An infrared image enhancement method for computer room monitoring

Abstract

The present invention relates to the field of image processing, and more specifically to an infrared image enhancement method for computer room monitoring, the method comprising: obtaining a preprocessed computer room infrared monitoring image, and evenly dividing the infrared monitoring image into a plurality of blocks; constructing a corresponding local window centered on each pixel in the blocks, and calculating a heat source anomaly index of the pixel; calculating a gain coefficient for the corresponding pixel when performing enhancement based on the heat source anomaly index, taking the computer room infrared monitoring image as input, and performing enhancement processing on the computer room infrared monitoring image based on the gain coefficient. The present invention enhances the computer room infrared monitoring image by calculating the heat source anomaly index of the pixel and determining the gain coefficient accordingly, thereby effectively highlighting key details in the image, making originally blurred or difficult-to-distinguish parts clearer, and facilitating monitoring personnel to more accurately observe the operating status of equipment in the computer room.

Description

Technical Field

[0001] The present invention relates to the field of image processing, and more particularly to an infrared image enhancement method for monitoring a machine room. Background Art

[0002] The computer room monitoring system is a comprehensive solution for ensuring room safety and stable equipment operation. Infrared video surveillance systems are crucial. This system utilizes passive infrared thermal imaging or active infrared fill-light technology to achieve 24-hour, uninterrupted monitoring by sensing infrared radiation or reflected light from objects. Even in extreme environments without visible light, it can produce clear thermal or near-infrared images, overcoming the lighting limitations of traditional monitoring. This ensures complete monitoring coverage in challenging conditions such as nighttime, equipment obstruction, and power outages, providing reliable assurance of room safety.

[0003] Prior art, such as the patent application document with publication number CN118154491A, discloses an infrared image processing method, device, and storage medium. The infrared image processing method first compresses the original infrared image, then uses target detection and instance segmentation models to process the target infrared image, identifies and segments the area to be processed containing the target power device, and then performs detail enhancement processing on the area to be processed, including grayscale enhancement and noise suppression operations on the base layer and the detail layer, respectively, to enhance the detail information of the image.

[0004] However, in a computer room, when the equipment is operating at a low load and the ambient temperature is relatively stable, the difference in infrared radiation intensity between the equipment surface and the surrounding environment is small. At this time, the contrast of the infrared image processed according to existing technology may still be insufficient for monitoring personnel to quickly and accurately detect potential abnormalities, such as local overheating of the equipment and other hidden faults, thereby affecting the effectiveness of the computer room monitoring system and failing to fully play its role in ensuring the safety of the computer room and the stable operation of the equipment. Summary of the Invention

[0005] In order to solve the above-mentioned technical problem that it is difficult to effectively highlight the characteristics of the device due to the limitations of infrared image enhancement processing, the present invention provides the following technical solution.

[0006] An infrared image enhancement method for computer room monitoring, comprising:

[0007] Obtain a pre-processed infrared surveillance image of the computer room and evenly divide the infrared surveillance image into multiple blocks;

[0008] Based on the block, a corresponding local window is constructed with each pixel as the center, and the heat source anomaly index of the pixel is calculated;

[0009] The gain coefficient of the corresponding pixel point when performing enhancement is calculated according to the heat source anomaly index, and the infrared monitoring image of the computer room is used as input and enhanced based on the gain coefficient;

[0010] Among them, the temperature feature vector of the pixel point is calculated, clustering is performed according to the temperature feature vector, and the pixel point connected domain analysis is performed on the clustering result to obtain multiple iso-feature domains. The area change and temperature change of the iso-feature domains to which the same pixel point belongs between a set number of selected image frames are compared to obtain the heat source anomaly index of the pixel point; the temperature feature vector includes the temperature smoothing index within the local window of the pixel point.

