby monitoring battery health. This invention further improves the accuracy of battery health monitoring.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of image data processing, in particular to a real-time battery health monitoring system based on an infrared thermal imager. BACKGROUND

[0002] An infrared thermal imager is a non-contact temperature measuring instrument that converts the infrared radiation emitted by a target object into a visual thermal image to quickly and accurately reflect the temperature distribution of the target object, which has significant advantages in discovering abnormal temperature points and thermal faults. The health status of a battery affects its performance and life, and abnormal conditions such as aging, short circuit, and overheating may occur during long-term use. If not discovered and addressed in a timely manner, it may cause serious safety accidents. Therefore, applying an infrared thermal imager to the field of battery health monitoring can timely discover abnormal conditions and take appropriate measures to extend the service life and safety of the battery.

[0003] In battery health monitoring, an infrared thermal imager is usually used to obtain the temperature distribution inside the battery, and then the average temperature or the temperature of a specific point is selected for heat calculation, and finally the battery health is monitored according to the heat changes at different times. However, when using an infrared thermal imager to collect thermal radiation distribution data of the battery area, due to the system error of the infrared thermal imager or external factor interference, isolated noise points may appear in the battery thermal imaging image, which cannot reflect the real temperature condition of the location. At the same time, when the battery shell falls off, the thermal conductivity of the area changes, and the obtained temperature data may not accurately reflect the real temperature condition of the area, which reduces the accuracy of battery health monitoring. SUMMARY

[0004] The present application provides a real-time battery health monitoring system based on an infrared thermal imager to solve the existing problems.

[0005] The real-time battery health monitoring system based on an infrared thermal imager of the present application adopts the following technical scheme:

[0006] One embodiment of the present application provides a real-time battery health monitoring system based on an infrared thermal imager, which includes the following modules:

[0007] An image acquisition module is used to acquire battery thermal imaging images at consecutive time points and obtain the battery area of each battery thermal imaging image.

[0008] The temperature change detection module is configured to: acquire a gray scatter plot of any pixel point according to the gray value of the pixel point in time sequence; acquire an isolated noise abnormality probability of any data point in the gray scatter plot according to the gray value change of the data point; acquire a battery shell falling abnormality probability of the data point according to the gray value difference between a wave trough data point and adjacent two side data points in the gray scatter plot; acquire a similar trend parameter of each data point in the gray scatter plot of the pixel point according to the gray value change time sequence difference between the pixel point and other pixel points in the neighborhood; acquire normal data points in the gray scatter plot according to the isolated noise abnormality probability, the battery shell falling abnormality probability and the similar trend parameter of the data point in the gray scatter plot; acquire the gray distribution of the pixel point in time sequence according to all the normal data points in the gray scatter plot of the pixel point; and acquire a temperature change index of the battery region according to the gray distribution of all the pixel points in the battery region.

[0009] The battery health monitoring module is configured to acquire a battery region heat change parameter according to the temperature change index of the battery region, and monitor the battery health according to the battery region heat change parameter.

[0010] Preferably, the method for acquiring the gray scatter plot of any pixel point according to the gray value of the pixel point in time sequence comprises the following specific steps:

[0011] The gray scatter plot of the pixel point at any coordinate in the battery region is constructed according to the gray value of the pixel point in time sequence, wherein the horizontal coordinate of the gray scatter plot represents the time interval of the battery thermal imaging image in each second, and the vertical coordinate represents the gray value of the pixel point in each frame.

[0012] Preferably, the method for acquiring the isolated noise abnormality probability of any data point in the gray scatter plot according to the gray value change of the data point comprises the following specific steps:

[0013] Any pixel point in the battery region is denoted as pixel point Q, any data point in the gray scatter plot of the pixel point Q is denoted as data point O, the angle between the connecting line of the data point O and other data points in the gray scatter plot and the x axis of the gray scatter plot is denoted as the inclination angle of the data point O and other data points, and the average value of the inclination angles of the data point O and adjacent A data points is denoted as the adjacent inclination angle average value of the data point O, wherein A is a preset adjacent data point parameter.

[0014] The specific calculation method of the isolated noise abnormality probability of the data point O in the gray scatter plot is as follows:

[0015]

[0016] In the formula, P' represents the isolated noise abnormality probability of the data point O, θ represents the adjacent inclination angle average value of the data point O, and P represents the isolated noise abnormality probability of the data point O. O O,m ​represents the inclination angle of data point O and the mth data point on both sides; μ O represents the average of adjacent inclination angles of data point O; ΔT O,min represents the minimum value of the gray value difference between data point O and other data points in the gray scatter plot; || represents the absolute value symbol.

