An air source heat pump condensate water discharge abnormality monitoring method based on image enhancement

Abstract

This invention relates to the fields of machine vision and image processing technology, specifically to a method for monitoring abnormal condensate discharge from air-source heat pumps based on image enhancement. The method includes: acquiring a continuous video stream of the air-source heat pump's water tray area; calculating the temporal mean background image within the sampling period and eliminating background and illumination interference; combining the cumulative difference in the temporal domain with local texture features in the spatial domain to calculate the flow intensity index of each pixel within the region of interest to extract the dynamic features of the water flow; performing distance-weighted summation of the flow intensity index based on the physical distance from each pixel to the center of the drain outlet to construct a drainage blockage risk coefficient; and finally determining the condensate discharge status based on this coefficient. This invention effectively overcomes the metamerism problem by mining temporal micro-fluctuations and spatial texture features, significantly improving the robustness and accuracy of condensate discharge monitoring in complex agricultural and industrial environments.

Description

Technical Field

[0001] This invention relates to the fields of machine vision and image processing technology. More specifically, this invention relates to a method for monitoring abnormal condensate discharge from air-source heat pumps based on image enhancement. Background Technology

[0002] Air source heat pumps, as highly efficient and energy-saving temperature control devices, are widely used in greenhouses, agricultural product drying rooms, and industrial constant-temperature workshops. During the operation of the heat pump unit, the condensate produced by the evaporator needs to be collected and discharged through a drip tray. However, in agricultural and industrial environments, high concentrations of suspended dust, plant lint, and microbial spores easily settle in the tray, mixing with water to form viscous dirt or promote algae growth. This can easily cause physical sensors such as float switches and electrode sensors to malfunction due to jamming or corrosion. If the accumulated water cannot be drained and overflows, it can corrode the equipment base, damage electrical components, and even cause safety accidents such as short circuits and fires.

[0003] To address the issue of physical sensor failure, existing technologies have proposed monitoring solutions based on machine vision. These solutions primarily rely on color segmentation or edge detection of single-frame images to monitor water trays, identifying water accumulation status through non-contact methods and attempting to solve the problem of physical damage and functional loss of sensors caused by dirt buildup.

[0004] However, in real-world scenarios, the dark brown grime at the bottom of the water tray after long-term operation is extremely similar to the turbid water in color, brightness, and texture, presenting a "metachromatic" problem that prevents traditional gradient edge detection algorithms from extracting effective water level contours. Furthermore, environments such as agricultural greenhouses, with their high concentrations of fog and flickering supplemental lighting, easily cause grayscale threshold-based algorithms to fail and generate numerous false alarms, making accurate water level monitoring difficult. Summary of the Invention

[0005] The purpose of this invention is to propose an image enhancement-based method for monitoring abnormal condensate discharge from air-source heat pumps, in order to solve the problem that existing technologies cannot accurately monitor the condensate discharge status under low contrast and strong interference environments; to this end, this invention provides a solution in one aspect.

[0006] This invention provides a method for monitoring abnormal condensate discharge from air-source heat pumps based on image enhancement, comprising:

[0007] A continuous video stream of the air-source heat pump's water tray area is acquired, and the temporal mean background map within the sampling period is calculated. Based on the temporal mean background map, a normalized fluctuation map is calculated for each frame to eliminate background and illumination interference. Based on the normalized fluctuation map, the flow intensity index of each pixel within the region of interest is calculated by combining the cumulative difference in the temporal domain with local texture features in the spatial domain to extract the dynamic characteristics of the water flow. The flow intensity index is then weighted and summed based on the physical distance from each pixel to the center of the drain outlet, and a drainage blockage risk coefficient is constructed using a reference energy threshold to assess the smoothness of drainage. The drainage blockage risk coefficient is compared with a preset judgment threshold to determine the current condensate discharge status, and an alarm is triggered when an anomaly is detected.