[0011] The present invention calculates features such as the temperature smoothing index within a local window of a pixel point, combines cluster analysis with connected domain analysis, and realizes dual verification from local temperature distribution to spatial topological structure; by comparing the area change and temperature change of the iso-feature domain to which the same pixel point belongs between a set number of selected image frames, it can be determined whether the heat source is abnormal. If the area of ​​the iso-feature domain to which a pixel point belongs increases sharply in a short period of time, and the temperature also rises significantly, then the heat source anomaly index of this pixel point will be very high, which may indicate an abnormality such as overheating of the equipment in the computer room; based on the heat source anomaly index, the gain coefficient is calculated, and the abnormal area is differentially enhanced (such as the brightness of the high-temperature area is increased), while retaining the details of the normal area to avoid image distortion caused by global over-enhancement, so that monitoring personnel can more intuitively find the location of the heat source and take timely measures.

[0012] Preferably, the temperature feature vector also includes a temperature distribution anomaly index within a local window of a pixel point.

[0013] Preferably, the process of obtaining the temperature smoothing index includes:

[0014] Obtaining the gradient magnitude and gradient angle of all pixels in the block;

[0015] The number of times the mode of the gradient angle within the local window of the pixel point occurs and the difference between the gradient angle of each pixel point and the mode of the gradient angle are obtained, and then the temperature smoothing index within the local window of the pixel point is calculated based on the gradient size and gradient angle.

[0016] The gradient magnitude reflects the severity of temperature changes, while the gradient angle indicates the direction of temperature change. Combined, these two factors reflect the smoothness of the temperature distribution within a local area. In a device's normal heat dissipation area, temperature changes may be relatively gentle, with a small gradient. In an abnormal heat source area, however, temperature changes may be more dramatic, with a larger gradient.

[0017] Preferably, the process of obtaining the temperature distribution anomaly index includes:

[0018] Divide the grayscale values ​​and gradients of all pixels contained in the local pixel window into a set number of grayscale intervals and gradient intervals respectively;

[0019] Count the number of pixels in each combination of grayscale interval and gradient interval;

[0020] The heat source weight of each pixel in the local window is calculated, and the total number of pixels in any combination of grayscale interval and gradient interval is multiplied by the heat source weight of all pixels in the combination and the sum is calculated to obtain the temperature distribution anomaly index.

[0021] By dividing the grayscale value (reflecting temperature) and gradient value (reflecting the temperature change rate) into intervals, the characteristics of high temperature areas and temperature mutation areas can be captured at the same time. By assigning a heat source weight to each combination of grayscale interval and gradient interval, the contribution of abnormal signals can be highlighted and normal background noise can be suppressed.

[0022] Preferably, the clustering is a K-means clustering algorithm.

[0023] Preferably, the process of obtaining the heat source anomaly index includes:

[0024] Calculate the relative change rate of the pixel based on the area change and temperature change of the iso-feature domain to which the pixel belongs between the current frame and the previous frame;

[0025] Traverse all frames of infrared monitoring images of the computer room before the current frame, calculate the relative change rate of pixels between two adjacent frames, sum and average them, and obtain the heat source anomaly index of the pixel.

[0026] When equipment in a computer room first starts to overheat abnormally, the temperature change may be relatively small. By calculating the relative rate of change between the current frame and the previous frame (combining area and temperature changes), sudden heat source anomalies in a short period of time can be quickly identified. Then, all historical frames are traversed and averaged to reflect the cumulative effect of heat source anomalies, thereby capturing small temperature changes.

[0027] Preferably, the gain coefficient satisfies the relationship:

[0028] Where, Indicates the pixel of the current frame The gain factor when performing enhancement, Indicates the pixel of the current frame The heat source anomaly index, Indicates the mean heat source anomaly index of all pixels in the current frame.

[0029] Preferably, the relative change rate satisfies the relationship:

[0030] Where, Represents pixel points The relative rate of change between the current frame and the previous frame of infrared surveillance image, Pixel The area of ​​the iso-feature domain in the current frame, Pixel The area of ​​the iso-feature domain in the frame before the current frame, Pixel The grayscale mean of all pixels in the equal feature domain in the current frame, Pixel The grayscale mean of all pixels in the same feature domain in the previous frame of the current frame; is a constant and 1 is a hyperparameter.

