A device energy efficiency state visual monitoring method based on infrared thermal image feature analysis

Abstract

This invention belongs to the field of image processing technology, specifically relating to a visual monitoring method for equipment energy efficiency status based on infrared thermal imaging feature analysis. The method includes: acquiring a single-frame infrared image of the equipment surface and converting it into a two-dimensional temperature matrix, while simultaneously generating an image temperature histogram; calculating the local thermal radiation convergence degree of each pixel based on the two-dimensional temperature matrix; simultaneously constructing the heat island salience potential energy of each pixel; performing frequency reconstruction on the image temperature histogram, calculating the histogram frequency reconstruction factor corresponding to each discrete temperature level; generating a reconstruction probability distribution, simultaneously executing the maximum inter-class variance algorithm to determine the optimal threshold, and using the optimal threshold to perform binarization segmentation of the two-dimensional temperature matrix to identify areas of insulation layer damage; determining the equipment energy efficiency status based on a preset threshold, and outputting the corresponding energy efficiency anomaly level. This invention improves the accuracy of identifying thermal leaks in infrared equipment.

Description

Technical Field

[0001] This invention relates to the field of image processing technology. More specifically, this invention relates to a visual monitoring method for the energy efficiency status of equipment based on infrared thermal imaging feature analysis. Background Technology

[0002] In continuous production in food processing enterprises, infrared thermal imaging technology is often used to monitor the local heat loss of the equipment's insulation layer due to its non-contact temperature measurement advantage. By periodically or online scanning the outer surface of the equipment with an infrared thermal imager, the temperature field distribution can be presented, providing a visual basis for energy efficiency assessment and maintenance decisions.

[0003] In order to extract damaged areas from complex thermodynamic backgrounds, existing technologies often use the maximum inter-class variance algorithm to find a threshold for automatic image segmentation. This algorithm relies on the inter-class variance statistics of the global histogram, which implicitly assumes that the image temperature histogram exhibits a clear bimodal distribution.

[0004] However, in the early stages of insulation layer damage, the proportion of pixels occupied by thermal leaks in the entire large equipment surface area is extremely small, resulting in a strong single-peak long-tail characteristic in the global histogram. Under this distribution, the global optimal threshold calculated by the algorithm will be dominated by the large area of ​​low-temperature normal insulation layer background and will be seriously high. This directly leads to the early tiny thermal leaks being regarded as background erased, causing extremely serious small target missed detection, making it impossible for the system to detect energy loss in the early stages. Since heat loss is cumulative and diffuse, the initial micron-level cracks may expand into centimeter-level damage within a few days, accelerating the aging of insulation materials and amplifying the thermal bridging effect.

[0005] Existing segmentation methods lack sensitivity to these low-contrast, small-scale anomalies, making it difficult to meet the stringent requirements of food companies for refined energy efficiency management and zero-accident operation. This not only weakens the core value of infrared thermal imaging technology in preventive maintenance, but may also lead to the continuous accumulation of energy waste, aggravated local overheating of equipment, and even induce safety accidents. Summary of the Invention

[0006] To address the technical problem of small target thermal leaks being missed by the traditional Otsu's method under a single-peak histogram, this invention provides a visual monitoring method for equipment energy efficiency status based on infrared thermal image feature analysis. The method includes: acquiring a single-frame infrared image of the equipment surface and converting it into a two-dimensional temperature matrix; discretizing the two-dimensional temperature matrix by temperature levels to generate an image temperature histogram; calculating the local thermal radiation convergence of each pixel within its local window neighborhood based on the two-dimensional temperature matrix; and calculating the local thermal radiation convergence, the absolute temperature of the pixel, and the global average temperature of the entire image. The maximum temperature value is used to construct the heat island salience potential energy for each pixel. Based on the heat island salience potential energy, the image temperature histogram is reconstructed by frequency, and the histogram frequency reconstruction factor corresponding to each discrete temperature level is calculated. Based on the histogram frequency reconstruction factor, a reconstruction probability distribution is generated. Based on the reconstruction probability distribution, the maximum inter-class variance algorithm is executed to determine the optimal threshold, and the optimal threshold is used to perform binarization segmentation of the two-dimensional temperature matrix to identify the damaged areas of the insulation layer. When the cumulative area of ​​the damaged areas of the insulation layer exceeds the preset threshold, the energy efficiency status of the equipment is determined, and the corresponding energy efficiency anomaly level is output.