[0008] Preferably, the step of calculating the normalized fluctuation map of each frame image based on the temporal mean background map includes: obtaining the original gray value of a pixel at the same coordinate position in each frame image within the sampling period; calculating the arithmetic mean of the original gray values ​​of the pixel in all frame images as the background gray value at that coordinate position in the temporal mean background map; and subtracting the background gray value from the original gray value at that coordinate position in each frame image to obtain the gray value at that coordinate position in the normalized fluctuation map of each frame image.

[0009] Through the above steps, static background components (such as the bottom of a dirty plate) in the image are filtered out. When a sudden change in illumination occurs, the mean background image will change accordingly, thereby offsetting the interference of global illumination fluctuations on local pixel analysis and achieving the effect of resisting illumination interference.

[0010] Preferably, the calculation of the flow intensity index for each pixel within the region of interest is performed as follows: In the formula, For pixels The flow intensity index, For the first The grayscale value of this pixel in the frame-normalized fluctuation map. The total number of frames in the captured video stream. This represents the standard deviation of the spatial texture within the neighborhood of that pixel. This is the normalized scaling constant.

[0011] By taking advantage of the large temporal fluctuations and continuous spatial texture of water flow, the weak water flow signal is powerfully extracted from the static background and random noise, thus solving the problem of feature extraction under low contrast.

[0012] Preferably, the spatial texture standard deviation is obtained by selecting intermediate frames of the video stream and calculating the standard deviation in coordinates. The standard deviation of pixel grayscale within the neighborhood window centered on the pixel.

[0013] Preferably, the normalized scaling constant is 100.

[0014] Preferably, the method for constructing the drainage blockage risk coefficient is as follows: In the formula, For the risk factor of drainage blockage, To monitor the region of interest, The flow intensity index, For pixels Euclidean distance from the center of the drain outlet The distance attenuation constant is As a reference energy threshold, For the natural constant An exponential function with base 0.

[0015] Through the above steps, a nonlinear risk assessment model was constructed. The negative exponential function ensures that the closer to the drain outlet and the stronger the flow, the lower the risk coefficient (approaching 0), and vice versa, the higher the risk coefficient (approaching 1), thus realizing a scientific assessment of the degree of blockage.

[0016] Preferably, the reference energy threshold is obtained by: calculating the sum of the distance-weighted flow intensity indices in the region of interest during the initial stage of equipment installation and when drainage is confirmed to be unobstructed, and using this sum as the reference energy threshold.

[0017] Preferably, the distance attenuation constant is 2 to 3 times the radius of the drain outlet.

[0018] Preferably, determining the current condensate discharge status includes: if the drainage blockage risk coefficient is less than 0.4, it is determined to be normal; if the drainage blockage risk coefficient is greater than or equal to 0.4 and less than 0.7, it is determined to be slowed flow; if the drainage blockage risk coefficient is greater than or equal to 0.7, it is determined to be severe blockage or water accumulation.

[0019] Through the above steps, quantitative and graded early warning is achieved. It is no longer a simple binary judgment, but can provide continuous monitoring of blockage trends, which facilitates preventive cleanup by operation and maintenance personnel.

[0020] Preferably, when acquiring continuous video streams, the total number of frames N in a single sampling is preset to 100 frames.

[0021] The beneficial effects of this invention are as follows: By constructing a flow intensity index, this invention identifies water flow using micro-fluctuation signals at the physical level, enabling accurate monitoring even when turbid water and dirty chassis are the same color. Simultaneously, this invention effectively suppresses abrupt changes in illumination and sensor thermal noise through time-series mean normalization and spatial texture weighting, greatly enhancing the system's anti-interference capability and achieving highly reliable monitoring in complex environments. Attached Figure Description

[0022] Figure 1 The flowchart illustrating the steps of the image enhancement-based air source heat pump condensate discharge anomaly monitoring method in this embodiment is shown in the schematic diagram.

[0023] Figure 2 The illustration shows a comparison of feature extraction responses under environmental interference according to an embodiment of the present invention, wherein Figure (a) is the original collected data, Figure (b) is the processing result of the prior art (frame difference method), and Figure (c) is the result of the flow intensity index calculated by the present invention.