[0031] By combining the joint changes of the spatial range and thermal radiation intensity of the pixel area with the Sigmoid function to achieve nonlinear mapping, the spatiotemporal dynamic characteristics of the pixel points are converted into comparable indicators. This not only improves the positioning accuracy of single-frame anomalies, but also realizes the prediction of anomaly development trends through time series accumulation.

[0032] Preferably, the process of obtaining the temperature smoothing index includes:

[0033] The entropy of the grayscale distribution in the local window is calculated, and the inverse of the entropy of the grayscale distribution in the local window is used as the temperature smoothing index.

[0034] Preferably, the process of obtaining the temperature distribution anomaly index includes:

[0035] Calculate the grayscale mean and grayscale standard deviation within the local window of the pixel;

[0036] The temperature distribution anomaly index is obtained by calculating the difference between the grayscale value of the pixel and the grayscale mean value in the local window and dividing the difference by the grayscale standard deviation.

[0037] The beneficial effects of the present invention are:

[0038] This method comprehensively considers multiple factors, including the temperature smoothing index and temperature distribution anomaly index within a pixel's local window, to more comprehensively and accurately assess heat source anomalies at that pixel. By comparing the area and temperature changes within the iso-feature domains belonging to the same pixel across a set number of selected image frames, it can effectively identify abnormally hot areas within equipment within the computer room.

[0039] Compared with traditional image enhancement methods, through a series of feature analysis and clustering steps, it can more accurately distinguish normal heat sources from abnormal heat sources, reduce the misjudgment rate caused by poor image quality or interference factors, and improve the reliability and accuracy of the monitoring system. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 This is a method flow chart of steps S1 to S3 in an infrared image enhancement method for computer room monitoring according to an embodiment of the present invention. DETAILED DESCRIPTION

[0041] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, but not all of the embodiments.

[0042] Reference Figure 1 , an infrared image enhancement method for computer room monitoring includes steps S1 to S3, specifically as follows:

[0043] S1: Obtain a pre-processed infrared surveillance image of the computer room and evenly divide the infrared surveillance image into multiple blocks.

[0044] In the computer room, the equipment generates heat when it is running, which can be captured by infrared cameras. Infrared cameras can detect the thermal radiation of objects and convert it into image signals.

[0045] For infrared surveillance video captured by an infrared camera, a single-frame infrared surveillance image is extracted from the video. In the embodiment of the present invention, the single-frame infrared surveillance image is an 8-bit grayscale image, which means that the grayscale value of each pixel ranges from 0 to 255. The higher the grayscale value, the higher the thermal radiation intensity at that point, which generally also means a higher temperature.

[0046] Furthermore, denoising is performed on the single-frame infrared monitoring image, such as median filtering denoising, to obtain a preprocessed computer room infrared monitoring image.

[0047] After denoising, the image is divided into blocks so that each block can be analyzed and processed independently. For example, if the image resolution is 640×480, it can be divided into multiple 80×60 sub-blocks.

[0048] S2: Based on the block, a corresponding local window is constructed with each pixel point as the center, and the heat source anomaly index of the pixel point is calculated.

[0049] When equipment in a computer room is operating normally, its heat source distribution is relatively uniform and stable. However, when equipment malfunctions or overheats, localized bright areas appear in the infrared image. These areas exhibit significant temperature gradient variations, significantly deviating from the normal heat source distribution pattern. By calculating the heat source anomaly index, we can quantify whether the heat source at each pixel represents an anomaly, thereby helping to identify these abnormal heat sources.

[0050] Specifically, the thermal radiation of normal equipment has the characteristics of radial diffusion, while abnormal heat sources (such as short circuit and overload) show random direction and drastic changes, that is, the gradient direction is chaotic (angle dispersion) and the gradient size suddenly changes (dramatic temperature changes).