[0007] This invention reconstructs the temperature histogram of the original infrared image by introducing local thermal radiation convergence and heat island salience potential energy, and performs maximum inter-class variance segmentation on this basis. This effectively overcomes the problem of missed detection of small targets caused by the single-peak long-tail distribution of the histogram in the early stage of insulation layer damage in traditional methods. At the same time, by comparing the damaged area with a preset threshold, the energy efficiency status is automatically graded, realizing end-to-end closed-loop monitoring from thermal image acquisition to operation and maintenance decision-making, which significantly improves the early identification capability and automation level of equipment energy efficiency anomalies.

[0008] Preferably, an image temperature histogram is generated by using an equal-interval quantization method on the two-dimensional temperature matrix.

[0009] Preferably, the formula for calculating the local thermal radiation convergence is: ; For pixels The degree of local thermal radiation convergence; The spatial coordinates of the center pixel currently being processed; For pixels Offset coordinates within the neighborhood; For local window neighborhood; For pixels Absolute physical temperature values ​​in a two-dimensional temperature matrix; For neighboring pixels Temperature value; This is a preset reference temperature gradient constant; is the Euclidean distance constant of the offset coordinates.

[0010] By defining a formula for calculating the local thermal radiation convergence, and comprehensively considering the weighted sum of squares of the temperature difference and spatial distance between the central pixel and its neighborhood, the concentration of local thermal anomalies can be effectively characterized, enhancing the spatial sensitivity to tiny thermal leaks, suppressing the interference of slow temperature changes in the background, and providing local feature inputs with high signal-to-noise ratio for subsequent potential energy construction.

[0011] Preferably, the local window neighborhood The preferred size The pixel matrix.

[0012] Preferably, the formula for calculating the potential energy of the heat island is: In the formula, For pixels The heat island effect highlights potential energy; For pixels The degree of local thermal radiation convergence; For pixels Absolute physical temperature values ​​in a two-dimensional temperature matrix; The global average temperature; The maximum extreme temperature; It is a natural exponential function.

[0013] The heat island highlights the potential energy by integrating the local thermal radiation convergence and the normalized offset of absolute temperature relative to the global temperature field. This makes the high temperature anomaly not only dependent on local contrast, but also modulated by its prominence in the overall temperature field, thereby strengthening the heat leakage signal that is truly meaningful for energy efficiency and further suppressing false hotspot interference.

[0014] Preferably, the formula for calculating the histogram frequency reconstruction factor is: In the formula, For the temperature level Histogram frequency reconstruction factor for all pixels; For a specific temperature level after discretization; A set of pixel coordinates; For pixels The heat island effect highlights potential energy; For the temperature level The sum of the potential energy of all pixels; For the temperature level The absolute total number of pixels of all pixels.

[0015] The histogram frequency reconstruction factor evaluates the overall anomaly weight of a temperature level by averaging the heat island potential energy of all pixels within the same temperature level and adding a constant bias, thus avoiding the influence of extreme values ​​of a single pixel and making the frequency reconstruction more statistically robust and physically reasonable.

[0016] Preferably, the formula for calculating the reconstructed probability distribution is: In the formula, For the temperature level The reconstruction probability distribution of all pixels; For the temperature level The total number of original pixels for all pixels; For the temperature level Histogram frequency reconstruction factor for all pixels; The total discrete temperature series; For the step size variable during traversal; The total number of original pixels at the m-th temperature level; is the histogram frequency reconstruction factor at the m-th temperature level.

[0017] The reconstructed probability distribution normalizes the original pixel frequency by multiplying it with the potential energy weight, which significantly improves the probability space of high temperature anomaly regions. This fundamentally changes the histogram shape, prompting the maximum inter-class variance algorithm to search for the optimal threshold that is closer to the actual damage boundary, thus greatly improving the segmentation accuracy.

[0018] Preferably, the method further includes marking pixels with values ​​above or equal to an optimal threshold as damaged areas and classifying pixels with values ​​below the optimal threshold as normal areas.

[0019] Preferably, the preset thresholds include a mild alarm threshold and a severe alarm threshold.

[0020] Preferably, the energy efficiency anomaly level includes: minor leakage and urgent need for maintenance.

