Intelligent detection method and system for desulfurization strain activity parameters based on hyperspectral image

Abstract

The present application belongs to the technical field of image processing, and particularly relates to a desulfurization strain activity parameter intelligent detection method and system based on hyperspectral images, which comprises the following steps: extracting the original reflectivity of each pixel point in the desulfurization bacterial liquid hyperspectral image under continuous wave bands; anchoring the local background baseline by using the original reflectivity at the first and second reference wave bands, and determining the metabolic absorption depth in combination with the original reflectivity at the characteristic metabolic wave band; determining the spatial spectral gradient bias based on the original reflectivity difference of each pixel point and its spatial neighborhood pixel points at the structural scattering wave band and the metabolic absorption depth difference between the pixel points; and determining the strain activity evaluation parameter according to the spatial spectral gradient bias and the metabolic absorption depth of all pixel points. The present application can eliminate the interference of background water fluctuation and high-density inactivated bacterial liquid, realize the real characterization of microbial metabolic intensity, and improve the accuracy and stability of the desulfurization system monitoring.

Description

Technical Field

[0001] This invention relates to the field of image processing technology. More specifically, this invention relates to an intelligent detection method and system for desulfurization bacterial activity parameters based on hyperspectral images. Background Technology

[0002] In the treatment of industrial sulfur-containing wastewater, the vitality of desulfurizing bacteria is a core factor determining the treatment efficiency and effluent quality stability of biological desulfurization systems. Therefore, real-time monitoring of the bacterial community status within the reactor is crucial. Currently, hyperspectral imaging technology, with its advantages of non-contact and high resolution, has been introduced into the monitoring of desulfurizing bacteria. Existing technologies typically acquire images of the bacterial solution inside the bioreactor using a hyperspectral camera and utilize the raw average reflectance under specific characteristic bands, or extract the reflectance of characteristic absorption bands and near-infrared reference bands and calculate their static ratio, to assess the overall concentration and activity status of the mixed bacterial solution within the desulfurization reactor.

[0003] However, the actual operating environment of industrial bioreactors is complex. During the metabolic decomposition of sulfides in wastewater, highly active desulfurizing bacteria accumulate a large number of insoluble elemental sulfur particles on their cell surface. This accumulation of biological metabolites significantly alters the optical scattering characteristics of individual colonies and their surrounding microenvironment. When a large number of high-density inactivated desulfurizing colonies exist in the wastewater treatment system, the overall macroscopic optical scattering intensity exhibited is very similar to that of a very small number of highly active, high-quality desulfurizing colonies. Existing technologies rely solely on the average reflectance of a single region or a fixed global spectral index, ignoring the spatial spectral heterogeneity of highly active desulfurizing bacteria due to the uneven aggregation of metabolites in the microscopic spatial distribution. This leads to the monitoring system frequently misjudging high-density inactivated bacterial solutions as highly active metabolic states, resulting in distorted feedback signals to the automatic control system. This not only fails to provide accurate guidance for nutrient dosing but also easily triggers the collapse of the desulfurization system, severely affecting the continuity of wastewater treatment and making it difficult to meet effluent quality standards. Summary of the Invention

[0004] To address the technical problem in the prior art where high-density inactivated bacterial solutions are misjudged as highly active due to the neglect of spatial spectral heterogeneity caused by the aggregation of metabolites of highly active bacterial communities, the present invention provides solutions in the following aspects.

[0005] In a first aspect, the present invention provides an intelligent detection method for desulfurization bacterial activity parameters based on hyperspectral images, comprising:

[0006] The raw reflectance of each pixel in the hyperspectral image of the desulfurization bacterial solution is extracted in continuous bands. Characteristic metabolic bands and first and second reference bands are determined from the continuous bands. The raw reflectance of each pixel in the first and second reference bands is used to anchor the local background baseline. A ratio bias calculation is performed based on the raw reflectance in the characteristic metabolic bands to obtain the metabolic absorption depth. Structural scattering bands are determined from the continuous bands. Attenuation weights are constructed based on the difference in raw reflectance between each pixel and its spatial neighbors in the structural scattering bands. The differences in metabolic absorption depth among pixels are weighted and summed using these attenuation weights to obtain the spatial spectral gradient bias. A coefficient of variation modulation term is constructed based on the mean and standard deviation of the spatial spectral gradient biases of all pixels. The metabolic absorption depth is weighted using the spatial spectral gradient biases of each pixel, and the weighted result is fused with the coefficient of variation modulation term to obtain the bacterial activity evaluation parameters.