[0017] Preferably, the method for obtaining the battery shell shedding anomaly probability of the data point according to the gray value difference between the data point in the valley of the gray scatter plot and the adjacent two data points on both sides comprises the following specific method:

[0018] Obtain the valley screening parameter in the gray scatter plot of the pixel point Q;

[0019] The specific calculation method of the battery shell shedding anomaly scatter probability of data point O in the gray scatter plot is as follows:

[0020]

[0021] i∈{1,2…m1 O},j∈{1,2…m2 O}

[0022] In the formula, P” O represents the battery shell shedding anomaly scatter probability of data point O; T O represents the gray value of data point O; T O,i represents the gray value of the ith data point on the left side of data point O; T O,j represents the gray value of the jth data point on the right side of data point O; F O represents the valley screening parameter of data point O; MIN() represents the minimum value of all values; || represents the absolute value symbol; m1 O represents the number of all data points on the left side of data point O; m2 O represents the number of all data points on the right side of data point O; α is a preset first parameter.

[0023] Preferably, the specific method for obtaining the valley screening parameter is as follows:

[0024] Pre-set the valley screening parameter;

[0025] For data point O, obtain the absolute value of the gray value difference between data point O and all data points on the left and right sides, and record it as the left adjacent gray difference and the right adjacent gray difference of data point O. When the left adjacent gray difference and the right adjacent gray difference on the left and right sides of data point O are both greater than 0, the valley screening parameter is equal to 1; otherwise, the valley screening parameter is equal to 0.

[0026] Preferably, the specific method for obtaining the similarity trend parameter of each data point in the gray scatter plot of the pixel point according to the gray value change time difference between the pixel point and other pixel points in the neighborhood comprises:

[0027] For the data point O on the gray scatter plot of the pixel point Q, taking the gray values of the data point O and the N data points on the adjacent time sequence as the gray data set of the data point O, N is a preset second parameter; obtaining the Pearson correlation coefficient of the gray data set of the data point O and the gray data set of all data points in the gray scatter plot of any pixel point, selecting the maximum value of the Pearson correlation coefficient; taking the data points corresponding to the two gray data sets in the maximum value of the Pearson correlation coefficient, and taking the other data point except the data point O as the similar data point of the data point O in the gray scatter plot of any pixel point; taking the time sequence difference value of the data point O and the corresponding similar data point in the respective gray scatter plot as the similar time sequence difference value;

[0028] According to the gray data set of the data point O in the gray scatter plot of the pixel point Q and the gray data set of all data points in the gray scatter plot of the jth pixel point in the eight neighborhood of the pixel point Q, the similar time sequence difference value of the pixel point Q and the similar data point in the gray scatter plot of the jth pixel point in the eight neighborhood is obtained.

[0029]

[0030] In the formula, S O represents the similar trend parameter of the data point O in the gray scatter plot of the pixel point Q; t o represents the time sequence of the data point O in the gray scatter plot of the pixel point Q; t l,o represents the time sequence of the similar data point of the data point O in the gray scatter plot of the lth pixel point; B represents the number of pixel points in the eight neighborhood of the pixel point Q; exp() represents the exponential function with natural constant as the base; || represents the absolute value symbol.

[0031] Preferably, the method for obtaining the normal data point in the gray scatter plot according to the isolated noise anomaly probability, the battery shell falling anomaly probability and the similar trend parameter of the data point in the gray scatter plot comprises the following specific method:

[0032] According to the isolated noise anomaly probability, the battery shell falling anomaly probability and the similar trend parameter of the data point in the gray scatter plot, the normal degree of all data points in the gray scatter plot is obtained.

[0033] When the normal degree of the data point is greater than a preset threshold θ, the data point is determined as a normal data point.

[0034] Preferably, the specific method for obtaining the normal degree of the data point is as follows:

[0035] The isolated noise anomaly probability, the battery shell falling anomaly probability and the similar trend parameter of any data point in the gray scatter plot are obtained.

[0036] The specific calculation method of the normal degree of the data point O in the gray scatter plot is as follows:

[0037]

[0038] In the formula, P O represents the normal degree of the data point O; P ' O represents the isolated noise anomaly probability of the data point O; P" O represents the battery shell shedding anomaly probability of the data point O; S O represents the similar trend parameter of the data point O; and a is a preset first parameter; sigmoid() represents a sigmoid normalization function.

[0039] Preferably, the gray scale distribution of the pixel point in time sequence is obtained according to all normal data points in the gray scatter plot of the pixel point, the temperature change index of the battery region is obtained according to the gray scale distribution of all pixel points in the battery region, and the specific method comprises the following steps:

[0040] All normal data points in the gray scatter plot are fitted by using a polynomial regression to obtain a gray scale change curve of the pixel point in time sequence, and the average curvature of all data points on the gray scale change curve is calculated; the average curvature of the gray scale change curve of all pixel points in the battery region is obtained.