[0051] In one embodiment, whether a local window constructed with a pixel point as the center conforms to the thermal radiation pattern of a normal heat source is determined by quantifying the consistency of the gradient direction and the smoothness of the gradient magnitude within the local window.

[0052] Taking any block obtained by dividing the infrared surveillance image of the current frame in S1 as an example, we use it as the target block and construct a local window centered on any pixel in the target block. For example, the size of the local window is set to 7×7 or 9×9. Then, we use the Sobel operator to process the block and obtain the gradient magnitude and gradient angle of all pixels in the block. The gradient angle range is 0° to 360°.

[0053] Furthermore, the number of occurrences of the mode of the gradient angle within the local window of the pixel point and the difference between the gradient angle of each pixel point and the mode of the gradient angle are obtained, and then the temperature smoothing index within the local window of the pixel point is calculated based on the gradient size and gradient angle.

[0054] Exemplarily, the temperature smoothing index satisfies the following relationship:

[0055]

[0056] Pixel The temperature smoothing index within the local window of Pixel The number of times the mode of the gradient angle in the local window appears (i.e. the number of pixels in the main direction), Pixel The total number of pixels contained in the local window, Pixel Pixels in the local window The absolute value of the difference between the gradient angle of and the mode of the gradient angle in the local window, Pixel Pixels in the local window The gradient size of . The 1 in the denominator is a hyperparameter used to avoid the denominator being zero.

[0057] in, Reflects the consistency of the gradient direction within the local window. When the main direction pixel ratio is high (such as edge or smooth area), Close to 1; within the local window, the smaller the difference between the gradient angle of different pixels and the mode angle, that is, The smaller it is, the more uniform the heat radiation direction of the heat source is, and the more it conforms to the characteristics of a normal heat source. At the same time, the smaller the gradient of the pixel point in the local window is, The smaller the value, the smoother the temperature gradient changes, and the more consistent it is with the characteristics of a normal heat source. Therefore, the larger the calculated temperature smoothness index, the smoother the temperature changes at the pixel and its surrounding area, and the more likely it is a normal heat source.

[0058] In one embodiment, the specific characteristics of temperature distribution are also taken into account. Heat from a normal heat source gradually diffuses into the surrounding environment along a heat dissipation path, resulting in a gradually darkening temperature gradient in the infrared image. This means the temperature remains within the normal range and the temperature gradient is relatively small. In contrast, heat from an abnormal heat source may diffuse over a smaller range and be concentrated in a localized area. This results in an isolated bright area in the infrared image. This means the temperature exceeds the normal range and the temperature gradient is relatively large.

[0059] Still taking any block obtained by the above S1 division as an example, it is used as the target block, and the above-mentioned local window of the same size is constructed with any pixel point in the target block as the center.

[0060] Furthermore, the grayscale value ranges of all pixels contained in the pixel local window are divided into A=5 grayscale intervals, and the gradient magnitude ranges of all pixels contained in the pixel local window are divided into B=6 gradient intervals.

[0061] Count the number of pixels in each grayscale interval and gradient interval combination to form an A×B matrix. For example, Indicates the number of pixels whose grayscale is in interval a and whose gradient is in interval b.

[0062] Next, calculate the heat source weight of each pixel in the local window, and then convert the above Multiply the heat source weight of the corresponding pixel point and sum it up to get the temperature distribution anomaly index in the local window.

[0063] Then the above temperature distribution anomaly index satisfies the relationship:

[0064]

[0065] Where, Pixel The temperature distribution anomaly index within the local window is Represents the interval index of the grayscale value in the local window, ranging from 1 to A; Represents the number of intervals of pixel grayscale values ​​within the local window; Represents the interval index of the gradient in the local window, ranging from 1 to B; Indicates the number of intervals of pixel gradient magnitude within the local window; is the heat source weight of the pixel whose grayscale is in interval a and gradient is in interval b, reflecting the combined influence of grayscale and gradient.