[0007] This invention extracts the original reflectance of the hyperspectral image of desulfurizing bacterial solution in continuous bands and dynamically anchors the local background baseline using the first and second reference bands, which are minimally affected by metabolism. Combined with characteristic metabolic bands, a ratio bias calculation is performed. This offsets the spectral baseline drift caused by water turbidity fluctuations during the calculation of metabolic absorption depth, achieving a true characterization of the intensity of cytochrome metabolic consumption in microscopic local areas. Furthermore, this invention constructs an attenuation weight by comparing the original reflectance difference between the central pixel and its spatial neighbors in the structural scattering band. This weight is then used to modulate the spatial difference of metabolic absorption depth to obtain the spatial spectral gradient bias. This effectively filters out abrupt interference from the physical edges of suspended particulate matter using physical similarity constraints, capturing the microscopic heterogeneous distribution characteristics caused by the life activities of desulfurizing bacteria. This solves the problem of interference identification caused by the homogenization of the internal biochemical structure of high-density inactivated bacterial communities. This invention utilizes the spatial spectral gradient bias to globally adaptively weight the metabolic absorption depth, and integrates it with the variation coefficient modulation term constructed from the mean and standard deviation of the spatial spectral gradient bias of all pixels. This makes the weights of the homogenized background water and dead bacterial clumps adaptively approach zero, thereby improving the evaluation index's resistance to industrial water quality fluctuations.

[0008] Preferably, determining the characteristic metabolic band and the first and second reference bands includes: comparing samples of highly active desulfurizing bacteria and inactive desulfurizing bacteria, and defining the interval in the spectral curve of the highly active desulfurizing bacteria sample where the original reflectance is minimized due to photon absorption by cytochrome as the empirical range of the characteristic metabolic band; finding the band intervals on both sides of the characteristic metabolic band that are not significantly affected by the activity state of the desulfurizing bacteria and where the change in original reflectance is the most gradual, and defining them as the empirical ranges of the first and second reference bands respectively; calculating the average original reflectance of each wavelength in the hyperspectral image of the desulfurizing bacterial solution, and determining the wavelength with the smallest average original reflectance within the empirical range of the characteristic metabolic band as the characteristic metabolic band; and determining the wavelengths with the largest average original reflectance within the empirical ranges of the first and second reference bands respectively as the first and second reference bands.

[0009] This invention achieves accurate positioning of the characteristic metabolic band and the empirical range of the reference band by comparing different active samples and finding local minima and flat intervals during the calibration stage. Then, it uses real-time average reflectance for adaptive optimization, which effectively avoids the distortion of biochemical characteristics caused by the shift of cytochrome characteristic absorption peaks due to fluctuations in the reactor's internal environment. This ensures that the subsequently extracted metabolic data is always anchored in the core response region where microbial life metabolic activities are most intense and least affected by background interference.

[0010] Preferably, the metabolic absorption depth satisfies the expression: In the formula, The first image in the hyperspectral image of the desulfurization bacterial solution is shown. Line number The metabolic absorption depth of the pixels in the column; The first image in the hyperspectral image of the desulfurization bacterial solution is shown. Line number The pixels in the column are in the characteristic metabolic band The original reflectance at that location; The first image in the hyperspectral image of the desulfurization bacterial solution is shown. Line number The pixels of the column in the first reference band The original reflectance at that location; The first image in the hyperspectral image of the desulfurization bacterial solution is shown. Line number The pixels of the column are in the second reference band. The original reflectance at that location; This represents the first minimum constant to prevent division by zero.

[0011] This invention utilizes the reflectance of a reference band that is unaffected by the concentration of local metabolites to dynamically anchor the local background baseline, and performs a ratio biasing process between the reference band reflectance and the characteristic band reflectance. This physically and logically cancels the background interference caused by the common translation of absolute reflectance and eliminates the influence of spectral baseline drift caused by global water turbidity fluctuations. As a result, the extracted depth parameters can purely and realistically reflect the intensity of cytochrome metabolic consumption in microscopic local areas.

[0012] Preferably, determining the structure scattering band from the continuous band includes: extracting spectral data of desulfurizing bacterial suspensions of different physical densities in the near-infrared band; defining the continuous spectral range where the correlation coefficient between the original reflectance and physical density reaches the positive maximum confidence interval, and where the difference in original reflectance and the gradient of adjacent wavelength changes are the smallest among different active samples at the same physical density, as the empirical range of the structure scattering band; calculating the average original reflectance of each wavelength in the hyperspectral image of the desulfurizing bacterial solution, and determining the wavelength with the largest average original reflectance within the empirical range of the structure scattering band as the structure scattering band.

[0013] This invention provides a purely physical spatial morphological constraint benchmark for the system, free from biochemical metabolic interference, by extracting a continuous spectral range in the near-infrared band that is strictly positively correlated with physical density and unaffected by the active state as the structural scattering band. This ensures that the scattering characteristics characterizing the physical density of bacterial cell clusters can be accurately obtained in complex industrial wastewater suspension environments, providing a reliable physical basis for the subsequent accurate identification of inactive bacterial communities with homogenized internal biochemical structures.

[0014] Preferably, the spatial spectral gradient bias satisfies the expression: In the formula, The first image in the hyperspectral image of the desulfurization bacterial solution is shown. Line number Spatial spectral gradient bias of the pixels in the column; The first image in the hyperspectral image of the desulfurization bacterial solution represents the... Line number The spatial neighborhood centered on the pixel in the column; Representing spatial neighborhood The row and column coordinates of any neighboring pixel within the range; , These represent the first and second images of the desulfurization bacterial solution in the hyperspectral image. Line number Column pixels, neighboring pixels The depth of metabolic absorption; , These represent the first and second images of the desulfurization bacterial solution in the hyperspectral image. Line number Column pixels, neighboring pixels In the structural scattering band The original reflectance at that location; Represents an exponential function with the natural constant as its base; This represents the second minimum constant to prevent division by zero.

