Concrete external wall maintenance state self-adaptive spraying control method based on depth vision analysis

Abstract

The present application relates to the technical field of concrete curing, and specifically discloses a concrete outer wall curing state self-adaptive spraying control method based on deep visual analysis, which collects wall surface time sequence images and environmental data; decomposes the images into space-time humidity characteristic quantities and combines the environmental data to construct a diffusion coefficient space-time field, and inversely calculates an evaporation rate field conforming to the water continuity equation; fuses the evaporation rate field and the current humidity field to generate a drying risk distribution map; maps the wall surface to each spraying subzone according to the drying risk map and calculates a pulse duty cycle sequence; after executing the subzone spraying, feeds back the images and environmental data collected after spraying, extracts abnormal pixels by comparing the actual and theoretical humidification amounts, and constructs a correction field to iteratively update the evaporation rate field; the present application fuses physical laws and visual observation, and closes the loop to correct the evaporation model through the spraying effect, can long-term maintain the matching between the spraying decision and the actual demand of the wall surface, and saves curing water.

Description

Technical Field

[0001] This invention relates to the field of concrete curing technology, specifically to an adaptive spraying control method for the curing status of concrete exterior walls based on depth visual analysis. Background Technology

[0002] After concrete is poured, it needs sufficient moisture retention and curing to prevent early cracking and ensure final strength. Exterior wall curing is greatly affected by environmental factors such as sunlight, wind speed, and temperature. Traditional methods often involve timed spraying or manual watering with handheld hoses. Timed spraying cannot adjust the water volume according to the actual dryness of the wall surface, easily leading to water waste and potential under-curing in some areas. In recent years, some solutions have utilized cameras to capture images of the wall surface, identified damp areas through image processing, and then controlled sprinklers for targeted spraying.

[0003] Existing spray control methods based on image recognition in concrete exterior wall curing rely solely on the appearance features of images to determine the surface wetness and dryness. They neither incorporate the physical laws of moisture diffusion within the concrete into the evaporation rate prediction process nor utilize the measured humidification after spraying to perform closed-loop correction of the evaporation model. As a result, the evaporation rate prediction model gradually becomes ineffective after changes in environmental conditions (such as wind direction and solar radiation) or changes in wall conditions (such as crack development and surface crusting), and cannot maintain a long-term match between spray decisions and actual needs. Summary of the Invention

[0004] The purpose of this invention is to provide an adaptive spraying control method for the curing status of concrete exterior walls based on depth visual analysis, so as to solve the problems mentioned above.

[0005] The objective of this invention can be achieved through the following technical solutions: An adaptive spraying control method for the curing status of concrete exterior walls based on depth visual analysis includes the following steps: S1, acquire time-series image sequences of the concrete exterior wall surface, and simultaneously acquire ambient temperature, humidity, wind speed and solar radiation data; S2, decompose the time-series image sequence into spatiotemporal humidity characterization quantities by pixels, and construct the spatiotemporal field of diffusion coefficient based on environmental temperature and humidity, wind speed and solar radiation data, and then use the spatiotemporal field of spatiotemporal humidity characterization quantities and diffusion coefficient to invert the evaporation rate field that conforms to the water continuity equation. S3, integrates the evaporation rate field with the humidity distribution field at the current moment to generate a dry risk distribution map that reflects the urgency of water shortage at each spatial location; S4. Based on the dryness risk distribution map, map each spatial location to the coverage area of ​​the preset spray array, and calculate the duty cycle sequence within the pulsating spray cycle for each area. S5 executes zoned spraying according to the duty cycle sequence, and simultaneously feeds back the newly acquired time-series image sequence and environmental data after spraying to S2 to iteratively update the evaporation rate field.

[0006] As a further aspect of the present invention: the inversion of the evaporation rate field conforming to the water continuity equation specifically includes: The humidity attenuation gradient of corresponding pixels in three consecutive frames is extracted from the spatiotemporal humidity characterization to form the observation residual sequence; Substituting the observed residual sequence and the spatiotemporal field of the diffusion coefficient into the spatial discretization form of the moisture continuity equation, and minimizing the sum of squared residuals by adjusting the implicit evaporation term, the initial value of the evaporation rate of each pixel is obtained. The initial evaporation rate is anisotropically smoothed according to the direction of solar radiation, so that the evaporation rate on the sunlit side is higher than that on the shaded side, thus generating the final evaporation rate field.

[0007] As a further aspect of the present invention: obtaining the initial value of the evaporation rate of each pixel specifically includes: The spatial discretization form of the moisture continuity equation is decomposed into two one-dimensional iterative operators along the horizontal and vertical directions of the wall. Traverse each pixel in the order of horizontal first and then vertical, and calculate the correction amount of the implicit evaporation term based on the weighted average of the observation residual of the pixel and the diffusion coefficients in the four directions of its neighborhood. Repeat the horizontal and vertical scans until the change in the sum of squared residuals between two adjacent scans is lower than a preset threshold. Then, output the implicit evaporation term of each pixel at this point as the initial value of the evaporation rate.