[0066] in, It increases with the increase of grayscale interval and gradient interval, strengthening the contribution of "high grayscale + high gradient" area, making the anomaly index more sensitive to areas with drastic edge changes or mutations.

[0067] If the pixels with larger grayscale and gradient values ​​are distributed in the lower right corner of the matrix, that is, when a and b are larger, the corresponding The larger the value is, the higher the temperature in the window is and the greater the temperature variation is, the greater the abnormal distribution of the heat source is, and therefore the larger the calculated temperature distribution anomaly index is.

[0068] In summary, the temperature smoothness index and the temperature anomaly index evaluate the temperature distribution within a local window from different perspectives. The temperature smoothness index focuses on the smoothness of temperature changes, while the temperature anomaly index focuses on the specific characteristics of the temperature distribution. By combining these two indices, a more comprehensive assessment of the temperature distribution within a local window can be achieved, leading to more accurate identification and differentiation of normal and abnormal heat sources.

[0069] According to the above operation, the temperature smoothing index and temperature distribution anomaly index in the local window of all pixels in the target block can be calculated by the same logic. Furthermore, the temperature smoothing index and temperature distribution anomaly index in the local window of all pixels on the infrared monitoring image can be obtained by the same logic.

[0070] In another embodiment, the entropy of the grayscale distribution in the local window may be calculated, and the inverse of the entropy of the grayscale distribution in the local window may be used as the temperature smoothing index.

[0071] In another embodiment, a simple method for calculating the temperature distribution anomaly index is provided, namely:

[0072] The grayscale mean and grayscale standard deviation in the local window of the pixel are calculated; the difference between the grayscale value of the pixel and the grayscale mean in the local window is calculated and divided by the grayscale standard deviation to obtain the temperature distribution anomaly index.

[0073] In one embodiment, the temperature smoothing index and the temperature distribution anomaly index are combined into a temperature feature vector to describe the temperature variation pattern of a pixel. The temperature feature vectors of all pixels are used as input and clustered using the K-means clustering algorithm, where K is 4. After clustering, pixels belonging to the same cluster have similar temperature variation characteristics.

[0074] Connected domain analysis is performed on the pixels belonging to the same cluster to obtain several connected domains, which are called iso-feature domains. The pixels in the iso-feature domains have similar temperature change characteristics.

[0075] It should be noted that in the infrared surveillance video of the computer room, the monitoring position and the position of each device in the computer room remain fixed. Therefore, in the infrared images of the computer room in two adjacent frames, the actual positions represented by the pixels at the same position are also the same or extremely similar.

[0076] If the temperature of the iso-feature domain of a pixel at the same location in two consecutive frames of the computer room infrared image increases rapidly and over a larger area, this indicates a possible heat source anomaly, such as local overheating caused by a fault or localized heat release due to a short circuit. Therefore, for each pixel in the current frame of the infrared surveillance image, the area change and temperature change of the iso-feature domain in that pixel are compared with those in the previous M frames (e.g., the previous 20 frames) to calculate the heat source anomaly index for that pixel.

[0077] For example, the relative change rate between the current frame and the previous frame is calculated first, that is, the relationship is satisfied:

[0078]

[0079] Where, Represents pixel points The relative rate of change between the current frame and the previous frame of infrared surveillance image, Pixel The area of ​​the iso-feature domain in the current frame, Pixel The area of ​​the iso-feature domain in the frame before the current frame, Pixel The grayscale mean of all pixels in the equal feature domain in the current frame, Pixel The grayscale mean of all pixels in the same feature domain in the previous frame of the current frame. is a constant (usually taken to a very small value to ensure that the denominator is not zero), and 1 is used as a hyperparameter.

[0080] in, The combined effect area (spatial dimension) and temperature level (intensity dimension) of the heat source are used to measure the overall abnormality of the heat source. For example, if the temperature of a certain area rises slightly but the area expands significantly (such as poor heat dissipation), or if the area remains unchanged but the temperature rises suddenly (such as short circuit heat release), both may lead to The larger the ratio, the more significant the anomaly. Furthermore, after the ratio is reduced by 1, it is mapped to the interval [0, 1] using the Sigmoid function. When the ratio approaches 1 (i.e., a small change), a larger gradient is generated, thereby amplifying weak anomaly signals and improving detection sensitivity.