[0015] This invention constructs attenuation weights by utilizing the physical similarity of structural scattering bands, and uses this to nonlinearly modulate the spatial difference of metabolic absorption depth. This enables the use of physical spatial attenuation laws to constrain the spatial changes of metabolic gradients, thereby adaptively weakening the microscopic pseudo-gradients caused by water interfaces. It effectively filters out abrupt interference caused by the physical edges of suspended particulate matter, allowing the extracted gradient bias to be highly focused on the microscopic biochemical heterogeneity distribution caused by the real life activities of desulfurizing bacteria.

[0016] Preferably, the bacterial strain activity evaluation parameters satisfy the following expression: In the formula, Indicates parameters for evaluating the activity of bacterial strains; and These represent the height and width of the hyperspectral image of the desulfurization bacterial solution, respectively. The first image in the hyperspectral image of the desulfurization bacterial solution is shown. Line number The metabolic absorption depth of the pixels in the column; The first image in the hyperspectral image of the desulfurization bacterial solution is shown. Line number Spatial spectral gradient bias of the pixels in the column; The standard deviation of the spatial spectral gradient bias of all pixels in the hyperspectral image of the desulfurization bacterial solution; This represents the mean of the spatial spectral gradient bias of all pixels in the hyperspectral image of the desulfurization bacterial solution. Represents the natural constant; Represents a logarithmic function with the natural constant as its base; This represents the third minimum constant to prevent division by zero.

[0017] This invention introduces the coefficient of variation of the spatial spectral gradient bias as a nonlinear gain term for global modulation, and uses it to adaptively weight and fuse the metabolic depth of each pixel. This allows high-quality colony regions exhibiting high metabolic heterogeneity to dominate the global evaluation value, thereby eliminating the false contribution of homogenized background water and dead bacterial clumps to the overall activity at the macroscopic level. This enhances the computational stability and resistance to water quality fluctuations of the bacterial activity evaluation parameters under extremely complex working conditions.

[0018] Preferably, the method further includes: generating a comprehensive control threshold based on historical microbial activity evaluation parameters within a time sliding window, comparing the current microbial activity evaluation parameters with the comprehensive control threshold, and outputting corresponding process control instructions to the automatic control system based on the comparison results.

[0019] Preferably, the step of generating a comprehensive control threshold based on historical microbial activity evaluation parameters within a time sliding window includes: obtaining the current hydraulic retention time of the sulfur-containing wastewater bioreactor; establishing a time sliding window with the current hydraulic retention time as the time span; extracting all historically continuously output microbial activity evaluation parameters within the time sliding window; obtaining the mean and standard deviation of all microbial activity evaluation parameters within the time sliding window; calculating the difference between the mean of the microbial activity evaluation parameters and three times the standard deviation of the microbial activity evaluation parameters; using the difference as an adaptive activity dynamic threshold; and extracting the maximum value between the adaptive activity dynamic threshold and a preset absolute activity baseline threshold as the comprehensive control threshold.

[0020] Preferably, the step of outputting corresponding process control instructions to the automatic control system based on the comparison results includes: in response to the current microbial activity evaluation parameter being greater than or equal to the comprehensive control threshold, determining that the current desulfurization microbial community is in a stable and healthy metabolic state, and outputting instructions to the automatic control system to maintain the current operating parameters; in response to the current microbial activity evaluation parameter being less than the comprehensive control threshold, determining that an abnormal desulfurization microbial inactivation has occurred in the current sulfur-containing wastewater bioreactor, generating a low-activity warning signal, and outputting process control instructions to the automatic control system to increase the dosage of nutrient agent and increase the opening frequency of the bottom sludge discharge valve.

[0021] Secondly, the present invention provides an intelligent detection system for desulfurization bacterial activity parameters based on hyperspectral images, comprising a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the above-mentioned intelligent detection method for desulfurization bacterial activity parameters based on hyperspectral images is implemented.

[0022] By adopting the above technical solution, the intelligent detection method for desulfurization bacterial activity parameters based on hyperspectral images is generated into a computer program and stored in a memory for loading and execution by a processor. This allows for the creation of a terminal device based on the memory and processor, facilitating its use.