[0041] The specific calculation method of the temperature change index of the battery region under the infrared thermal imager is as follows:

[0042]

[0043] In the formula, E represents the temperature change index of the battery region; n represents the number of pixel points of the battery region; D k represents the Euclidean distance of the kth pixel point in the battery region to the center of the battery region; p k represents the average curvature of the gray scale change curve of the kth pixel point in the battery region in time sequence, which represents the gray scale distribution of the i th pixel point.

[0044] Preferably, the specific method for obtaining the heat change parameter of the battery region through the temperature change index of the battery region and monitoring the battery health according to the heat change parameter of the battery region comprises the following steps:

[0045] The heat change parameter of the battery region is obtained by using a heat conduction formula according to the temperature change index of the battery.

[0046] A preset heat threshold is used to judge the health status of the battery, and when the heat change parameter of the battery region is greater than the heat threshold, the health of the battery is in trouble, and there may be a situation of overheating or imminent explosion, and a warning is given.

[0047] The beneficial effects of the technical scheme of the present application are: obtaining a gray scatter diagram of any pixel point in the battery area under time sequence, obtaining the probability of any data point in the gray scatter diagram being an abnormal scatter point according to the characteristics of isolated noise points and abnormal scatters caused by the falling of the battery shell, determining the normal degree of the data point in the pixel point gray scatter diagram according to the temperature change trend of the battery area pixel point and the consistency in its neighborhood, and screening all normal data points in the gray scatter diagram. In this way, the gray change of each pixel point is obtained, and the temperature change of the entire battery area is represented by the gray change of all pixel points. Finally, the battery heat change is obtained according to the battery temperature change, and the battery health is monitored. The present application further improves the accuracy of battery health monitoring. BRIEF DESCRIPTION OF DRAWINGS

[0048] In order to more clearly illustrate the technical scheme in the embodiments of the present application or the prior art, the drawings needed to be used in the embodiments or the prior art description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0049] Figure 1 The structure block diagram of the real-time battery health monitoring system based on the infrared thermal imager of the present application. DETAILED DESCRIPTION

[0050] In order to further illustrate the technical means and effects adopted by the present application to achieve the predetermined invention purpose, the specific implementation, structure, features and effects of the real-time battery health monitoring system based on the infrared thermal imager according to the present application are described in detail as follows. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. In addition, the specific features, structures or characteristics in one or more embodiments can be combined in any suitable form.

[0051] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the present application belongs.

[0052] The specific scheme of the real-time battery health monitoring system based on the infrared thermal imager provided by the present application is described in detail below with reference to the drawings.

[0053] Please refer to Figure 1 which shows the structure block diagram of the real-time battery health monitoring system based on the infrared thermal imager provided by one embodiment of the present application. The system includes the following modules:

[0054] The image acquisition module 101 acquires battery thermal imaging images and segments the battery thermal imaging images to obtain target images.

[0055] It should be noted that when batteries experience aging, overheating, or other problems, they release a large amount of heat, which appears as a bright color on the thermal image. As the heat spreads, the background area will also appear in a similar color on the thermal image, which has a significant impact on monitoring battery temperature changes. Therefore, the battery thermal image is segmented to obtain the battery region, and then the heat change parameters in the battery region are obtained. The battery health status is then monitored based on the heat change parameters.

[0056] To implement the real-time battery health monitoring system based on an infrared thermal imager proposed in this embodiment, it is first necessary to acquire a thermal image of the battery. The specific process is as follows:

[0057] Step (1.1): Collect battery thermal imaging images at continuous times using an infrared thermal imager and obtain the battery area of ​​each battery thermal imaging image.

[0058] It should be noted that when a problem occurs in the battery health status, a rapid thermal change will occur within one second. Therefore, 12 consecutive frames of battery thermal imaging images are acquired every second for battery health monitoring.

[0059] Step (1.2) uses the Otsu method to perform threshold segmentation on the battery thermal imaging image to obtain a binary image of the battery region. Then, opening and closing operations are used to perform morphological processing on the binary image of the battery region to obtain the battery region. This battery region is used as a battery region matching template. The SIFT algorithm is used to match the battery thermal imaging image with the battery region matching template for all subsequent battery thermal imaging images. The battery region in all acquired battery thermal imaging images is determined based on the matching result.

[0060] It should be noted that, due to the thermal effect of current, the temperature change of the battery is usually quite significant after the battery has been powered on for 3 minutes. Therefore, in this embodiment, an infrared thermal imager is used to collect the thermal image of the battery at this moment to obtain the battery region matching template.

[0061] It should be further noted that Otsu's method and SIFT algorithm are well-known techniques in image processing, so they will not be described in detail in this embodiment.

[0062] Thus, the battery region in the battery thermal imaging image is obtained using the above method.