[0021] The beneficial effects of this invention are as follows:

[0022] This invention reconstructs the temperature histogram of the original infrared image by introducing local thermal radiation convergence and heat island salience potential energy, and performs maximum inter-class variance segmentation on this basis. This effectively overcomes the problem of missed detection of small targets caused by the single-peak long-tail distribution of the histogram in the early stage of insulation layer damage in traditional methods. At the same time, by comparing the damaged area with a preset threshold, the energy efficiency status is automatically graded, realizing end-to-end closed-loop monitoring from thermal image acquisition to operation and maintenance decision-making, which significantly improves the early identification capability and automation level of equipment energy efficiency anomalies. Attached Figure Description

[0023] Figure 1 This is a flowchart illustrating a visual monitoring method for device energy efficiency status based on infrared thermal imaging feature analysis according to the present invention.

[0024] Figure 2 It is the original infrared thermal image and its temperature distribution map;

[0025] Figure 3 This is an illustration showing the enhanced potential energy characteristics of the urban heat island effect.

[0026] Figure 4 This is a comparison chart of the final processing results of binarized segmentation. Detailed Implementation

[0027] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0028] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0029] This invention discloses a visual monitoring method for equipment energy efficiency status based on infrared thermal imaging feature analysis, referring to... Figure 1 This includes steps S1 to S5:

[0030] S1: Acquire a single-frame infrared image of the device surface and convert it into a two-dimensional temperature matrix, while generating an image temperature histogram.

[0031] The infrared thermal image video stream of the device surface is acquired, and a single frame of infrared image for the current monitoring period is extracted. In this embodiment, an uncooled infrared thermal imager can be used to perform non-contact imaging of the device to be monitored. The preferred resolution of the infrared thermal imager is... The temperature measurement range covers the highest operating temperature that the equipment may reach and the ambient temperature of the workshop.

[0032] It should be noted that after acquiring a single frame of infrared image, each pixel in the image is parsed and mapped into a two-dimensional temperature matrix. This establishes the correspondence between spatial coordinates and absolute physical temperature. This step directly converts the thermal radiation intensity reflected in the infrared image into an absolute physical temperature distribution that can be used for numerical calculation. At the same time, the two-dimensional temperature matrix is ​​discretized using an equal-interval quantization method, that is, the temperature value range is divided into several equally wide temperature intervals, and the temperature value of each pixel is mapped to the corresponding temperature level index, thereby generating an image temperature histogram.

[0033] S2: Calculate the local thermal radiation convergence of each pixel based on the two-dimensional temperature matrix.

[0034] Early thermal leaks are heat energy that seeps through tiny cracks in the insulation layer, exhibiting a clear physical characteristic of radiating and diffusing in all directions. In image space, this manifests as a temperature spike with a high center and low periphery. Based on this physical phenomenon, this step calculates the degree to which the heat energy of a pixel target tends to converge towards the center within its neighborhood, using this as the standard for initial spatial anomaly measurement. The formula for calculating the local thermal radiation convergence degree is as follows:

[0035]

[0036] For pixels The local thermal radiation convergence degree represents the intensity of the heat source center. The higher the value, the more obvious the temperature change at that point is relative to the surrounding environment, which is more consistent with the early heat leak point's temperature peak characteristics of high temperature at the center and low temperature around the edges. The spatial coordinates of the center pixel currently being processed; For pixels In-neighbor offset coordinates, referring to the coordinates relative to the center pixel. The horizontal and vertical offsets can be used to access all neighboring pixels around the center pixel by traversing these offsets. For local window neighborhood, referring to The core calculation scope; For pixels Absolute physical temperature values ​​in a two-dimensional temperature matrix; For neighboring pixels Temperature value; The preset reference temperature gradient constant is used to normalize the temperature gradient into a dimensionless quantity. Because the background temperature gradient in infrared images is affected by equipment operating conditions, environmental disturbances, and sensor noise, its typical fluctuation range is relatively stable. If the setting is too small, it will be overly sensitive to normal temperature fluctuations, easily misjudging non-defect areas as hot leaks, leading to increased false alarm rates and wasted maintenance resources; if... Setting the threshold too high will weaken the response to weak but real early heat loss, causing small targets to be missed, delaying the early warning of energy efficiency degradation, and increasing energy consumption and safety risks. Therefore, the reasonable range for this parameter is 0.05 to 0.3; in this embodiment, it is set to 0.1. This setting is determined based on a comprehensive consideration of the thermal sensitivity of mainstream industrial infrared thermal imagers, the typical background gradient of food processing equipment surfaces, and the intensity of early crack heat leakage signals. This avoids false alarms caused by an overly strict threshold while ensuring effective capture of functional heat loss defects, thus achieving a balance between detection reliability and engineering practicality. In other embodiments, the implementer can flexibly adjust the threshold within the range according to the equipment operating temperature, insulation layer material, or monitoring accuracy requirements. The specific value to be taken; The Euclidean distance constant for offset coordinates refers to the geometric distance from neighboring pixels to the center pixel. It acts as a distance penalty factor. The farther a pixel is from the center, the smaller the contribution weight of its temperature difference to the convergence, thus ensuring that the calculation results can accurately pinpoint localized, subtle, and sudden thermal penetration.