[0008] As a further aspect of the present invention: the generation of a drying risk distribution map reflecting the degree of water shortage urgency at various spatial locations specifically includes: Based on the current humidity distribution field, pixels with humidity below a preset humidity threshold are marked as areas to be evaluated. Within the region to be evaluated, the local gradient direction of the evaporation rate field and the humidity distribution field is multiplied by a dot to obtain the projection component of the moisture loss vector in the direction of the fastest decrease in humidity. The projected component is multiplied by the time accumulation of the evaporation rate field, and after normalization, the drying risk value of each pixel is obtained and filled into the corresponding position to form a drying risk distribution map.

[0009] As a further aspect of the present invention: the process of obtaining the projection components is as follows: A square neighborhood is selected centered on each pixel in the region to be evaluated. The central difference of the humidity value in the neighborhood is calculated to obtain the orientation angle of the local gradient direction. Based on the orientation angle, the evaporation rate field at the pixel is decomposed into components parallel to the local gradient direction and orthogonal components. The values ​​of the parallel components are extracted and multiplied by the magnitude of the local gradient direction to obtain the projection components.

[0010] As a further aspect of the present invention: the output process of the duty cycle sequence is as follows: The dry risk distribution map is integrated according to each coverage zone of the preset spray array to obtain the cumulative risk value of each zone; After sorting the cumulative risk values ​​of each partition, they are divided into three risk levels: high, medium, and low. The high-risk level is assigned a pre-pulse group within the pulsation cycle, the medium-risk level is assigned a mid-pulse group, and the low-risk level is assigned a post-pulse group. Within each pulse group, a high-low level alternating timing sequence is generated with the pulse width proportional to the cumulative risk value of the partition, and the result is obtained by splicing the sequences together to obtain the duty cycle sequence of the partition.

[0011] As a further aspect of the present invention: obtaining the cumulative risk value for each partition specifically includes: Each coverage area is evenly divided into multiple sub-regions, and the dryness risk value at the center of each sub-region is extracted. The aridity risk values ​​of each center are weighted and summed using the shortest distance from the center of the sub-region to the boundary of the partition as the weight. Divide the accumulated value by the total area of ​​the sub-region to obtain the cumulative risk value of the partition.

[0012] As a further aspect of the present invention: the iterative update of the evaporation rate field specifically includes: The actual humidification at each location is obtained by subtracting the newly acquired time-series image sequence after spraying from the current humidity distribution field by pixels. The actual humidification amount is compared with the theoretical humidification amount calculated based on the duty cycle sequence, and abnormal pixels whose deviation from the theoretical humidification amount exceeds the preset tolerance are extracted. A correction field is constructed by spreading outward from the abnormal pixel as the center. The correction field is then superimposed on the evaporation rate field output by S2 to complete the iterative update of the evaporation rate field.

[0013] As a further aspect of the present invention: the extraction process of the abnormal pixels is as follows: Calculate the absolute value of the difference between the actual humidification amount and the theoretical humidification amount for each pixel, and use it as the original deviation; Using the global median of the original deviation as a benchmark, calculate the local deviation of each pixel from the median of its eight neighboring pixels; Pixels whose local deviation exceeds 1.3 times the global median of the original deviation and is also greater than the mean of the neighborhood deviation are marked as abnormal pixels.

[0014] The beneficial effects of this invention are: (1) By combining visual observation of the changes in humidity on the concrete surface with the physical equation of water diffusion, an evaporation rate field that is more in line with the actual evaporation law can be derived. This allows for zoned pulsed spraying based on the actual water shortage urgency of each area, avoiding local over-wetting or uneven drying caused by traditional timed spraying, effectively saving water resources and reducing maintenance costs.

[0015] (2) By using the measured humidification data collected after spraying as feedback, abnormal areas that deviate significantly from theoretical expectations are automatically identified, and a correction field is constructed with the abnormal point as the center to iteratively update the evaporation rate field, so that the control system can continuously adapt to the actual changes in the water loss characteristics of the wall surface (such as cracks, surface crusting or sudden weather changes), and maintain the long-term accuracy of spraying decisions without manual recalibration. Attached Figure Description

[0016] The invention will now be further described with reference to the accompanying drawings.

[0017] Figure 1 This is a flowchart of the method of the present invention; Figure 2 This is a flowchart illustrating the process of reversing the evaporation rate field that conforms to the water continuity equation in this invention. Figure 3 This is a flowchart of the process for generating the drying risk distribution map in this invention. Detailed Implementation

[0018] 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 embodiments of the present invention, and not all embodiments. 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.