[0023] The beneficial effects of this invention are as follows: By deeply exploring the unique microscopic accumulation phenomenon of elemental sulfur during the metabolism of desulfurizing bacteria, this invention utilizes a reference band to dynamically anchor the local background baseline and performs ratio bias calculations to obtain the metabolic absorption depth. This effectively eliminates the influence of spectral baseline drift caused by global water turbidity fluctuations, achieving a true characterization of the intensity of cytochrome metabolic consumption in microscopic local regions. This invention combines the physical similarity of structural scattering bands to nonlinearly modulate the spatial difference of metabolic absorption depth, enabling the constructed spatial spectral gradient bias to accurately identify active bacterial floc regions with heterogeneous biochemical processes. This solves the technical defect of existing technologies that misjudge high-density inactive bacterial solutions as highly active states due to neglecting microscopic spatial distribution characteristics. This invention uses the spatial spectral gradient bias, representing the microscopic local biochemical heterogeneity characteristics, as the core basis for global adaptive weighting and integrates the variation coefficient modulation term of the global gradient for positive gain compensation. This makes the weights of homogenized background water and dead bacterial clumps approach zero, improving the output parameters' resistance to disturbances caused by industrial water quality fluctuations. This invention generates a comprehensive control threshold by introducing a dual threshold constraint mechanism based on historical statistical characteristics and physical and biochemical limits. This eliminates the possibility of false health misjudgments that may occur under continuous low activity conditions, ensuring the rapid recovery and stable operation of the biological desulfurization system, and guaranteeing the continuity of wastewater treatment and the stability of effluent quality. Attached Figure Description

[0024] Figure 1 This is a flowchart illustrating the intelligent detection method for desulfurization bacterial activity parameters based on hyperspectral images in this invention;

[0025] Figure 2 This is a graph showing the variation of the original reflectance of the structural scattering band and the spatial spectral gradient bias of the pixels along the main diagonal of the desulfurization bacterial solution in the hyperspectral image. Detailed Implementation

[0026] 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.

[0027] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0028] This invention discloses an intelligent detection method for desulfurization bacterial activity parameters based on hyperspectral images, referring to... Figure 1 This includes steps S1-S5:

[0029] S1: Extract the original reflectance of each pixel in the hyperspectral image of the desulfurization bacterial solution in continuous bands.

[0030] It should be noted that the internal environment of industrial desulfurization reactors is closed and undergoes drastic dynamic changes. Traditional contact sampling cannot continuously obtain information on the spatial distribution morphology and spectral response of bacterial colonies. Therefore, this invention utilizes hyperspectral imaging equipment to acquire hyperspectral images in real time to obtain basic data carriers containing the spatial distribution morphology and continuous spectral response of microorganisms.

[0031] Specifically, a hyperspectral imaging device is used to acquire hyperspectral images of the desulfurization bacterial solution in a sulfur-containing wastewater bioreactor in real time, and the original reflectance of each pixel in the hyperspectral image of the desulfurization bacterial solution is extracted in a continuous band, which covers the visible to near-infrared spectral range.

[0032] S2: Determine the characteristic metabolic band and the first and second reference bands from the continuous bands, anchor the local background baseline using the original reflectance of each pixel at the first and second reference bands, and perform a ratio bias calculation by combining the original reflectance at the characteristic metabolic band to obtain the metabolic absorption depth.

[0033] It should be noted that the water inside industrial-grade bioreactors experiences significant turbidity fluctuations, and the original reflectance of the hyperspectral image of the desulfurization bacterial solution is highly susceptible to baseline drift caused by global scattering from suspended particulate matter, resulting in the true spectral features being masked by background noise. Furthermore, due to fluctuations in pH and temperature within the reactor during actual operation, as well as the microscopic metabolic succession of the dominant desulfurization strains, the actual characteristic absorption peaks of cytochromes undergo slight red-shifts or blue-shifts. Directly using fixed empirical wavelength values ​​can easily deviate from the true absorption extreme points, leading to severe distortion of the extracted biochemical features. Therefore, this invention adaptively searches for real-time characteristic bands within a preset empirical band range. Based on the physical response logic of envelope removal, a dynamic linear baseline is constructed using two adjacent reference bands to extract the true spectral features from the background noise, obtaining pure cytochrome metabolic absorption characteristics.

[0034] Specifically, in a laboratory setting, a laboratory spectrometer was used to perform continuous band spectral scanning and comparison on samples of highly active desulfurizing bacteria and inactive desulfurizing bacteria. On the spectral curve of the highly active desulfurizing bacteria samples, the spectral range containing the local minimum of original reflectance caused by strong photon absorption by cytochromes was identified. This range corresponds to the band where microbial metabolic activity is most intense, and it is defined as the empirical range of the characteristic metabolic band. The function of this characteristic metabolic band is to locate the core response region of the desulfurizing bacteria's biochemical metabolism. Based on the calibration results of typical desulfurizing strains, the empirical range of the characteristic metabolic band is as follows: nanometer to nanometer.

[0035] To eliminate global spectral baseline drift caused by fluctuations in water turbidity, it is necessary to find the band intervals on both sides of the characteristic metabolic band that are not significantly affected by the activity state of desulfurizing bacteria and have the most gradual changes in original reflectance. These band intervals are then calibrated as the empirical range of the first reference band and the empirical range of the second reference band, respectively. The purpose of calibrating the first and second reference bands is to provide a pure background reference point.