[0037] Among them, when the temperature of the center pixel Temperature significantly higher than that of surrounding neighboring pixels At that time, the difference between the two increases; simultaneously, this difference is affected by the distance constant. The penalty calculation shows that pixels farther from the center have a smaller weight for their temperature difference contribution; after summing the squares of the distance penalty and logarithmic processing, the local thermal radiation convergence is... The change will increase nonlinearly; this mathematical change precisely corresponds to the physical phenomenon of sudden heat penetration in a small area in the actual production scenario, and highly reflects the physical thermal field convergence characteristics of strong outward diffusion from the early small area heat dissipation point.

[0038] It should be noted that the local window neighborhood The preferred size The pixel matrix; when the window size is smaller than When the calculation range is too narrow, it is easily affected by single noise points from the infrared sensor, leading to serious deviations in the convergence calculation; when the window size is larger than... At that time, the neighborhood may contain too many normal low-temperature background pixels, thus diluting the physical characteristics of the thermal field convergence; therefore, this embodiment adopts... The window accurately captures the thermal field diffusion trend of early microcracks at the most efficient computational cost; The set value can balance computational efficiency, noise suppression capability and thermal anomaly sensitivity, while avoiding amplification of noise interference due to an excessively small window. It effectively captures the thermal field diffusion trend caused by early microcracks in the insulation layer, achieving a good balance between detection reliability and engineering practicality. In other embodiments, the implementer can flexibly adjust the window size within the range according to factors such as equipment operating temperature, insulation material type and infrared imaging resolution.

[0039] S3: Construct heat islands for each pixel to highlight potential energy.

[0040] Due to the noise from infrared camera sensors and the slight pseudo-edge interference easily generated by moisture disturbance in the workshop, this embodiment introduces global temperature information based on local features. In the real heat dissipation scenario of high-temperature equipment, the actual heat leak point not only has a strong local abrupt change, but its absolute temperature level will also necessarily exceed the overall reference temperature environment of the equipment surface. Combining global environmental temperature parameters and local features, the true potential energy of the pixel as an isolated heat source can be accurately mapped. Therefore, based on the local heat radiation convergence, the absolute temperature of the pixel, the global average temperature of the entire image, and the maximum temperature value, the formula for calculating the heat island prominence potential energy of each pixel is as follows:

[0041]

[0042] In the formula, For pixels The heat island highlights potential energy, representing the energy level of the pixel as a real heat leakage source. It not only contains local contrast information, but also superimposed with the gain of global temperature weight. For pixels The local thermal radiation convergence reflects the actual physical temperature value of the current central pixel. For pixels Absolute physical temperature values ​​in a two-dimensional temperature matrix; The global average temperature is the average temperature of all pixels in the entire infrared thermal image. It serves as a baseline for measuring the overall thermal environment of the device surface and is used to distinguish between normal insulation backgrounds and abnormal heating areas. The maximum extreme temperature is the pixel value with the highest temperature in the entire image; it plays a normalization role in the calculation formula, defining the boundary of the hottest leak point that may exist in the image. It is a natural exponential function.

[0043] Among them, the center temperature An increase in temperature directly drives the internal variable of the exponential term to increase; when the temperature of a specific pixel increases... Far exceeding the global average temperature and toward the maximum extreme temperature As the heat island approaches, the exponential term expands rapidly, thus acting as a multiplicative gain coefficient to enhance its potential energy. The temperature increases dramatically; this mechanism can effectively suppress high-frequency interference noise in low-temperature backgrounds because although noise pixels may have a certain local gradient, their absolute temperature level is difficult to approach the maximum extreme temperature.