[0019] Please see Figure 1 As shown, this invention is an adaptive spraying control method for the curing status of concrete exterior walls based on depth visual analysis, comprising the following steps: S1, acquire time-series image sequences of the concrete exterior wall surface, and simultaneously acquire ambient temperature, humidity, wind speed and solar radiation data; S2, decompose the time-series image sequence into spatiotemporal humidity characterization quantities by pixels, and construct the spatiotemporal field of diffusion coefficient based on environmental temperature and humidity, wind speed and solar radiation data, and then use the spatiotemporal field of spatiotemporal humidity characterization quantities and diffusion coefficient to invert the evaporation rate field that conforms to the water continuity equation. S3, integrates the evaporation rate field with the humidity distribution field at the current moment to generate a dry risk distribution map that reflects the urgency of water shortage at each spatial location; S4. Based on the dryness risk distribution map, map each spatial location to the coverage area of ​​the preset spray array, and calculate the duty cycle sequence within the pulsating spray cycle for each area. S5 executes zoned spraying according to the duty cycle sequence, and simultaneously feeds back the newly acquired time-series image sequence and environmental data after spraying to S2 to iteratively update the evaporation rate field.

[0020] In S1, a time-series image sequence of the concrete exterior wall surface is acquired, and environmental temperature, humidity, wind speed, and solar radiation data are acquired simultaneously, specifically including: An industrial-grade visible light camera, fixedly installed on a construction hoist platform or scaffold crossbar, continuously acquires a sequence of temporal images of the concrete exterior wall surface at a sampling frequency of 1 frame per minute. The camera lens plane is kept parallel to the wall surface, and a dust cover and a light shield are installed at the front of the lens to avoid direct sunlight and dust interference. Temperature and humidity sensors, a three-cup anemometer, and a total radiation meter are deployed at the same horizontal position as the camera. The temperature and humidity sensors use capacitive sensitive elements and are placed inside a radiation-proof ventilation cover. The anemometer is installed 50 cm away from the wall with the windward side unobstructed. The total radiation meter is placed horizontally above the wall to avoid shadow obstruction. The aforementioned sensors synchronously collect ambient temperature, relative humidity, wind speed, and solar radiation intensity values ​​at the same sampling frequency as the camera. The collected time-series image sequences are transmitted to an industrial control computer via coaxial cable. The ambient temperature, humidity, wind speed, and solar radiation data are aggregated to the same industrial control computer via RS-485 bus using the Modbus protocol. All data are uniformly stored as structured records according to the timestamp of the collection time. Each record contains a wall image and its corresponding ambient temperature, relative humidity, wind speed, and solar radiation intensity values.

[0021] Please see Figure 2 As shown, in S2, the time-series image sequence is decomposed into spatiotemporal humidity characteristics by pixel, and a spatiotemporal field of diffusion coefficient is constructed based on environmental temperature, humidity, wind speed, and solar radiation data. Then, the spatiotemporal humidity characteristics and the spatiotemporal field of diffusion coefficient are used to inversely derive the evaporation rate field that conforms to the moisture continuity equation. Specifically, this includes: The time-series image sequence is decomposed into spatiotemporal humidity representations by pixel: The industrial control computer reads the stored continuous time-series image sequence, performs grayscale processing on each frame, and converts the grayscale value of each pixel into a relative humidity value. The relative humidity value ranges from 0 to 1, where 0 represents complete dryness and 1 represents saturated humidity. The conversion method uses a pre-calibrated grayscale-humidity mapping curve, which is obtained by fitting images of the same batch of concrete test blocks with different moisture contents taken under laboratory conditions. For each pixel in each frame, its grayscale value is substituted into the mapping curve to obtain the corresponding relative humidity value, thus constructing a spatiotemporal humidity representation indexed by the two-dimensional coordinates of the wall surface and the timestamp.

[0022] A spatiotemporal field for the diffusion coefficient is constructed based on the ambient temperature, relative humidity, wind speed, and solar radiation intensity values ​​collected by S1. The diffusion coefficient characterizes the rate of moisture diffusion within concrete, and its value is affected by temperature and humidity. For each pixel location and its corresponding timestamp, a baseline diffusion coefficient is first calculated based on the ambient temperature and relative humidity values ​​at that moment. The calculation process is as follows: the baseline diffusion coefficient equals a basic constant multiplied by a power function with the temperature value as the exponent, and then multiplied by a linear factor related to the relative humidity value. The basic constant is taken as 0.002, the base of the exponent is taken as 1.07, the temperature value is substituted in degrees Celsius, and the linear factor is equal to 1 plus 0.5 times the relative humidity value.

[0023] The baseline diffusion coefficient is corrected based on wind speed and solar radiation intensity values: if the wind speed is greater than 2 meters per second, the diffusion coefficient increases by 5% for every 1 meter per second increase in wind speed; if the solar radiation intensity is greater than 200 watts per square meter, the diffusion coefficient increases by 8% for every 100 watts per square meter increase. Finally, the diffusion coefficient value for each pixel at each moment is obtained, and these values ​​are arranged according to spatial coordinates and time to form the spatiotemporal field of the diffusion coefficient.

[0024] The humidity attenuation gradients of corresponding pixels in three consecutive frames are extracted from the spatiotemporal humidity representation to form an observation residual sequence. For each pixel location, the relative humidity values ​​corresponding to three temporally adjacent frames (denoted as times T-1, T, and T+1) are taken. The first gradient is calculated as the humidity at time T minus the humidity at time T-1, and the second gradient is calculated as the humidity at time T+1 minus the humidity at time T. The average of these two gradient values ​​is then used as the humidity attenuation gradient at the center time T. The humidity attenuation gradients of all pixels at all times are arranged in chronological order to form the observation residual sequence.