[0036] Specifically, for each candidate wavelength within the spectral ranges to the left and right of the characteristic metabolic band, a reference band stability index is calculated based on the spectral scanning data of highly active desulfurizing bacteria samples and inactive desulfurizing bacteria samples. Satisfying the expression:

[0037]

[0038] In the formula, Indicates candidate wavelength Reference band stability index at the location; This indicates that the highly active desulfurizing bacteria sample is at the candidate wavelength. The original reflectance at that location; Indicates the inactivated desulfurization bacteria sample at the candidate wavelength The original reflectance at that location; Indicates two samples at candidate wavelengths The absolute value of the original reflectance difference at the location; This indicates that the highly active desulfurizing bacteria sample is at the candidate wavelength. The first derivative of the original reflectivity at that point; Indicates the inactivated desulfurization bacteria sample at the candidate wavelength The first derivative of the original reflectivity at that point; and These represent the absolute values ​​of the corresponding first derivatives; This represents an exponential function with the natural constant as its base.

[0039] Among them, the absolute value of the difference in original reflectance This reflects the degree to which the wavelength is affected by the metabolic state of desulfurizing bacteria; the smaller the value, the less affected the wavelength is by the activity state of the desulfurizing bacteria. The absolute value of the first derivative reflects the degree of drastic fluctuation in the spectral curve at that wavelength; the smaller the absolute value of the first derivative, the smoother the spectral curve, and the closer the value of the exponent term is to 1. When the reference band stability index... When the value is at its minimum, it indicates that the candidate wavelength is not only minimally affected by the activity of desulfurizing bacteria, but also has the flattest local spectral curve, making it less prone to baseline jumps due to slight wavelength shifts.

[0040] Therefore, the stability index of the reference band is sought in the left and right intervals of the characteristic metabolic band, respectively. The minimum wavelength point was identified, and a minimal fixed redundancy range was expanded outwards from this wavelength point to both sides, which were respectively designated as the empirical ranges of the first and second reference bands. Based on the calibration results of typical desulfurization strains, the empirical range of the first reference band was... nanometer to nanometers, the empirical range of the second reference band is nanometer to nanometer.

[0041] Furthermore, the average raw reflectance of all pixels in the hyperspectral image of the desulfurization bacterial solution at each wavelength will be obtained. nanometer to The wavelength with the lowest average original reflectance within the nanometer range was determined as the characteristic metabolic band. , will nanometer to The wavelength with the highest average original reflectivity within the nanometer range was determined as the first reference band. , will nanometer to The wavelength with the highest average original reflectivity within the nanometer range was determined as the second reference band. .

[0042] Based on the original reflectance of each pixel in the hyperspectral image of the desulfurized bacterial solution at the characteristic metabolic band, the first reference band, and the second reference band, the metabolic absorption depth of each pixel in the hyperspectral image of the desulfurized bacterial solution is calculated:

[0043]

[0044] In the formula, The first image in the hyperspectral image of the desulfurization bacterial solution is shown. Line number The metabolic absorption depth of the pixels in the column; The first image in the hyperspectral image of the desulfurization bacterial solution is shown. Line number The pixels in the column are in the characteristic metabolic band The original reflectance at that location; The first image in the hyperspectral image of the desulfurization bacterial solution is shown. Line number The pixels of the column in the first reference band The original reflectance at that location; The first image in the hyperspectral image of the desulfurization bacterial solution is shown. Line number The pixels of the column are in the second reference band. The original reflectance at that location; This represents the first division-to-zero minimum constant, used to prevent computational overflow anomalies where the original reflectivity at the characteristic metabolic band approaches 0, resulting in a denominator of 0. In this embodiment, it is set to... To provide strict numerical protection in the event of optical acquisition blind spots or extreme dark field noise interference in the image, in other embodiments, implementers can make a first adjustment of the order of magnitude of the division-to-zero minimum constant based on the sampling bit width of the hyperspectral camera.

[0045] The more vigorous the desulfurization bacteria activity in the microscopic region corresponding to a pixel, the more effective its metabolic bands become. The more intense the photon absorption at a given point, the lower the original reflectance at the characteristic metabolic band. Fluctuations in water turbidity cause a synchronous overall vertical shift in the original reflectance of pixels across various bands, resulting in spectral baseline drift. This invention utilizes the sum of the original reflectance of a first reference band and a second reference band, which are unaffected by local metabolic product concentrations. To dynamically anchor the local background baseline, the ratio is calculated to the original reflectance of the characteristic metabolic band affected by metabolism, and a bias is applied by subtracting 1. The more vigorous the desulfurization bacteria activity, the higher the denominator. The value is much smaller than the numerator. The value of makes the metabolic absorption depth The larger the value, the more it offsets the background interference caused by the common shift in absolute reflectance, eliminates the influence of spectral baseline drift caused by global water turbidity, and improves the extraction of metabolic absorption depth. It can purely and realistically characterize the intensity of cytochrome metabolic consumption in microscopic local areas.

[0046] S3: Determine the structure scattering band from the continuous band, construct the attenuation weight based on the difference in original reflectance between each pixel and its spatial neighboring pixels in the structure scattering band, and use the attenuation weight to perform a weighted summation of the differences in metabolic absorption depth between pixels to obtain the spatial spectral gradient bias.