[0044] For example, Figure 2This is the original infrared thermal image and its temperature distribution map; this image shows a single-frame infrared image of the acquired equipment surface; the image as a whole presents a complex background noise distribution. Due to the extremely small early damage area of ​​the equipment's insulation layer, the contrast between the thermal leakage point and the background is extremely low, and its energy characteristics are completely masked in the huge normal low temperature background; at this time, the temperature histogram of the image presents a typical single-peak long-tail distribution. If it is directly segmented, it is very easy to cause the missed detection of small targets.

[0045] For example, Figure 3 This is an image showing the enhanced potential energy characteristics of the heat island. This image is an intermediate processing result after nonlinear potential energy transformation. By fusing local thermal radiation convergence with global temperature weights, a heat island highlighting potential energy model is constructed. It can be seen that random thermal noise and clutter in the background area are deeply purified and suppressed to extremely low levels, while the originally weak point-like and line-like thermal anomaly features are significantly stretched and amplified at a high magnification in the potential energy space. This step effectively enhances the saliency of the target signal, providing a high signal-to-noise ratio feature input for subsequent accurate extraction of the damaged boundary.

[0046] S4: Perform frequency reconstruction on the image temperature histogram and calculate the histogram frequency reconstruction factor corresponding to each discrete temperature level.

[0047] Traditional Otsu's method calculates inter-class differences based on the absolute number of pixels at each discrete temperature level. However, since early heat island pixels are few, direct statistics would cause them to be masked by a large number of normal background pixels in the global frequency distribution. This step transforms the salience potential at the spatial pixel level into dimensional reconstruction weight parameters for the image statistical histogram, artificially intervening to enhance the statistical importance of small targets within a specific temperature range. Based on the heat island salience potential, the image temperature histogram is reconstructed by frequency, and the formula for calculating the histogram frequency reconstruction factor corresponding to each discrete temperature level is as follows:

[0048]

[0049] In the formula, For the temperature level Histogram frequency reconstruction factor for all pixels; For a specific temperature level after discretization, a range of 0-255 is typically selected; It is a set of pixel coordinates, representing all temperature values ​​in the infrared image that fall exactly within the level. A set of pixel coordinates within a given range; For pixels The heat island effect highlights potential energy; For the temperature level The sum of the potential energy of all pixels; if the temperature level contains real hot spots, the molecular value of its pixels will increase significantly due to their high potential energy. For the temperature level The absolute total number of pixels of all pixels; 1 is a constant placed in the smoothing term of the denominator to prevent division by zero mathematical anomalies when the number of pixels is zero at a certain temperature level.

[0050] In the above calculation formula, the numerator is the sum of pixel potential energies within a specific temperature level, and the denominator is the total number of pixels within that level; when a specific temperature level The few pixels contained within have extremely high heat island potential accumulation values ​​due to their high-temperature characteristics, while the total number of pixels at this level... When the value is extremely small, according to the division operation logic, the histogram frequency reconstruction factor... This will result in extremely high numerical amplification; which allows high-temperature pixel groups that were originally statistically weak due to their small physical area but contain extremely high leakage characteristics to successfully obtain huge weight gain compensation.

[0051] S5: Generate the reconstruction probability distribution, and simultaneously execute the maximum inter-class variance algorithm to determine the optimal threshold. Use the optimal threshold to perform binarization segmentation on the two-dimensional temperature matrix to identify the damaged areas of the insulation layer. Based on the preset threshold, determine the energy efficiency status of the equipment and output the corresponding energy efficiency anomaly level.