[0025] The observed residual sequence and the spatiotemporal field of the diffusion coefficient are simultaneously substituted into the spatial discretization form of the moisture continuity equation. By adjusting the implicit evaporation term to minimize the sum of squared residuals, the initial value of the evaporation rate for each pixel is obtained. The moisture continuity equation is spatially discretized into a grid of horizontal and vertical lines on the wall, with each grid corresponding to a pixel. The spatial discretization form can be expressed in words as: the rate of change of humidity at the current moment equals the diffusion coefficient multiplied by the sum of the second derivatives of humidity in the horizontal and vertical directions, minus the evaporation rate. An iterative method is used to solve for the evaporation rate. The spatial discrete form of the moisture continuity equation is decomposed into two one-dimensional iterative operators along the wall, one horizontally and one vertically, that is, first only considering horizontal diffusion, and then only considering vertical diffusion.

[0026] The process iterates through each pixel in a horizontal-then-vertical order. For the current pixel, a correction to the implicit evaporation term is calculated based on its observation residual (i.e., the actual humidity change rate measured from the time-series image) and the weighted average of the diffusion coefficients in its four neighboring directions (left, right, top, and bottom). The weighted average is calculated by summing the diffusion coefficients in the four directions and dividing by 4. The correction is calculated as follows: the correction equals the observation residual minus the weighted average of the diffusion coefficients multiplied by the sum of the humidity differences between the current pixel and its neighboring pixels, then divided by a relaxation factor of 0.8. This correction is added to the current implicit evaporation term value, resulting in the updated implicit evaporation term for that pixel. A complete horizontal and vertical traversal is called a scan.

[0027] Repeat the horizontal and vertical scans described above. After each scan, calculate the sum of squared residuals for all pixels, which is the sum of the squares of the difference between the observed residual of each pixel and the theoretical rate of change calculated based on the current implicit evaporation term and diffusion coefficient. Stop the iteration when the change in the sum of squared residuals between two adjacent scans falls below a preset threshold. This threshold is set to be less than one-thousandth of the absolute value of the difference between the sums of squared residuals between two adjacent scans. When the iteration stops, the current implicit evaporation term value for each pixel is the initial value of the evaporation rate.

[0028] Anisotropic smoothing of the initial evaporation rate value is performed based on the direction of solar radiation to generate the final evaporation rate field. The solar azimuth is determined based on the trend of collected solar radiation intensity values. Specifically, during the period of fastest increase in solar radiation intensity within a continuous one-hour period, the direction of maximum radiation received by the total radiation meter during this period is recorded as the solar radiation direction. This direction is divided into four quadrants (east, south, west, and north) with the wall normal as the reference. Then, the initial evaporation rate value of each pixel is corrected according to its position relative to the solar radiation direction: if the pixel is on the sunlit side (facing the solar radiation direction), its initial evaporation rate value is multiplied by an enhancement factor of 1.3; if it is on the shaded side, it is multiplied by a reduction factor of 0.8. After correction, the average value of a 3x3 neighborhood centered on each pixel is taken for smoothing to obtain the final evaporation rate value for that pixel. The final evaporation rate values ​​of all pixels are arranged in spatial coordinates to form the evaporation rate field.

[0029] Please see Figure 3 As shown, in S3, the evaporation rate field is fused with the humidity distribution field at the current moment to generate a dryness risk distribution map reflecting the urgency of water shortage at each spatial location, specifically including: Based on the current humidity distribution field, pixels with humidity levels below a preset humidity threshold are marked as evaluation areas. The humidity distribution field consists of the relative humidity values ​​of each pixel at the current moment, output by S2, ranging from 0 to 1. The preset humidity threshold is 0.6; that is, when the relative humidity value of a pixel is less than 0.6, the location is considered insufficiently moist and must be included in the evaluation area; if the relative humidity value is greater than or equal to 0.6, the location is considered sufficiently moist and is not included in subsequent calculations. For each pixel within the evaluation area, a subsequent step will calculate its drying risk value.

[0030] Within the region to be evaluated, the local gradient directions of the evaporation rate field and the humidity distribution field are multiplied by a dot product to obtain the projection component of the moisture loss vector in the direction of the fastest humidity decrease. Specifically, a square neighborhood with a side length of 5 pixels is selected centered on each pixel within the region to be evaluated (i.e., the neighborhood includes the center pixel and two pixels above, below, to the left, and to the right). The central difference of the humidity values ​​within the neighborhood is calculated to obtain the local gradient direction.