[0024] Figure 3 A trend chart illustrating the risk monitoring of condensate drainage blockage according to an embodiment of the present invention is shown. Detailed Implementation

[0025] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0026] like Figure 1 As shown in this embodiment, an image enhancement-based method for monitoring abnormal condensate discharge from an air-source heat pump includes the following steps:

[0027] Step S1: Acquire a continuous video stream of the air source heat pump water receiving pan area, calculate the temporal mean background map within the sampling period, and calculate the normalized fluctuation map of each frame image based on the temporal mean background map to eliminate background and illumination interference.

[0028] Specifically, an industrial-grade miniature camera is installed above the water tray of the air source heat pump to capture a continuous video stream covering the drainage area. Assume the total number of frames in the captured video stream is... In this embodiment, N is set to 100 frames. To eliminate the interference of illumination fluctuations on the analysis, the temporal mean background map within the sampling period is first calculated, and then the normalized fluctuation map of each frame relative to the mean is calculated.

[0029] The time-domain mean background plot and the normalized fluctuation plot are obtained as follows:

[0030] ;

[0031] ;

[0032] In the formula, Pixels in the background image with time-domain mean grayscale value, It is the first Frame image at pixels The original grayscale value at that location, For the first Pixels in frame normalized fluctuation graph The grayscale value.

[0033] For example, suppose that for a specific pixel in an image ,exist The sequence of original grayscale values ​​acquired within a short period of the frame. for: .

[0034] First, the temporal mean background value of the pixel within the sampling period is calculated to be 100. Then, the normalized fluctuation value of each frame is calculated to obtain the pixel value. The sequence is .

[0035] Thus, by calculating the time-domain mean and performing normalization, a high-pass filter is essentially applied, filtering out the DC component (i.e., the stationary background) from the image, making the still water and dirty bottom of the plate visible. The values ​​are close to 0, and only the parts that change (such as water flow fluctuations) retain values, effectively eliminating the interference of global illumination changes on local pixel determination.

[0036] Step S2: Based on the normalized wave map, the flow intensity index of each pixel in the region of interest is calculated by combining the cumulative difference in the time domain and the local texture features in the spatial domain, so as to extract the dynamic features of the water flow.

[0037] Specifically, to distinguish between water flow fluctuations and sensor noise at low contrast, this invention constructs a flow intensity index (FII), which combines the temporal difference between adjacent frames and the local texture standard deviation in the spatial domain for signal amplification. The expression for the flow intensity index is:

[0038] ;

[0039] In the formula, For pixels The flow intensity index, For the first The grayscale value of this pixel in the frame-normalized fluctuation map. The total number of frames in the captured video stream. This represents the standard deviation of the spatial texture within the neighborhood of that pixel. This is the normalization scaling constant, usually taken as 100.

[0040] For example, continue using the data in S1, The sequence is .

[0041] First, calculate the absolute value of the difference between adjacent frames (the cumulative difference in the time domain), and the sum of the absolute values ​​of the differences between adjacent frames is 11. Then, the average fluctuation amplitude in the time domain is: .

[0042] Assuming the local spatial texture standard deviation is obtained by calculating the neighborhood of the intermediate frame. (This indicates that the texture around this point has some fluctuations, possibly resembling water ripples.) Set a normalization scale constant. Then the spatial texture gain term can be calculated as follows: Therefore, the flow intensity index of this pixel can be derived as: .

[0043] In comparison, if the point is a random noise point, although there are fluctuations in the time domain, its... If the gain term is usually very small (e.g., 2), then the gain term is only... The signal is hardly amplified; while in flowing water areas, due to the shimmering effect, If the signal is large, it will be significantly amplified.

[0044] Thus, by combining temporal fluctuations with spatial texture features, the multiplication effect significantly enhances the characteristic expression of water flow signals, thereby enabling the sensitive extraction of weak flow signals from static backgrounds and random noise.

[0045] Step S3: The flow intensity index is summed by distance weighting based on the physical distance from each pixel to the center of the drain outlet, and a drainage blockage risk coefficient is constructed by combining it with a reference energy threshold to assess the smoothness of drainage.