[0063] The temperature change detection module 102 acquires a grayscale scatter plot of all pixels within the battery area, analyzes the characteristics of all data points in the grayscale scatter plot of the pixels, the temporal differences in grayscale changes between the pixels and other pixels in the neighborhood, and acquires normal data points in the grayscale scatter plot. Based on all normal data points in the grayscale scatter plot of the pixels, it acquires the grayscale distribution of the pixels in time. Based on the grayscale distribution of all pixels within the battery area, it acquires the temperature change index of the battery area.

[0064] It should be noted that due to systematic errors in infrared thermal imagers or interference from external factors, isolated noise points may appear in the battery area. Additionally, when the battery casing detaches, the thermal conductivity of that area changes, causing the data acquired at a certain moment to not accurately reflect the true temperature of that area. These situations will result in some abnormal points in the grayscale scatter plot of some pixels in the battery area, which significantly impacts battery health monitoring. Therefore, it is necessary to remove abnormal points from the grayscale scatter plot of individual pixels, then obtain the grayscale change curve of each pixel over time, and based on the grayscale change curve, obtain the grayscale change of each pixel. Finally, obtain the grayscale change of the entire battery area to obtain the temperature change index of the battery area under the infrared thermal imager. In the grayscale scatter plot of each pixel in the battery area, when a pixel is an isolated noise point, its grayscale value initially appears as a normal value in the thermal image, then the grayscale value undergoes a sudden change, appearing isolated in the grayscale scatter plot, and finally the pixel's grayscale value returns to its normal value. Based on the above features, obtain the probability that each point in the grayscale scatter plot belongs to an isolated noise point.

[0065] Specifically, in step 2.1, firstly, for any pixel in the battery region at any coordinate, construct a grayscale scatter plot of its grayscale value across consecutive frames. The horizontal axis of the grayscale scatter plot represents the time interval of the battery thermal imaging image in consecutive frames per second, and the vertical axis represents the grayscale value of the pixel in each frame. Let any pixel in the battery region be denoted as pixel Q, and any data point in the grayscale scatter plot of pixel Q be denoted as data point O. The angle between the line connecting data point O to other data points and the x-axis of the grayscale scatter plot is denoted as the angle between data point O and the x-axis of the grayscale scatter plot. The tilt angles of other data points; the mean of the tilt angles of data point O and the A adjacent data points on both sides is recorded as the mean of adjacent tilt angles, where A is a preset parameter for adjacent data points; then, the tilt angles of data point O and the A adjacent data points on both sides, the mean of adjacent tilt angles of data point O, and the minimum value of the gray level difference between data point O and other data points in the gray level scatter plot are obtained, and the isolated noise anomaly probability of data points is constructed using the above data; the isolated noise anomaly probability characterizes the probability that data point O is an anomalous scatter plot caused by isolated noise points.

[0066] It should be noted that, according to the construction method of grayscale scatter plot, since the grayscale scatter plot is formed by arranging the grayscale values ​​of all corresponding pixels at the same coordinates in the battery region of the continuous frame battery thermal imaging image in chronological order, each data point in the grayscale scatter plot has a corresponding pixel in the battery region at the corresponding time sequence.

[0067] It should be noted that in this embodiment, the adjacent data point parameter A is preset to 4 based on experience, and can be adjusted according to the actual situation. This embodiment does not impose a specific limitation. In addition, there may be an out-of-boundary problem in this embodiment. That is, when data point O is a data point at one end of the grayscale scatter plot, it is impossible to obtain the leftmost and rightmost data points. Therefore, in order to avoid the above-mentioned out-of-boundary problem, this embodiment selects the data points A adjacent to the other side for calculation for this type of data (for example, if data point O is the leftmost data point in the grayscale scatter plot, then select the A data points adjacent to the right side of data point O).

[0068] As one example, the expression for calculating the isolated noise anomaly probability of data point O in a grayscale scatter plot is:

[0069]

[0070] In the formula, P' O θ represents the probability of isolated noise anomalies at data point O; O,m μ represents the tilt angle between data point O and the m-th adjacent data point on both sides. O ΔT is the mean of the adjacent tilt angles of data point O; O,min || represents the minimum grayscale difference between data point O and other data points in the grayscale scatter plot; || represents the absolute value sign.

[0071] It should be noted that |θ O,m -μ O This is used to describe the grayscale changes of data points in a grayscale scatter plot. The greater the difference between the tilt angle of data point O and the m-th adjacent data point on both sides and the average tilt angle of adjacent data point O, the greater the degree of grayscale change of data point O, and the greater the probability that the data point is an anomalous scatter plot caused by isolated noise. The greater the difference between the grayscale value of data point O and other data points in the grayscale scatter plot, the greater the probability that the point is an isolated point, and the greater the probability that the data point is an anomalous scatter plot caused by isolated noise.