[0047] It should be noted that the intense metabolism and uneven accumulation of elemental sulfur particles inside highly active desulfurizing bacterial micelles lead to dramatic changes in absorption gradients within the microscopic space at the cell scale. In contrast, the biochemical structure inside a large number of inactive desulfurizing bacterial aggregates tends to be homogenized and cannot form localized dramatic gradients. This makes it difficult for conventional methods to distinguish between high-density dead bacteria and highly active live bacteria. Therefore, this invention combines the physical scattering similarity of non-metabolic regions to constrain and evaluate the spatial difference in metabolic absorption depth in order to identify active bacterial micelle regions with heterogeneous biological metabolism.

[0048] To extract wavelengths that purely characterize the scattering properties of the physical density of bacterial cell aggregates, this invention pre-calibrates the spectral characteristics of actual industrial wastewater suspensions in a laboratory environment. Specifically, a laboratory spectrometer is used to perform continuous spectral scanning on desulfurization bacterial suspensions of different physical concentrations. For each candidate wavelength in the near-infrared long-wavelength band of the non-metabolic absorption region, the statistical correlation coefficient between the original reflectance sequence and the corresponding physical density sequence of each sample group at that candidate wavelength is calculated. Wavelengths with correlation coefficients reaching the positive maximum confidence interval are selected as the initial set of wavelengths, thereby ensuring that the selected wavelengths have a strictly linear positive response to changes in the physical density of the suspended matter. In the initial band set, the original reflectance difference between highly active desulfurizing bacteria samples and inactive desulfurizing bacteria samples at each candidate wavelength is extracted under the same physical density. The gradient of the change in original reflectance between each candidate wavelength and adjacent wavelengths is also extracted. The continuous spectral range where the original reflectance difference approaches 0 and the gradient of the change in adjacent wavelengths is minimized is calibrated as the empirical range of the structure scattering band. Through the dual constraints of the original reflectance difference and the gradient of change, it can be ensured that the selected band is completely unaffected by the metabolic state of desulfurizing bacteria, eliminating the existence of biochemical absorption valleys, while ensuring minimal fluctuations in the spectral curve, avoiding baseline abrupt changes during subsequent feature extraction. The purpose of calibrating the structure scattering band in this invention is to provide a purely physical spatial morphological constraint benchmark completely unaffected by the biochemical metabolism of desulfurizing bacteria. Based on the calibration results of the near-infrared transmission physical characteristics of actual industrial wastewater, the empirical range of the structure scattering band is 750 nm to 850 nm. In other embodiments, implementers can recalibrate the boundary settings of the empirical range according to the actual turbidity and suspended particulate matter characteristics of the wastewater.

[0049] Furthermore, the average original reflectance of all pixels in the hyperspectral image of the desulfurization bacterial solution at various wavelengths is obtained, and the wavelength with the highest average original reflectance in the range of 750 nm to 850 nm is adaptively determined as the structural scattering band.

[0050] Based on the metabolic absorption depth and original reflectance of each pixel in the hyperspectral image of the desulfurization bacterial solution, the spatial spectral gradient bias of each pixel in the hyperspectral image of the desulfurization bacterial solution is calculated:

[0051]

[0052] In the formula, The first image in the hyperspectral image of the desulfurization bacterial solution is shown. Line number Spatial spectral gradient bias of the pixels in the column; The first image in the hyperspectral image of the desulfurization bacterial solution represents the... Line number The spatial neighborhood centered on the pixel in the column; Representing spatial neighborhood The row and column coordinates of any neighboring pixel within the range; The first image in the hyperspectral image of the desulfurization bacterial solution is shown. Line number The metabolic absorption depth of the pixels in the column; Represents neighboring pixels The depth of metabolic absorption; This represents the absolute value of the difference in metabolic absorption depth between the center pixel and its neighboring pixels. The first image in the hyperspectral image of the desulfurization bacterial solution is shown. Line number The pixels in the column in the structure scattering band The original reflectance at that location; Represents neighboring pixels In the structural scattering band The original reflectance at that location; This represents the absolute value of the difference in original reflectivity of the structural scattering band between the center pixel and its neighboring pixels. Represents an exponential function with the natural constant as its base; This represents the second division-to-zero minimum constant, used to prevent calculation overflow anomalies where the denominator is zero due to the original reflectivity of the structure's scattering band approaching zero. In this embodiment, it is set to... To provide strict numerical protection in situations where there are optical acquisition blind spots or extreme dark environments, in other embodiments, implementers can make a second adjustment to prevent division by zero by an order of magnitude based on the sampling bit width of the device.

[0053] When the physical scattering characteristics of the central pixel and its neighboring pixels are more similar, it indicates that they belong to the same bacterial floc. The smaller the value, The closer the value is to 1, the greater the absolute value of the difference in metabolic absorption depth. If the physical scattering characteristics are completely preserved, it indicates that the fluid boundary between the bacterial floc and the water body has been crossed. The value is drastically reduced to near zero, adaptively weakening the microscopic pseudo-gradients caused by the water interface. This invention utilizes the spatial attenuation weight of structural scattering characteristics to modulate the difference in metabolic gradients, filtering out abrupt interference from the physical edges of suspended particulate matter, so that the extracted spatial spectral gradient bias is focused on the microscopic heterogeneous distribution caused by the life activities of desulfurizing bacteria.