[0081] Finally, traverse the infrared surveillance images of the previous M frames of the current frame, calculate the relative change rate of the pixels between two adjacent frames, sum and average them, and get the pixel point Heat source anomaly index.

[0082] In general, the calculation of the heat source anomaly index for a pixel relies on the clustering results of the temperature smoothing index and the temperature distribution anomaly index (partitioning the equal-feature domain), thereby mapping spatial features to the temporal dimension. A larger temperature smoothing index indicates smoother temperature variations in the region, more consistent with the radial diffusion characteristics of a normal heat source. Since a normal heat source (i.e., a high temperature smoothing index) typically does not exhibit significant anomalous variations, the heat source anomaly index is relatively low. The temperature distribution anomaly index reflects the persistence of heat source anomalies. A high temperature distribution anomaly index (the presence of anomalous hotspots) may lead to an increase in the heat source anomaly index because the anomalous hotspots persist or expand over time.

[0083] S3: Calculate the gain coefficient of the corresponding pixel point when performing enhancement according to the heat source anomaly index, take the infrared monitoring image of the computer room as input, and perform enhancement processing on it based on the gain coefficient.

[0084] In the infrared image of the computer room, pixels of abnormal heat sources (such as equipment overload, line short circuit, etc.) need to be enhanced first so that operation and maintenance personnel can quickly locate the problem; enhancement should be reduced in areas with normal temperature to avoid excessive interference information.

[0085] By comparing the heat source anomaly index, an independent gain coefficient is calculated for each pixel point to achieve the effect of "more enhancement in abnormal areas and less enhancement in normal areas".

[0086] The gain coefficient of the pixel point calculated according to the heat source anomaly index satisfies the relationship:

[0087]

[0088] Where, Indicates the pixel of the current frame The gain factor when performing enhancement, Indicates the pixel of the current frame The heat source anomaly index, Indicates the mean heat source anomaly index of all pixels in the current frame.

[0089] Then, each frame of the infrared monitoring image of the computer room is used as input, and for each pixel in the image, its gain coefficient is calculated according to the above formula , use the calculated gain coefficient to enhance each pixel point, highlight the abnormal heat source part, and finally generate an enhanced infrared image of the computer room, which is convenient for monitoring personnel to identify abnormal heat sources more clearly and check safety hazards in time.

[0090] If the heat source anomaly index of a certain pixel point is greater than the average of the heat source anomaly indices of all pixel points, it means that the degree of heat source anomaly at the location represented by the pixel point is higher than the overall anomaly of the heat source in the image. In this case, the point is more likely to be an abnormal heat source and needs to be enhanced more significantly. Conversely, if the heat source anomaly index of a certain pixel point is less than the average of the heat source anomaly indices of all pixel points, it means that the degree of heat source anomaly at the location represented by the pixel point is lower than the overall anomaly of the heat source in the image. In this case, the point is more likely to be a normal heat source point and needs to be enhanced less.

[0091] At this point, the enhancement processing of the infrared image in the computer room monitoring is completed.

[0092] It should be noted that those skilled in the art may make various modifications and improvements without departing from the scope of the present invention, and these modifications and improvements fall within the scope of protection of the present invention. Therefore, the scope of protection of the patent for this invention shall be based on the appended claims.