[0072] It should be further noted that when a battery malfunctions and explodes, the heat in the battery area gradually increases, causing the grayscale value of each pixel in the battery area to gradually increase, exhibiting an irreversible trend. However, when the battery casing detaches, it also causes a short-term heat change, affecting the accuracy of subsequent battery health monitoring. On the thermal imaging map, the detached battery casing appears as a bright-dark-bright color change, while the grayscale value of the pixels in the battery area changes from large to small to large, demonstrating a reversible characteristic. In the aforementioned grayscale scatter plot, when the battery casing detaches, the pixels in the battery area exhibit a continuous trough phenomenon within a one-second timeframe. Therefore, the closer the grayscale values ​​of the peaks on either side of any abnormal scatter point are to the trough of the grayscale scatter plot, the greater the probability that the abnormal point is caused by the detached battery casing. Based on these characteristics, the probability of a data point being an abnormal point caused by the detached battery casing is constructed, and this probability characterizes the likelihood that the data point is an abnormal point caused by the detached battery casing.

[0073] As one example, the expression for calculating the abnormal probability of battery casing detachment at data point O in the grayscale scatter plot of pixel Q is:

[0074]

[0075] i∈{1,2……m1 O},j∈{1,2……m2 O}

[0076] In the formula, P” O T represents the probability of an abnormal battery casing detachment at data point O; O T represents the grayscale value of data point O; O,i T represents the grayscale value of the i-th data point to the left of data point O; O,j F represents the grayscale value of the j-th data point to the right of data point O; O This represents the trough filtering parameter for data point O; MIN() means taking the minimum value among all values; || represents the absolute value sign; m1 O This represents the number of all data points to the left of data point O; m2 O This represents the number of all data points to the right of data point O; α is a preset first parameter to prevent the denominator from being 0, which would render the formula meaningless.

[0077] It should be further explained that F oThe valley filtering parameter represents the data point O. It filters all valley points in the grayscale scatter plot. If any data point in the grayscale scatter plot is located in a valley position, the data point may be an abnormal scatter point, and the abnormal scatter point filtering parameter is set to 1; otherwise, the abnormal scatter point filtering parameter is set to 0. In addition, the first parameter α is preset to 1 in this embodiment, which can be adjusted according to the actual situation. This embodiment does not impose specific limitations.

[0078] As one embodiment, the method for obtaining the trough screening parameters is as follows:

[0079] Preset valley filtering parameters; for data point O, obtain the absolute value of the difference between the gray values ​​of data point O and all data points on the left and right sides, and record them as the gray value difference between the left and right neighbors of data point O. When both the gray value difference between the left and right neighbors of data point O are greater than 0, the valley filtering parameter is equal to 1; otherwise, the valley filtering parameter is equal to 0.

[0080] It should be noted that when the battery health condition is abnormal, its temperature will rise rapidly and spread to the surrounding area. Therefore, for pixel Q within the battery region, its temperature change trend is basically consistent with the temperature change trend of all pixels in its eight neighboring regions; however, in terms of timing, the temperature change of other pixels in its eight neighboring regions is slightly delayed or advanced compared to pixel Q. Therefore, based on the aforementioned characteristics, similarity trend parameters are obtained for any data point in the grayscale scatter plot of any pixel within the battery region.

[0081] Specifically, in step 2.2, firstly, for data point O on the grayscale scatter plot of pixel Q, its grayscale value is taken with that of N adjacent data points in the time series (where N is a preset second parameter; in this embodiment, N=2, which can be adjusted according to the actual situation; this embodiment does not impose a specific limitation) and recorded as the grayscale dataset of data point O. The Pearson correlation coefficient between the grayscale dataset of data point O and the grayscale datasets of all data points in the grayscale scatter plot of any pixel is calculated, and the maximum value of the Pearson correlation coefficient is obtained. The data points corresponding to the two grayscale datasets in the maximum Pearson correlation coefficient are taken, and the other data point excluding data point O is recorded as a similar data point of data point O in the grayscale scatter plot of any pixel. The time series difference between data point O and its similar data points in their respective grayscale scatter plots is recorded as the similarity time series difference.

[0082] Then, based on the grayscale dataset of data point O in the grayscale scatter plot of pixel Q and the grayscale dataset of all data points in the grayscale scatter plot of the j-th pixel within the eight neighborhoods of pixel Q, the similarity time difference between pixel Q and its similar data points in the grayscale scatter plot of the j-th pixel is obtained. Finally, the similarity time difference of all pixels in the eight neighborhoods of pixel Q is traversed and summed to obtain the similarity trend parameter of data point O in the grayscale scatter plot of pixel Q.