[0054] It should be further noted that the physical size of highly active bacterial flocs in industrial desulfurization reactors is typically on the order of tens of micrometers. Combined with the microscopic spatial resolution of a hyperspectral camera, the physical mapping distance of the 8-neighborhood precisely matches the scale of the core metabolic region of the bacterial flocs. Therefore, in this embodiment, the spatial neighborhood... The set is configured as an 8-neighborhood pixel set. In other embodiments, the implementer can set the spatial neighborhood range according to the spatial resolution of the actual hyperspectral camera.

[0055] For example, Figure 2 The graph shows the variation curves of the original reflectance and spatial spectral gradient bias of the structural scattering band pixels in the hyperspectral image of the desulfurization bacterial solution along the main diagonal. In the original reflectance curve of the structural scattering band, both inactivated and highly active bacterial clusters exhibit similar high-level reflectance intensities when passing through corresponding regions due to their high physical density. This demonstrates that relying solely on macroscopic scattering intensity or the average reflectance value is insufficient to effectively distinguish between dead and live bacteria with similar physical densities. Furthermore, in the corresponding spatial spectral gradient bias curve, the inactivated bacterial cluster region, due to the homogenization of its internal biochemical structure and the inability to form a local gradient, effectively mitigates interference signals. The effective filtering effect approaches 0. In contrast, highly active bacterial colonies generate a dramatic microscopic spatial absorption gradient due to the uneven accumulation of elemental sulfur metabolites on the cell surface. This results in significant high-frequency fluctuations and numerical enhancement in the spatial spectral gradient bias within the corresponding range. This demonstrates that by introducing the dimension of spatial heterogeneity, this invention can successfully eliminate the pseudo-high-value signal of homogeneous dead bacteria and lock in the microscopic biochemical characteristics representing real life activities. It solves the technical defects of existing technologies that use global reflectance averages or fixed band ratios to misjudge biological activity, and provides core underlying data support for achieving global high-fidelity activity assessment.

[0056] S4: Construct a coefficient of variation modulation term based on the mean and standard deviation of the spatial spectral gradient bias of all pixels, use the spatial spectral gradient bias of each pixel to weight the metabolic absorption depth, and fuse the weighted result with the coefficient of variation modulation term to obtain the bacterial activity evaluation parameters.

[0057] It should be noted that the local spatial spectral gradient bias can only distinguish between dead and live bacteria at the individual cell level. In the operation and control of the macroscopic biological desulfurization reactor, the automatic control system needs a global evaluation index. However, simply averaging the global values ​​can lead to the anomalous signal of high-density dead bacteria masking the overall trend and causing distortion in the macroscopic activity evaluation. Therefore, this invention transforms the spatial spectral gradient bias into a dynamic weighting coefficient and adaptively integrates the discrete variation of the global gradient to achieve a high-fidelity global evaluation of the overall vitality of the microbial species.

[0058] Specifically, the mean and standard deviation of the spatial spectral gradient bias of all pixels in the hyperspectral image of the desulfurization bacterial solution are obtained. Based on the spatial spectral gradient bias and metabolic absorption depth of each pixel in the hyperspectral image of the desulfurization bacterial solution, global bacterial activity evaluation parameters are calculated.

[0059]

[0060] In the formula, Indicates parameters for evaluating the activity of bacterial strains; and These represent the height and width of the hyperspectral image of the desulfurization bacterial solution, respectively. The first image in the hyperspectral image of the desulfurization bacterial solution is shown. Line number The metabolic absorption depth of the pixels in the column; The first image in the hyperspectral image of the desulfurization bacterial solution is shown. Line number Spatial spectral gradient bias of the pixels in the column; The standard deviation of the spatial spectral gradient bias of all pixels in the hyperspectral image of the desulfurization bacterial solution; This represents the mean of the spatial spectral gradient bias of all pixels in the hyperspectral image of the desulfurization bacterial solution. Represents the natural constant; Represents a logarithmic function with the natural constant as its base; This represents the third division-to-zero minimum constant, used to prevent calculation overflow anomalies where the sum or mean of the global spatial spectral gradient biases approaches zero when calculating bacterial activity evaluation parameters. In this embodiment, it is set as follows: This provides strict numerical protection under extreme conditions where the reactor experiences extreme toxic shocks that lead to the absolute homogeneous death of the microbial community, ensuring the computational stability of the macroscopic activity assessment system without causing it to collapse. In other embodiments, implementers can set a third prevention of division by zero minimum constant according to the system control accuracy requirements.