[0031] The center difference is calculated as follows: For a center pixel located in row i and column j, its humidity value is H(i,j). The horizontal gradient component Gx is equal to the humidity value in row i+1 and column j minus the humidity value in row i-1 and column j-1, then divided by twice the pixel spacing (pixel spacing is 1 pixel unit). The vertical gradient component Gy is equal to the humidity value in row i+1 and column j minus the humidity value in row i-1 and column j, then divided by twice the pixel spacing. This yields the gradient vector (Gx, Gy). The direction angle θ of this vector is calculated using the arctangent function, i.e., θ is equal to the arctangent of Gy divided by Gx, with the angle ranging from negative π to π. If both Gx and Gy are 0, the direction angle is 0. The direction angle θ is the opposite direction of the fastest humidity decrease (because the gradient points to the direction of the fastest humidity increase, and the direction of the fastest humidity decrease is its opposite direction). In practical use, the gradient direction angle is directly taken as the reference direction. Next, based on the orientation angle, the evaporation rate field at the pixel is decomposed into components parallel to the local gradient direction and orthogonal components. Let the value of the evaporation rate field at the pixel be E. Then, the parallel component is equal to E multiplied by the cosine of the orientation angle θ, and the orthogonal component is equal to E multiplied by the sine of the orientation angle θ. After extracting the parallel component, it is multiplied by the magnitude of the local gradient direction to obtain the projection component. The magnitude M of the local gradient direction is calculated using the following formula: Where Gx and Gy are the previously calculated transverse and longitudinal gradient components, respectively. The projected component P is equal to the parallel component multiplied by M. This projected component reflects the weight of the rate of water loss along the direction of the fastest descent of the humidity gradient at the current evaporation rate.

[0032] The drying risk value for each pixel is obtained by multiplying the projected component by the time accumulation of the evaporation rate field and then normalizing the result. The time accumulation refers to the elapsed time, in seconds, from the start of the current spray cycle to the current moment. Let T (seconds) be the interval between the end of the last spray and the current moment; then the time accumulation is T. For each pixel within the evaluation area, the intermediate product value... It equals the projected component P multiplied by the time accumulation T. Then, this product value is normalized: the normalization of all pixels within the entire evaluation region is calculated. Find the maximum value among them. and minimum value .like equal If the drying risk value is 0.5, then the drying risk value R of all pixels is uniformly set to 0.5; otherwise, the drying risk value R of each pixel is calculated according to the following formula: ;in This is the median product value of that pixel. The smallest in the area to be evaluated value, For the largest After normalization, the drying risk value R ranges from 0 to 1, with a higher value indicating a greater urgency of water shortage at that location. The drying risk values ​​of all pixels are filled into their corresponding spatial locations to form a drying risk distribution map. For pixels outside the area to be evaluated, their drying risk value is directly assigned to 0. The drying risk distribution map is the output of S3 for use in subsequent steps. In the above calculations, all division operations retain three significant decimal places. If a denominator is zero during the calculation, it is handled according to the aforementioned conditions. The pixel spacing is considered as 1 when calculating the gradient; the conversion of the actual physical size is handled uniformly during subsequent nozzle mapping.

[0033] In S4, based on the dryness risk distribution map, each spatial location is mapped to a pre-defined spray array coverage zone, and the duty cycle sequence within the pulsating spray cycle is calculated for each zone, specifically including: Based on the drying risk distribution map, each spatial location is mapped to a coverage area of ​​a preset spray array, and the duty cycle sequence within the pulsating spray cycle is calculated for each area. The preset spray array consists of several nozzles arranged in a rectangular grid, with each nozzle covering one rectangular area on the wall surface. All areas are non-overlapping and completely cover the entire exterior wall surface. Each area is rectangular in shape, with its side length consistent with the spacing between adjacent nozzles. The drying risk distribution map is a two-dimensional array composed of the drying risk values ​​(ranging from 0 to 1) of each pixel output by S3.

[0034] The drying risk distribution map is integrated across each coverage zone to obtain the cumulative risk value for each zone. The integration process is as follows: For each coverage zone, it is uniformly divided into multiple sub-regions. The side length of each sub-region is 2 pixels; that is, each zone is divided into its width (width divided by 2, rounded down) and its height (height divided by 2, rounded down). If the division is not exact, a complete pixel is retained at the edge as the last sub-region. The drying risk value of the pixel at the center of each sub-region is extracted. If the center of the sub-region is exactly located on the pixel boundary, the arithmetic mean of the drying risk values ​​of the four pixels adjacent to that center is taken.

[0035] The drying risk values ​​of each sub-region are weighted and summed using the shortest distance from its center to the boundary of the partition as the weight: For each sub-region, the shortest distance from its center point to the four boundaries of its partition is calculated in pixels. This shortest distance is incremented by 1 and used as the weight value. The drying risk value of each sub-region is multiplied by its weight to obtain the weighted contribution value. The weighted contribution values ​​of all sub-regions are summed to obtain the weighted sum.

[0036] The weighted sum is divided by the total area of ​​the sub-regions (multiplied by the area of ​​each individual sub-region, with each individual sub-region area being 4 square pixels) to obtain the cumulative risk value for that zone. The cumulative risk value ranges from 0 to 1, reflecting the overall water shortage level of the zone. Sub-regions closer to the zone boundary have a lower weight, while sub-regions closer to the zone center have a higher weight, to avoid double-counting of risks due to overlapping coverage by adjacent sprinklers at the zone edges.