[0083] As one example, the specific expression for calculating the similarity trend parameter of data point O in the grayscale scatter plot of pixel Q is:

[0084]

[0085] In the formula, S O The parameter t represents the similarity trend of data point O in the grayscale scatter plot of pixel Q; o This represents the temporal sequence of data point O in the grayscale scatter plot of pixel Q; t l,o represents the temporal sequence of similar data points in the grayscale scatter plot of the l-th pixel O; B represents the number of pixels in the eight-neighborhood of pixel Q; exp() represents the exponential function with the natural constant as the base; || represents the absolute value sign.

[0086] It should be noted that the smaller the similarity time difference between data point O in the grayscale scatter plot of pixel Q and similar data points in the grayscale scatter plots of all pixels in the neighborhood of that pixel, the higher the similarity between data point O and the similar data points. Therefore, the smaller the similarity trend parameter of data point O, the greater the normality of data point O.

[0087] It should be further noted that when obtaining the pixels in the eight-neighborhood of pixel Q, there is a problem of exceeding the boundary when pixel Q is located at the boundary of the battery area. Therefore, in this embodiment, the part that exceeds the battery area is not considered. Only the pixels in the battery area are obtained for the pixels in the eight-neighborhood of pixel Q. The specific number of pixels in the eight-neighborhood is subject to the actual number obtained.

[0088] The above obtains the isolated noise anomaly probability and the battery casing detachment anomaly probability of each data point in the grayscale scatter plot of any pixel, as well as the similarity trend parameter of each data point in the grayscale scatter plot; therefore, it is possible to obtain the degree to which each data point in the grayscale scatter plot belongs to normal data points, and then a preset threshold is set to remove all abnormal scatter points.

[0089] As one example, the expression for calculating the normality of data point O in the grayscale scatter plot of pixel Q is:

[0090]

[0091] In the formula, P O Indicates the normality of data point O; P' O P represents the probability of isolated noise anomalies at data point O; O S represents the probability of the battery casing detaching abnormally at data point O; OThe similarity trend parameter represents the data point O; α is the preset first parameter; sigmoid() represents the sigmoid normalization function, which is only used to map the output result to the range (0,1).

[0092] It should be noted that the lower the probability of isolated noise anomalies for any data point in the grayscale scatter plot of any pixel, the greater the normality of that data point; the lower the probability of battery casing detachment anomalies for that data point, the greater the normality of that data point; and the higher the similarity trend parameter of a data point, the greater the normality of that data point.

[0093] Finally, the normality of each data point in the grayscale scatter plot of any pixel is obtained. When the normality is greater than the preset threshold θ, it means that the probability of this data point being a normal point is relatively high, so the data point is determined to be a normal point.

[0094] This embodiment distinguishes anomalies in all data points in the grayscale scatter plot by giving a threshold θ and removes abnormal scatter points; in addition, the threshold θ = 0.8 is preset based on experience and can be adjusted according to the actual situation. This embodiment does not impose specific limitations.

[0095] After removing outliers, a multinomial regression is used to fit all data points in the grayscale scatter plot to obtain the grayscale change curve of the pixel over time. Then, the average curvature of all data points on the grayscale change curve is obtained. At the same time, the Euclidean distance between the pixel and the centroid of the battery region is set as the weight of the average curvature of the grayscale change curve, and the average curvature is adjusted. Finally, the average curvature of the grayscale change curve of all pixels in the battery region is adjusted, and then the results are summed to obtain the temperature change index of the battery region under the infrared thermal imager.

[0096] As one embodiment, the expression for calculating the temperature change index of the battery area under an infrared thermal imager is as follows:

[0097]

[0098] In the formula, E represents the temperature change index of the battery region; n represents the number of pixels in the battery region; D k ρ represents the Euclidean distance of pixel k from the centroid of the battery region; k It represents the average curvature of the grayscale change curve of pixel k over time.

[0099] It should be noted that the average curvature of the grayscale change curve of each pixel in time represents the temperature change of that pixel in consecutive frames. The larger the average curvature, the greater the temperature change index. Meanwhile, the reciprocal of the Euclidean distance between any pixel and the centroid of the battery region is used as the temperature change weight for that pixel. The smaller the Euclidean distance between the pixel and the centroid of the battery region, the greater the temperature change weight, indicating that the temperature change of that pixel is more effective in the overall temperature change index of the battery region; the closer the pixel is to the edge of the battery region, the smaller the temperature change weight.

[0100] Thus, the temperature change index of the entire battery area under infrared thermal imaging was obtained using the above method.

[0101] The battery health monitoring module 103 obtains battery heat change parameters through the battery temperature change index and monitors battery health based on the battery heat change parameters.

[0102] It should be noted that the battery's thermal change parameter is obtained through the battery's temperature change index. The larger the thermal change parameter, the more abnormal the battery's health condition. By setting a thermal threshold to judge the thermal change parameter, if the battery's health condition is abnormal, corresponding measures will be taken in a timely manner.