[0046] Specifically, considering that during normal drainage, the water flow is more rapid and the characteristics are more pronounced closer to the drain outlet, a simple summation is not sufficient; a distance weight must be introduced. Therefore, this invention constructs a drainage blockage risk coefficient. The expression for the drainage blockage risk coefficient is:

[0047] ;

[0048] In the formula, For the risk factor of drainage blockage, To monitor the region of interest, The flow intensity index, For pixels Euclidean distance (in pixels) from the center of the drain outlet. The distance attenuation constant is As a reference energy threshold, For the natural constant It is an exponential function with base 0. The distance attenuation constant is 2 to 3 times the radius of the drain outlet.

[0049] The reference energy threshold is obtained as follows: during the initial installation of the equipment and after confirming that the drainage is unobstructed, the sum of the distance-weighted flow intensity indices in the region of interest is calculated, and this sum is used as the reference energy threshold.

[0050] For example, assume the physical center coordinates of the drain outlet are... Examine a pixel within the region of interest. lie in Then its Euclidean distance from the center of the drain outlet Pixel. Assume a distance attenuation constant. Then the distance weight of that point is .

[0051] Assuming that the point and its surrounding area are under normal drainage conditions and the water flow is rapid, its High. Assuming the calculated result is 100, the contribution of this point to the molecule (weighted total energy) is: .

[0052] Assuming the entire region of interest The above calculations are performed on all pixels within the formula and summed to obtain a total sum (i.e., the numerator in the formula), which is 800.

[0053] Assume the reference energy threshold measured during system initialization calibration. (representing the total energy under ideal unobstructed conditions), the calculated drainage blockage risk coefficient is 0.45.

[0054] If it is completely blocked, the water flow will be still. If the entire graph approaches 0 and the numerator approaches 0, then... (Represents high risk); if drainage is very smooth, the molecules are close to ,but (This represents low risk).

[0055] Thus, by introducing a distance-weighted and negative exponential function model, a physically consistent risk assessment index was constructed: the closer the value is to 0, the smoother the drainage (higher energy); the closer the value is to 1, the higher the risk of blockage (lower energy), thus achieving a scientific quantitative assessment of the degree of blockage.

[0056] Step S4: Compare the drainage blockage risk coefficient with a preset judgment threshold to determine the current condensate discharge status, and trigger an alarm when the status is determined to be abnormal.

[0057] Specifically, based on the calculated Values ​​are used for hierarchical judgment and decision output:

[0058] like : Determined as normal; if If the flow rate is determined to be slowed (partial blockage), the system sends a "maintenance recommended" signal; if If the system is determined to be severely blocked or flooded, it will immediately trigger a shutdown protection and send an alarm.

[0059] In the previous example, the calculated It is 0.45, falling into The system will therefore determine that the flow rate has slowed down and prompt the operations and maintenance personnel to pay attention to it.

[0060] In this way, by comparing the quantified risk coefficient with the graded threshold, the system has realized the transformation from simple binary judgment to continuous trend monitoring, which can detect early signs of blockage in a timely manner and prevent accidents from developing into overflow accidents.

[0061] like Figure 2 As shown, this invention demonstrates its advantages in anti-interference. Figure 2 The diagram contains three sub-graphs, with the horizontal axis representing the time frame sequence. Figure (a) shows the drastic jump in grayscale signal during the illumination change in frames 40-60; Figure (b) shows the large spurious peak (false alarm) generated at the illumination change moment; Figure (c) shows the calculations of this invention. It remained stable in the light interference section, and the difference between the "flowing" signal (high value) and the "stationary" signal (low value) was significant, with extremely high feature discrimination.

[0062] like Figure 3 As shown, it demonstrates The index shows an S-shaped upward trend as drainage conditions worsen, clearly dividing the "normal drainage zone" (low risk), "flow rate slowdown zone" (medium risk), and "confirmed blockage zone" (high risk). The red alarm threshold line (0.7) intuitively demonstrates the judgment logic and verifies the effectiveness of the quantitative indicator.