Claims

1. An infrared image enhancement method for computer room monitoring, characterized in that: include: Obtain a pre-processed infrared surveillance image of the computer room and evenly divide the infrared surveillance image into multiple blocks; Based on the block, a corresponding local window is constructed with each pixel as the center, and the heat source anomaly index of the pixel is calculated; The gain coefficient of the corresponding pixel point when performing enhancement is calculated according to the heat source anomaly index, and the infrared monitoring image of the computer room is used as input and enhanced based on the gain coefficient; The temperature characteristic vector of the pixel point is calculated, clustering is performed based on the temperature characteristic vector, pixel connected domain analysis is performed on the clustering result to obtain multiple iso-characteristic domains, and the area change and temperature change of the iso-characteristic domains to which the same pixel point belongs between a set number of selected image frames are compared to obtain the heat source anomaly index of the pixel point; the temperature characteristic vector includes the temperature smoothing index within the local window of the pixel point and the temperature distribution anomaly index within the local window of the pixel point; The process of obtaining the temperature smoothing index includes: Obtaining the gradient magnitude and gradient angle of all pixels in the block; Obtain the number of occurrences of the mode of the gradient angle within the local window of the pixel point, as well as the difference between the gradient angle of each pixel point and the mode of the gradient angle, and then calculate the temperature smoothing index within the local window of the pixel point based on the gradient size and gradient angle; the temperature smoothing index satisfies: Pixel The temperature smoothing index within the local window of Pixel The number of times the mode of the gradient angle in the local window appears, that is, the number of pixels in the main direction, Pixel The total number of pixels contained in the local window, Pixel Pixels in the local window The absolute value of the difference between the gradient angle of and the mode of the gradient angle in the local window, Pixel Pixels in the local window The gradient size; The process of obtaining the temperature distribution anomaly index includes: Divide the grayscale values ​​and gradients of all pixels contained in the local pixel window into a set number of grayscale intervals and gradient intervals respectively; Count the number of pixels in each combination of grayscale interval and gradient interval; The heat source weight of each pixel in the local window is calculated, and the total number of pixels in any combination of grayscale interval and gradient interval is multiplied by the heat source weight of all pixels in the combination and the sum is calculated to obtain the temperature distribution anomaly index.

2. The infrared image enhancement method for computer room monitoring according to claim 1, characterized in that: The clustering is a K-means clustering algorithm.

3. The infrared image enhancement method for computer room monitoring according to claim 2, characterized in that: The process of obtaining the heat source anomaly index includes: Calculate the relative change rate of the pixel based on the area change and temperature change of the iso-feature domain to which the pixel belongs between the current frame and the previous frame; Traverse all frames of infrared monitoring images of the computer room before the current frame, calculate the relative change rate of pixels between two adjacent frames, sum and average them, and obtain the heat source anomaly index of the pixel.

4. The infrared image enhancement method for computer room monitoring according to claim 3, characterized in that: The gain coefficient satisfies the relationship: Where, Indicates the pixel of the current frame The gain coefficient when performing enhancement, Indicates the pixel of the current frame The heat source anomaly index, Indicates the mean heat source anomaly index of all pixels in the current frame.

5. The infrared image enhancement method for computer room monitoring according to claim 4, characterized in that: The relative rate of change satisfies the relationship: Where, Represents pixel points The relative rate of change between the current frame and the previous frame of infrared surveillance image, Pixel The area of ​​the iso-feature domain in the current frame, Pixel The area of ​​the iso-feature domain in the frame before the current frame, Pixel The grayscale mean of all pixels in the equal feature domain in the current frame, Pixel The grayscale mean of all pixels in the same feature domain in the previous frame of the current frame; is a constant and 1 is a hyperparameter.

6. The infrared image enhancement method for computer room monitoring according to claim 1, characterized in that: The process of obtaining the temperature smoothing index includes: The entropy of the grayscale distribution in the local window is calculated, and the inverse of the entropy of the grayscale distribution in the local window is used as the temperature smoothing index.

7. The infrared image enhancement method for computer room monitoring according to claim 1, characterized in that: The process of obtaining the temperature distribution anomaly index includes: Calculate the grayscale mean and grayscale standard deviation within the local window of the pixel; The temperature distribution anomaly index is obtained by calculating the difference between the grayscale value of the pixel and the grayscale mean value in the local window and dividing the difference by the grayscale standard deviation.

Images