[0103] Specifically, based on the battery temperature change index obtained from the temperature change detection module, the heat change parameters of the battery area are obtained using the heat conduction formula. A preset heat threshold is used to judge the battery health status. When the heat change parameters are greater than the heat threshold, the battery health is problematic, and there may be overheating or an impending explosion.

[0104] It should be further noted that, based on experience, the preset heat threshold σ = 100 kilojoules is used. σ can be adjusted according to the actual situation, and this embodiment does not impose a specific limitation.

[0105] This concludes the embodiment.

[0106] It should be noted that the exp(-x) model used in this embodiment is only used to represent negative correlation and constrain the output of the model to be within the (0,1) interval. In specific implementation, it can be replaced by other models with the same purpose. This embodiment only uses the exp(-x) model as an example for description and does not make specific limitations on it, where x refers to the input of the model.

[0107] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A real-time battery health monitoring system based on an infrared thermal imager, characterized in that, The system includes the following modules: The image acquisition module is used to acquire continuous thermal imaging images of the battery and obtain the battery area in each thermal imaging image. The temperature change detection module is used to obtain a grayscale scatter plot of any pixel in the battery area based on the grayscale value of any pixel in the time sequence; to obtain the isolated noise anomaly probability of any data point by the grayscale change of the data points in the grayscale scatter plot; and to obtain the battery casing detachment anomaly probability of the data point based on the difference in grayscale value between the trough data point and the adjacent data points in the grayscale scatter plot. Based on the temporal difference of grayscale change between a pixel and other pixels in its neighborhood, obtain the similarity trend parameter of each data point in the grayscale scatter plot of the pixel; based on the isolated noise anomaly probability, the battery casing detachment anomaly probability, and the similarity trend parameter of the data points in the grayscale scatter plot, obtain the normal data points in the grayscale scatter plot. Based on all normal data points in the grayscale scatter plot of the pixels, obtain the grayscale distribution of the pixels in time sequence; based on the grayscale distribution of all pixels in the battery area, obtain the temperature change index of the battery area. The battery health monitoring module is used to obtain the thermal change parameters of the battery area through the temperature change index of the battery area, and to monitor the battery health based on the thermal change parameters of the battery area.

2. The real-time battery health monitoring system based on an infrared thermal imager according to claim 1, characterized in that, The specific method for obtaining a grayscale scatter plot of any pixel based on its grayscale value over time within the battery area includes: For any pixel in the battery region at any coordinate, construct a grayscale scatter plot of its grayscale value over time. The horizontal axis of the grayscale scatter plot represents the time interval of consecutive frames of battery thermal imaging images per second, and the vertical axis represents the grayscale value of the pixel in each frame.

3. The real-time battery health monitoring system based on an infrared thermal imager according to claim 1, characterized in that, The specific method for obtaining the isolated noise anomaly probability of any data point by analyzing the grayscale changes of data points in a grayscale scatter plot includes: Let any pixel in the battery area be labeled as pixel Q, and any data point in the grayscale scatter plot of pixel Q be labeled as data point O; let the angle between the line connecting the coordinates of data point O and other data points in the grayscale scatter plot and the x-axis of the grayscale scatter plot be labeled as the tilt angle between data point O and other data points. The mean of the tilt angles between data point O and the A adjacent data points on both sides is denoted as the mean of the adjacent tilt angles of data point O, where A is a preset parameter for adjacent data points; The specific method for calculating the probability of isolated noise anomalies at data point O in a grayscale scatter plot is as follows: In the formula, P' O This represents the probability of isolated noise anomalies at data point O; θ O,m μ represents the tilt angle between data point O and the m-th adjacent data point on both sides. O ΔT is the mean of the adjacent tilt angles of data point O; O,min This represents the minimum grayscale difference between data point O and other data points in the grayscale scatter plot; | represents the absolute value sign.

4. The real-time battery health monitoring system based on an infrared thermal imager according to claim 1, characterized in that, The method for obtaining the probability of abnormal battery casing detachment based on the difference in grayscale values ​​between the trough data points and their adjacent data points in the grayscale scatter plot includes the following specific methods: Obtain the valley filtering parameters in the grayscale scatter plot of pixel Q; The specific calculation method for the probability of abnormal battery casing detachment at data point O in the grayscale scatter plot is as follows: In the formula, P” O T represents the probability of an abnormal battery casing detachment at data point O; O T represents the grayscale value of data point O; O,i T represents the grayscale value of the i-th data point to the left of data point O; O,j F represents the grayscale value of the j-th data point to the right of data point O; O This represents the trough filtering parameter for data point O; MIN() represents taking the minimum value among all values; | represents the absolute value sign; m1 O This represents the number of all data points to the left of data point O; m2 O This indicates the number of all data points to the right of data point O; α is the preset first parameter.