[0037] The cumulative risk values ​​of each zone are sorted from largest to smallest, and divided into three risk levels: high, medium, and low. Specifically, after sorting the cumulative risk values ​​of all zones, the highest value in the first 1 / 3 is marked as high-risk, the middle 1 / 3 as medium-risk, and the last 1 / 3 as low-risk. If the total number of zones is not divisible by 3, the risk is allocated proportionally: high-risk zones include the first few zones in the sorted list (rounded to the nearest integer), low-risk zones include the same number of zones at the end of the sorted list, and the remainder are medium-risk. A pre-pulse group within the pulsation cycle is allocated to high-risk zones; that is, the duty cycle sequence allocated to this zone is concentrated in the first 1 / 3 of the pulsation spray cycle. A mid-pulse group is allocated to medium-risk zones, concentrated in the middle 1 / 3 of the pulsation cycle. A post-pulse group is allocated to low-risk zones, concentrated in the last 1 / 3 of the pulsation spray cycle. The pulsation spray cycle length is preset based on the wall area and the water pump capacity, and is set to 10 seconds.

[0038] Within each pulse group, a high-low level alternation timing sequence is generated with the pulse width proportional to the cumulative risk value of that partition. These sequences are then concatenated to obtain the duty cycle sequence for that partition. For high-risk partitions, in the first 1 / 3 of the cycle (i.e., the first 3.33 seconds), the cumulative risk value is multiplied by the total cycle duration and then by a reference pulse width coefficient (0.5 seconds) to obtain the width of one pulse for that partition. Pulses of equal width are then output at equal intervals within this time period, with a fixed low-level interval of 0.1 seconds between pulses. The actual number of output pulses is equal to the time period length divided by (pulse width plus 0.1 seconds) and rounded down. For medium-risk partitions, pulses are concentrated in the middle 1 / 3 of the cycle (from 3.33 seconds to 6.66 seconds); for low-risk partitions, pulses are concentrated in the last 1 / 3 of the cycle (from 6.66 seconds to 10 seconds). Each zone ultimately generates a square wave timing sequence with alternating high and low levels, i.e., a duty cycle sequence. The duration of the high level is the pulse width, and the duration of the low level is a fixed interval of 0.1 seconds. This duty cycle sequence serves as a control signal to drive the solenoid valves of the corresponding zone's sprinkler heads. If the cumulative risk value of a zone is 0, its duty cycle sequence is entirely low, meaning no spraying. If the cumulative risk value is 1, a high level is continuously output during the corresponding time period, meaning continuous spraying for the entire time period. The pulse width of intermediate values ​​is linearly interpolated proportionally. The duty cycle sequences of each zone are independent and can be executed in parallel.

[0039] In S5, zoned spraying is performed according to the duty cycle sequence, and the newly acquired time-series image sequence and environmental data after spraying are fed back to S2 to iteratively update the evaporation rate field, specifically including: The zoned spraying is executed according to the calculated duty cycle sequence for each zone. The solenoid valve of the nozzle corresponding to each zone opens or closes according to the duty cycle sequence of that zone, with the nozzle spraying water during the high-level duration and the nozzle closing during the low-level duration.

[0040] Immediately after spraying, the same sampling frequency (1 frame per minute) and data acquisition method as S1 were used to re-acquire time-series image sequences of the concrete exterior wall surface after spraying, as well as ambient temperature, relative humidity, wind speed, and solar radiation intensity values. The newly acquired image sequences were used as post-spraying observation data for subsequent iterative updates.

[0041] The actual humidification amount at each location is obtained by subtracting the newly acquired time-series image sequence after spraying from the current humidity distribution field pixel by pixel. Specifically, the first newly acquired image frame (i.e., the image taken immediately after spraying) is converted to grayscale, and the grayscale value of each pixel is converted to the relative humidity value after spraying according to the same grayscale-humidity mapping curve in S1. The current humidity distribution field refers to the humidity distribution field output by S2 before spraying. For each pixel, the relative humidity value after spraying is subtracted from the relative humidity value before spraying. If the difference is positive, it represents the increase in humidity at that location due to spraying, which is the actual humidification amount; if the difference is negative or zero, the actual humidification amount is 0. The actual humidification amounts of all pixels constitute a two-dimensional array with the same size as the humidity distribution field.

[0042] The actual humidification is compared with the theoretical humidification calculated based on the duty cycle sequence, and abnormal pixels whose deviation exceeds a preset tolerance are extracted. The theoretical humidification is calculated as follows: for each zone, the proportion of the total high-level duration in its duty cycle sequence to the entire pulsed spray cycle (10 seconds) is multiplied by the rated flow coefficient of the spray head in that zone (pre-calibrated, in units of humidity increase per second), and then multiplied by the spray cycle duration to obtain the theoretical humidification for each pixel in that zone. Since the theoretical humidification is the same for all pixels in the same zone, the theoretical humidification for the entire zone is a constant. The theoretical humidification for each zone is filled into the corresponding pixel positions to obtain a theoretical humidification array with the same size as the humidity distribution field. For each pixel, the absolute value of the difference between its actual humidification and theoretical humidification is calculated, and this absolute value is used as the original deviation for that pixel. The original deviation reflects the difference between the spray effect and the expectation.