[0052] Based on the frequency reconstruction factors obtained above, this step redistributes the occurrence probability of the original pixels, completely transforming the underlying probability distribution environment of the maximum inter-class variance algorithm, so that the enhanced weak targets can guide the search direction for the variance extremum; the formula for constructing the reconstruction probability distribution is as follows:

[0053]

[0054] In the formula, For the temperature level The reconstruction probability distribution of all pixels, which is at the temperature level. The probability values ​​are redistributed after manual intervention; they replace the original probabilities in the traditional histogram and become the underlying data for subsequent calculation of inter-class variance. For the temperature level The total number of original pixels of all pixels, the number of pixels that this temperature level occupies in the original infrared image; For the temperature level The histogram frequency reconstruction factor of all pixels is directly changed as a multiplication operator. The physical and statistical significance of; The total discrete temperature levels represent the total number of discrete levels into which the temperature range of the entire infrared image is divided. Since the raw temperature data output by the infrared thermal imager is usually highly accurate, if L is set too small, it will lead to coarse temperature quantization and reduce segmentation sensitivity; if L is set too large, it will introduce too many empty or sparse temperature intervals, affecting the robustness of the maximum inter-class variance algorithm. Therefore, the reasonable range of this parameter is [128, 512], and it is set to 256 in this embodiment. This setting avoids losing early weak thermal anomaly features due to overly coarse grading, and also prevents amplification of noise interference due to overly fine grading, thus achieving a good balance between detection accuracy and engineering practicality. In other embodiments, the implementer can flexibly adjust this parameter within the range according to the temperature measurement resolution of the infrared thermal imager, the surface temperature characteristics of the equipment, and real-time requirements. This is a loop index used to iterate over the step size variable and sum all temperature levels across the entire graph. The total number of original pixels at the m-th temperature level; For being in the first Histogram frequency reconstruction factor for each temperature level; It serves as a normalization function for the weighted total number of pixels after reconstruction at all levels of the entire image, ensuring that the sum of the reconstruction probabilities at all levels equals 1, thus satisfying the basic mathematical definition of probability distribution.

[0055] Among them, as it is at the temperature level Histogram frequency reconstruction factor of all pixels The significant increase in the high-temperature, small-target range leads to a corresponding increase in the numerator term dominated by its product; after normalization, the probability distribution is reconstructed. The values ​​rise significantly in the small high-temperature range; since the subsequent calculation of the maximum inter-class variance is highly dependent on the probability distribution of each level, the significant increase in the statistical proportion of hidden small targets will completely change the trend of the variance curve, causing the final calculated optimal threshold to adaptively move towards the true temperature boundary between small-area hot spots and large-area normal insulation layers.

[0056] It should be noted that, based on the reconstructed probability distribution, the maximum inter-class variance criterion is calculated, all possible temperature levels are traversed as candidate split points, the inter-class variance corresponding to each candidate point is calculated, and the temperature level that maximizes the inter-class variance is selected as the optimal threshold.

[0057] After obtaining the above-mentioned optimal threshold, the entire infrared thermal image two-dimensional temperature matrix is ​​binarized and precisely cut using the optimal threshold. Specifically, pixels with values ​​higher than or equal to the optimal threshold are marked as damaged areas, and pixels with values ​​lower than the optimal threshold are judged as normal areas, thereby non-destructively peeling out all early insulation layer damage contours.

[0058] For example, Figure 4 This is a comparison of the final processing results of binarized segmentation. This image shows the final segmentation result after performing the maximum inter-class variance algorithm based on the reconstructed probability distribution. By reconstructing the histogram by frequency, the problem of the threshold being dominated by a large area of ​​background in traditional methods is overcome, and the optimal threshold is adaptively moved towards the real damage boundary. The image successfully eliminates all background interference and retains the geometric shape of the early heat leakage location completely and clearly in a pure binarized form, realizing accurate identification of the damaged area of ​​the insulation layer.

[0059] Furthermore, the physical area of ​​the damaged area is accumulated and compared with preset thresholds in turn. The preset thresholds include a minor alarm threshold and a major alarm threshold. When the accumulated area is less than the minor alarm threshold, it is determined to be operating normally. When the accumulated area is between the minor and major alarm thresholds, it is determined to be a minor leak. When the accumulated area exceeds the major alarm threshold, it is determined to be in urgent need of maintenance.

[0060] Since the preset threshold is used to define whether the damage to the equipment's insulation layer has the reliability boundary to cause serious energy loss, its reasonable value range is [0.25, 0.4]. In this embodiment, the mild alarm threshold is set to 0.3 and the severe alarm threshold is set to 0.35. If the preset threshold is set too small, it may lead to the misjudgment of slight temperature fluctuations or uneven distribution within the normal construction tolerance as unqualified, increasing unnecessary rework costs. If the preset threshold is set too large, it may miss defects with obvious positional deviations or loose structures, weakening node constraints and insulation performance, resulting in Structural safety hazards exist; therefore, setting the mild alarm threshold and severe alarm threshold to 0.3 and 0.35 respectively is based on balancing detection sensitivity and actual construction tolerance. This avoids frequent false alarms caused by overly stringent judgments while ensuring timely detection of functional defects, thus achieving a precise balance between equipment operation safety and maintenance operability. In other embodiments, implementers can flexibly adjust the specific values ​​of the mild alarm threshold and severe alarm threshold within the recommended range of [0.25, 0.4] according to the importance of the component and production process requirements.