5. The real-time battery health monitoring system based on an infrared thermal imager according to claim 4, characterized in that, The specific method for obtaining the valley screening parameters is as follows: Preset trough filtering parameters; For data point O, obtain the absolute value of the difference between the gray values ​​of data point O and all data points on the left and right sides, and record it as the gray value difference between the left and right neighbors of data point O. When both the gray value difference between the left and right neighbors of data point O are greater than 0, the valley filtering parameter is equal to 1. Conversely, the trough filtering parameter is equal to 0.

6. The real-time battery health monitoring system based on an infrared thermal imager according to claim 1, characterized in that, The specific method for obtaining the similarity trend parameter of each data point in the grayscale scatter plot of a pixel based on the temporal difference of grayscale changes between the pixel and other pixels in its neighborhood includes: For a data point O in the grayscale scatter plot of pixel Q, the grayscale values ​​of data point O and N adjacent data points in the time series are taken as the grayscale dataset of data point O, where N is a preset second parameter; the Pearson correlation coefficient between the grayscale dataset of data point O and the grayscale datasets of all data points in the grayscale scatter plot of any pixel is obtained, and the maximum value of the Pearson correlation coefficient is selected; the data points corresponding to the two grayscale datasets in the maximum value of the Pearson correlation coefficient are taken, and the other data point excluding data point O is taken as the similar data point of data point O in the grayscale scatter plot of any pixel; the time series difference between data point O and the corresponding similar data point in their respective grayscale scatter plots is taken as the similarity time series difference; Based on the grayscale dataset of data point O in the grayscale scatter plot of pixel Q and the grayscale dataset of all data points in the grayscale scatter plot of the j-th pixel in the eight neighborhood of pixel Q, obtain the similarity time difference between pixel Q and similar data points in the grayscale scatter plot of the j-th pixel in the eight neighborhood. In the formula, S O The similarity trend parameter represents the data point O in the grayscale scatter plot of pixel Q; t o The temporal sequence of data point O in the grayscale scatter plot of pixel Q; t l,o represents the temporal sequence of similar data points in the grayscale scatter plot of the first pixel O; B represents the number of pixels in the eight-neighborhood of pixel Q; exp() represents an exponential function with the natural constant as the base; || represents the absolute value sign.

7. The real-time battery health monitoring system based on an infrared thermal imager according to claim 1, characterized in that, The specific method for obtaining normal data points in the grayscale scatter plot based on the isolated noise anomaly probability, the battery casing detachment anomaly probability, and similarity trend parameters is as follows: Based on the probability of isolated noise anomalies, the probability of battery casing detachment anomalies, and similarity trend parameters of data points in the grayscale scatter plot, the normality of all data points in the grayscale scatter plot is obtained. When the normality of a data point is greater than the preset threshold θ, the data point is determined to be a normal data point.

8. The real-time battery health monitoring system based on an infrared thermal imager according to claim 7, characterized in that, The specific method for obtaining the normality level of the data points is as follows: Obtain the isolated noise anomaly probability, battery casing detachment anomaly probability, and similarity trend parameters for any data point in the grayscale scatter plot; The specific method for calculating the normality of data point O in a grayscale scatter plot is as follows: In the formula, P O Indicates the normality of data point O; P' O P represents the probability of isolated noise anomalies at data point O; O This represents the probability of the battery casing falling off at data point O. S O The similarity trend parameter represents the data point O; α is the preset first parameter; sigmoid() represents the sigmoid normalization function.

9. The real-time battery health monitoring system based on an infrared thermal imager according to claim 1, characterized in that, The specific methods for obtaining the grayscale distribution of pixels over time based on all normal data points in the grayscale scatter plot of pixels, and obtaining the temperature change index of the battery area based on the grayscale distribution of all pixels within the battery area, include: The grayscale scatter plot is fitted with a multinomial regression to obtain the grayscale change curve of the pixel over time. The average curvature of all data points on the grayscale change curve is calculated. The average curvature of the grayscale change curve of all pixels in the battery area is obtained. The specific calculation method for the temperature change index of the battery area under infrared thermal imager is as follows: In the formula, E represents the temperature change index of the battery region; n represents the number of pixels in the battery region; D k ρ represents the Euclidean distance from the k-th pixel in the battery region to the centroid of the battery region; k This represents the average curvature of the grayscale change curve of the k-th pixel in the battery region over time.

10. The real-time battery health monitoring system based on an infrared thermal imager according to claim 1, characterized in that, The method for obtaining thermal change parameters of the battery region through the temperature change index and monitoring battery health based on these parameters includes the following specific methods: Based on the temperature change index of the battery region, the heat change parameters of the battery region are obtained using the heat conduction formula. A preset heat threshold is used to judge the health status of the battery. When the heat change parameter of the battery area exceeds the heat threshold, the battery health is found to be problematic, and an early warning is issued.

Images