[0043] The specific process for extracting anomalous pixels is as follows: Calculate the global median of the original deviations of all pixels. The global median is calculated by sorting the original deviation values ​​of all pixels from smallest to largest and taking the value in the middle position after sorting; if the total number of pixels is even, take the average of the two middle values. Using this global median as a benchmark, calculate the local deviation of each pixel from the median deviation of its eight neighboring pixels. For each pixel, take the original deviation values ​​within a 3x3 neighborhood centered on that pixel (a total of 9 pixels, including itself), calculate the median of these 9 values, and record it as the neighborhood deviation median of that pixel. Calculate the absolute value of the difference between the original deviation of that pixel and the median of its neighborhood deviations; this is called the local deviation. Calculate the arithmetic mean of the nine original deviation values ​​within the neighborhood of that pixel; this is recorded as the neighborhood deviation mean. Pixels whose local deviation exceeds 1.3 times the global median and is also greater than the neighborhood deviation mean are marked as anomalous pixels. If the global median is 0, it will still be 0 after 1.3 times. In this case, as long as the local deviation is greater than 0 and greater than the mean of the neighborhood deviation, it will be marked as an abnormal pixel.

[0044] A correction field is constructed by spreading outwards from the abnormal pixel as the center. This correction field is then superimposed on the evaporation rate field to complete the iterative update of the evaporation rate field. The method for constructing the correction field is as follows: For each abnormal pixel, a circular region with a radius of 5 pixels is generated, centered on that pixel. A correction value is filled into all pixels within this region, equal to the original deviation of the abnormal pixel divided by the theoretical humidification rate (if the theoretical humidification rate is 0, the correction value is 0). If the circular regions of multiple abnormal pixels overlap, the correction value at the overlapping position is the arithmetic mean of all overlapping correction values. For pixels not covered by any abnormal pixel's circular region, the correction value is 0. All the above correction values ​​are arranged according to pixel position to form the correction field. The correction field is then superimposed pixel-by-pixel with the current evaporation rate field (i.e., the evaporation rate field before spraying) to obtain the updated evaporation rate field. The updated evaporation rate field is used for the S2 calculation of the next spraying cycle, thus achieving closed-loop adaptive control. If no abnormal pixels are detected in this round, the evaporation rate field remains unchanged, and no iterative update is performed.

[0045] The working principle of this invention is as follows: Time-series images of the wall surface and environmental data such as temperature, humidity, wind speed, and solar radiation are collected via cameras and sensors. The images are decomposed into spatiotemporal humidity representations by pixels, and a spatiotemporal field of diffusion coefficient is constructed by combining environmental data. An evaporation rate field conforming to the moisture continuity equation is then derived. The evaporation rate field is then fused with the current humidity distribution field to generate a dryness risk distribution map. Based on the dryness risk distribution map, the wall surface is mapped to each coverage zone of the spray array, and a duty cycle sequence within the pulsating spray cycle is calculated for each zone. Spraying is performed according to the duty cycle sequence, and newly acquired images and environmental data after spraying are fed back. Abnormal pixels are extracted by comparing the actual humidification with the theoretical humidification, and a correction field is constructed and superimposed onto the evaporation rate field to achieve iterative updates.

[0046] The foregoing has provided a detailed description of one embodiment of the present invention, but this description is merely a preferred embodiment and should not be construed as limiting the scope of the invention. All equivalent variations and modifications made within the scope of the claims of this invention should still fall within the patent coverage of this invention.

Claims

1. An adaptive spraying control method for the curing status of concrete exterior walls based on depth visual analysis, characterized in that, Includes the following steps: S1, acquire time-series image sequences of the concrete exterior wall surface, and simultaneously acquire ambient temperature, humidity, wind speed and solar radiation data; S2, decompose the time-series image sequence into spatiotemporal humidity characterization quantities by pixels, and construct the spatiotemporal field of diffusion coefficient based on environmental temperature and humidity, wind speed and solar radiation data, and then use the spatiotemporal field of spatiotemporal humidity characterization quantities and diffusion coefficient to invert the evaporation rate field that conforms to the water continuity equation. S3, integrates the evaporation rate field with the humidity distribution field at the current moment to generate a dry risk distribution map that reflects the urgency of water shortage at each spatial location; S4. Based on the dryness risk distribution map, map each spatial location to the coverage area of ​​the preset spray array, and calculate the duty cycle sequence within the pulsating spray cycle for each area. S5 executes zoned spraying according to the duty cycle sequence, and simultaneously feeds back the newly acquired time-series image sequence and environmental data after spraying to S2 to iteratively update the evaporation rate field.