[0063] In the description of this specification, "multiple" means at least two, such as two, three or more, etc., unless otherwise expressly and specifically defined.

[0064] While various embodiments of the invention have been shown and described in this specification, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Many modifications, alterations, and alternatives will occur to those skilled in the art without departing from the spirit and essence of the invention.

Claims

1. A method for monitoring abnormal condensate discharge from an air-source heat pump based on image enhancement, characterized in that, include: A continuous video stream of the water receiving pan area of ​​the air source heat pump is collected, the temporal mean background map within the sampling period is calculated, and the normalized fluctuation map of each frame image is calculated based on the temporal mean background map to eliminate background and illumination interference. Based on the normalized fluctuation map, the flow intensity index of each pixel within the region of interest is calculated by combining the cumulative difference in the temporal domain and the local texture features in the spatial domain. ; In the formula, For pixels The flow intensity index, For the first The grayscale value of this pixel in the frame-normalized fluctuation map. The total number of frames in the captured video stream. This represents the standard deviation of the spatial texture within the neighborhood of that pixel. This is a normalized scale constant used to extract the dynamic characteristics of water flow; The flow intensity index is summed using a distance-weighted method based on the physical distance of each pixel to the center of the drain outlet, and a drain blockage risk coefficient is constructed by combining this with a reference energy threshold. ; In the formula, For the risk factor of drainage blockage, To monitor the region of interest, The flow intensity index, For pixels Euclidean distance from the center of the drain outlet The distance attenuation constant is As a reference energy threshold, For the natural constant An exponential function with base 0.5 is used to assess the smoothness of drainage. The drainage blockage risk coefficient is compared with a preset judgment threshold to determine the current condensate discharge status, and an alarm is triggered when the status is determined to be abnormal.

2. The method for monitoring abnormal condensate discharge from an air-source heat pump based on image enhancement according to claim 1, characterized in that, The step of calculating the normalized fluctuation map of each frame of the image based on the time-domain mean background map includes: The original gray value of the pixel at the same coordinate position in each frame image within the sampling period is obtained, and the arithmetic mean of the original gray value of the pixel in all frame images is calculated as the background gray value at that coordinate position in the time domain mean background image. Subtract the background gray value from the original gray value at that coordinate position in each frame image to obtain the gray value of that coordinate position in the normalized fluctuation map of each frame image.

3. The method for monitoring abnormal condensate discharge from an air-source heat pump based on image enhancement according to claim 1, characterized in that, The spatial texture standard deviation is obtained by selecting an intermediate frame of the video stream and calculating it using coordinates. The standard deviation of pixel grayscale within the neighborhood window centered on the pixel.

4. The method for monitoring abnormal condensate discharge from an air-source heat pump based on image enhancement according to claim 1, characterized in that, The normalized scaling constant is set to 100.

5. The method for monitoring abnormal condensate discharge from an air-source heat pump based on image enhancement according to claim 1, characterized in that, The reference energy threshold is obtained as follows: during the initial installation of the equipment and after confirming that the drainage is unobstructed, the sum of the distance-weighted flow intensity indices in the region of interest is calculated, and this sum is used as the reference energy threshold.

6. The method for monitoring abnormal condensate discharge from an air-source heat pump based on image enhancement according to claim 1, characterized in that, The distance attenuation constant is 2 to 3 times the radius of the drain outlet.

7. The method for monitoring abnormal condensate discharge from an air-source heat pump based on image enhancement according to claim 1, characterized in that, Determining the current condensate discharge status includes: If the risk factor for drainage blockage is less than 0.4, it is considered normal. If the drainage blockage risk coefficient is greater than or equal to 0.4 and less than 0.7, it is judged as a slowdown in flow rate; If the risk coefficient of drainage blockage is greater than or equal to 0.7, it is judged as a serious blockage or water accumulation.

8. The method for monitoring abnormal condensate discharge from an air-source heat pump based on image enhancement according to claim 1, characterized in that, When acquiring continuous video streams, the preset total number of frames N for a single sampling is 100 frames.

Images