Claims

1. A method for visual monitoring of the energy efficiency state of equipment based on the analysis of infrared thermographic features, characterized in that, include: Acquire a single-frame infrared image of the device surface and convert it into a two-dimensional temperature matrix; The two-dimensional temperature matrix is ​​discretized by temperature levels to generate an image temperature histogram. According to the two-dimensional temperature matrix, a local heat radiation convergence degree of each pixel point in a local window neighborhood thereof is calculated , ; spatial coordinates of the center pixel point currently being processed; pixel point offset coordinates in the neighborhood; local window neighborhood; pixel point absolute physical temperature value in the two-dimensional temperature matrix; temperature value of the neighborhood pixel point temperature value of the neighborhood pixel point preset reference temperature gradient constant; Euclidean distance constant of offset coordinates; Based on the local thermal radiation convergence, the absolute temperature of the pixel, the global average temperature of the entire image, and the maximum temperature value, the heat island salience potential of each pixel is constructed. , In the formula, The global average temperature; The maximum extreme temperature; It is a natural exponential function; Based on the heat island highlighting potential energy, the image temperature histogram is reconstructed by frequency, and the histogram frequency reconstruction factor corresponding to each discrete temperature level is calculated. Based on the frequency reconstruction factor of the histogram, a reconstruction probability distribution is generated; based on the reconstruction probability distribution, the maximum inter-class variance algorithm is executed to determine the optimal threshold, and the optimal threshold is used to perform binarization segmentation of the two-dimensional temperature matrix to identify the damaged areas of the insulation layer. When the cumulative area of ​​the damaged insulation layer exceeds a preset threshold, the energy efficiency status of the equipment is determined and the corresponding energy efficiency anomaly level is output.

2. The method for visual monitoring of equipment energy efficiency status based on infrared thermal imaging feature analysis according to claim 1, characterized in that, An image temperature histogram is generated by using an equal-interval quantization method on a two-dimensional temperature matrix.

3. The method for visual monitoring of equipment energy efficiency status based on infrared thermal imaging feature analysis according to claim 1, characterized in that, The local window neighborhood The size is The pixel matrix.

4. The method for visual monitoring of equipment energy efficiency status based on infrared thermal imaging feature analysis according to claim 1, characterized in that, The formula for calculating the histogram frequency reconstruction factor is: In the formula, For the temperature level Histogram frequency reconstruction factor for all pixels; The temperature level is discretized. A set of pixel coordinates; For pixels The heat island effect highlights potential energy; For the temperature level The sum of the potential energy of all pixels; For the temperature level The absolute total number of pixels of all pixels.

5. The method for visual monitoring of equipment energy efficiency status based on infrared thermal imaging feature analysis according to claim 1, characterized in that, The formula for calculating the reconstructed probability distribution is: In the formula, For the temperature level The reconstruction probability distribution of all pixels; For the temperature level The total number of original pixels for all pixels; For the temperature level Histogram frequency reconstruction factor for all pixels; The total discrete temperature series; For the step size variable during traversal; The total number of original pixels at the m-th temperature level; is the histogram frequency reconstruction factor at the m-th temperature level.

6. The method for visual monitoring of equipment energy efficiency status based on infrared thermal imaging feature analysis according to claim 1, characterized in that, The method also includes marking pixels with values ​​above or equal to an optimal threshold as damaged areas and pixels with values ​​below the optimal threshold as normal areas.

7. The method for visual monitoring of equipment energy efficiency status based on infrared thermal imaging feature analysis according to claim 1, characterized in that, The preset thresholds include a mild alarm threshold and a severe alarm threshold.

8. The method for visual monitoring of equipment energy efficiency status based on infrared thermal imaging feature analysis according to claim 1, characterized in that, The energy efficiency anomaly levels include: minor leakage and urgent need for maintenance.

Images