2. The adaptive spraying control method for the curing status of concrete exterior walls based on depth visual analysis according to claim 1, characterized in that, The inverse evaporation rate field conforming to the water continuity equation specifically includes: The humidity attenuation gradient of corresponding pixels in three consecutive frames is extracted from the spatiotemporal humidity characterization to form the observation residual sequence; Substituting the observed residual sequence and the spatiotemporal field of the diffusion coefficient into the spatial discretization form of the moisture continuity equation, and minimizing the sum of squared residuals by adjusting the implicit evaporation term, the initial value of the evaporation rate of each pixel is obtained. The initial evaporation rate is anisotropically smoothed according to the direction of solar radiation, so that the evaporation rate on the sunlit side is higher than that on the shaded side, thus generating the final evaporation rate field.

3. The adaptive spraying control method for the curing status of concrete exterior walls based on depth visual analysis according to claim 2, characterized in that, The process of obtaining the initial evaporation rate value for each pixel specifically includes: The spatial discretization form of the moisture continuity equation is decomposed into two one-dimensional iterative operators along the horizontal and vertical directions of the wall. Traverse each pixel in the order of horizontal first and then vertical, and calculate the correction amount of the implicit evaporation term based on the weighted average of the observation residual of the pixel and the diffusion coefficients in the four directions of its neighborhood. Repeat the horizontal and vertical scans until the change in the sum of squared residuals between two adjacent scans is lower than a preset threshold. Then, output the implicit evaporation term of each pixel at this point as the initial value of the evaporation rate.

4. The adaptive spraying control method for the curing status of concrete exterior walls based on depth visual analysis according to claim 1, characterized in that, The generation of the aridity risk distribution map, reflecting the urgency of water shortage at various spatial locations, specifically includes: Based on the current humidity distribution field, pixels with humidity below a preset humidity threshold are marked as areas to be evaluated. Within the region to be evaluated, the local gradient direction of the evaporation rate field and the humidity distribution field is multiplied by a dot to obtain the projection component of the moisture loss vector in the direction of the fastest humidity decrease. The projected component is multiplied by the time accumulation of the evaporation rate field, and after normalization, the drying risk value of each pixel is obtained and filled into the corresponding position to form a drying risk distribution map.

5. The adaptive spraying control method for the curing status of concrete exterior walls based on depth visual analysis according to claim 4, characterized in that, The process of obtaining the projection components is as follows: A square neighborhood is selected centered on each pixel in the region to be evaluated. The central difference of the humidity value in the neighborhood is calculated to obtain the orientation angle of the local gradient direction. Based on the orientation angle, the evaporation rate field at the pixel is decomposed into components parallel to the local gradient direction and orthogonal components. The values ​​of the parallel components are extracted and multiplied by the magnitude of the local gradient direction to obtain the projection components.

6. The adaptive spraying control method for the curing status of concrete exterior walls based on depth visual analysis according to claim 1, characterized in that, The output process of the duty cycle sequence is as follows: The dry risk distribution map is integrated according to each coverage zone of the preset spray array to obtain the cumulative risk value of each zone; After sorting the cumulative risk values ​​of each partition, they are divided into three risk levels: high, medium, and low. The high-risk level is assigned a pre-pulse group within the pulsation cycle, the medium-risk level is assigned a mid-pulse group, and the low-risk level is assigned a post-pulse group. Within each pulse group, a high-low level alternating timing sequence is generated with the pulse width proportional to the cumulative risk value of the partition. After splicing, the duty cycle sequence of the partition is obtained.

7. The adaptive spraying control method for the curing status of concrete exterior walls based on depth visual analysis according to claim 6, characterized in that, The process of obtaining the cumulative risk value for each partition specifically includes: Each coverage area is evenly divided into multiple sub-regions, and the dryness risk value at the center of each sub-region is extracted. The aridity risk values ​​of each center are weighted and summed using the shortest distance from the center of the sub-region to the boundary of the partition as the weight. Divide the accumulated value by the total area of ​​the sub-region to obtain the cumulative risk value of the partition.

8. The adaptive spraying control method for the curing status of concrete exterior walls based on depth visual analysis according to claim 1, characterized in that, The iterative update of the evaporation rate field specifically includes: The actual humidification at each location is obtained by subtracting the newly acquired time-series image sequence after spraying from the current humidity distribution field by pixels. The actual humidification amount is compared with the theoretical humidification amount calculated based on the duty cycle sequence, and abnormal pixels whose deviation from the theoretical humidification amount exceeds the preset tolerance are extracted. A correction field is constructed by spreading outward from the abnormal pixel as the center. The correction field is then superimposed on the evaporation rate field output by S2 to complete the iterative update of the evaporation rate field.

9. The adaptive spraying control method for the curing status of concrete exterior walls based on depth visual analysis according to claim 8, characterized in that, The process for extracting the abnormal pixels is as follows: Calculate the absolute value of the difference between the actual humidification amount and the theoretical humidification amount for each pixel, and use it as the original deviation; Using the global median of the original deviation as a benchmark, calculate the local deviation of each pixel from the median of its eight neighboring pixels; Pixels whose local deviation exceeds 1.3 times the global median of the original deviation and is also greater than the mean of the neighborhood deviation are marked as abnormal pixels.

Images