[0061] This invention is based on spatial spectral gradient bias. Depth of metabolic absorption Weighting is applied to ensure that only metabolic uptake depths from regions exhibiting high metabolic heterogeneity are considered. Only then can the overall evaluation values ​​be dominated, eliminating the impact of large areas of dead bacteria or inorganic sediment on the overall activity. When highly active desulfurization bacteria dominate in the reactor, the metabolic evolution progress among different high-quality colonies naturally differs, leading to increased dispersion of the overall gradient bias and thus increasing the standard deviation of the spatial spectral gradient bias. A sharp increase, standard deviation With the mean term The ratio increases, and the coefficient of variation modulation term The value of increases nonlinearly, and positive gain compensation is applied to the activity evaluation, where the natural constant is... This ensures that the lower limit of the modulation term is 1 under extremely homogeneous conditions. The present invention uses a spatial spectral gradient bias representing the microscopic local biochemical heterogeneity characteristics. As the core basis for global data weighted modulation, the weights of homogenized background water and dead bacteria clumps adaptively approach 0, eliminating the false contributions of background water environmental fluctuations and passive accumulation of dead bacteria to overall activity, thereby ensuring that the output bacterial activity evaluation parameters are accurate. It possesses strong resistance to disturbances caused by fluctuations in industrial water quality.

[0062] S5: Control the desulfurization process based on the current microbial activity evaluation parameters.

[0063] It should be noted that in actual industrial biological desulfurization systems, relying solely on fixed thresholds to assess bacterial activity can easily lead to frequent false alarms due to normal operating loads or fluctuations in influent water quality. Utilizing historical statistical features based on a time-sliding window to generate adaptive thresholds can effectively identify sudden toxic shocks. However, when the reactor experiences long-term latent degradation or remains in a low-activity state, a simple adaptive sliding window can cause baseline drift, meaning the historical mean continuously decreases, leading to a downward compromise of the threshold and misjudging a persistent low-activity anomaly as normal. Therefore, this invention introduces a dual threshold constraint mechanism, combining a dynamic threshold based on real-time statistical features with an absolute baseline threshold based on physical and biochemical limits. A nonlinear maximum value function locks in the final safety warning line, preserving the system's high sensitivity to sudden anomalies while eliminating logical loopholes that could lead to false health assessments under persistently low operating conditions.

[0064] Specifically, the current hydraulic retention time of the sulfur-containing wastewater bioreactor is obtained. A time-sliding window is established with this current hydraulic retention time as the time span. All historically continuously output bacterial activity evaluation parameters within this time-sliding window are extracted, and the mean and standard deviation of all bacterial activity evaluation parameters within the time-sliding window are calculated. Based on the mean and standard deviation of all bacterial activity evaluation parameters within the time-sliding window, an adaptive activity dynamic threshold is calculated.

[0065]

[0066] In the formula, Indicates the adaptive activity dynamic threshold; This represents the mean value of all bacterial activity evaluation parameters within the time sliding window; This represents the standard deviation of all bacterial activity evaluation parameters within the time sliding window; numerical values. It is a constant based on the distribution properties of the statistical Laida criterion.

[0067] Furthermore, based on the adaptive dynamic activity threshold and the absolute activity baseline threshold, the comprehensive control threshold is calculated:

[0068]

[0069] In the formula, Indicates the overall control threshold; Represents the maximum value function; Indicates the absolute activity baseline threshold; This represents the adaptive activity dynamic threshold.

[0070] Among them, the absolute activity bottom line threshold This refers to the lower limit of activity necessary to maintain the minimum desulfurization biochemical cycle in the bioreactor. In this invention, extreme survival and toxicity decay experiments were conducted on this desulfurization strain in a laboratory environment beforehand. Evaluation parameters corresponding to the critical point where the bacterial community's biochemical metabolism was about to completely collapse and stop were extracted and calibrated as empirical values ​​for the absolute activity baseline threshold, which were then written into the automatic control system. The empirical range is as follows: to In other embodiments, implementers may set the absolute activity baseline threshold based on the minimum tolerance designed for the actual desulfurization process.

[0071] The current bacterial activity evaluation parameters With comprehensive control threshold Perform a logical comparison:

[0072] Response parameters for evaluating bacterial activity at the current moment Greater than or equal to the comprehensive control threshold The system determines that the desulfurization bacteria in the current sulfur-containing wastewater bioreactor are in a stable and healthy metabolic state, and outputs instructions to the automatic control system to maintain the current operating parameters.

[0073] Response parameters for evaluating bacterial activity at the current moment Less than the comprehensive control threshold The system determines that the desulfurization bacteria in the current sulfur-containing wastewater bioreactor have been severely inactivated or have encountered a severe shock, generates a low activity warning signal, and outputs process control instructions to the automatic control system to increase the dosage of nutrients and increase the opening frequency of the bottom sludge discharge valve, thereby accelerating the replacement of the high-density inactivated bacterial clusters settled at the bottom and quickly restoring the overall biological sulfur metabolism activity of the system.

[0074] This invention also discloses an intelligent detection system for desulfurization bacterial activity parameters based on hyperspectral images, comprising a processor and a memory. The memory stores computer program instructions, which, when executed by the processor, implement the intelligent detection method for desulfurization bacterial activity parameters based on hyperspectral images according to this invention.

[0075] The system also includes other components well known to those skilled in the art, such as communication buses and communication interfaces, the settings and functions of which are known in the art and will not be described in detail here.

Claims

1. A method for intelligent detection of desulfurization bacterial activity parameters based on hyperspectral imaging, characterized in that, include: Extract the original reflectance of each pixel in the hyperspectral image of the desulfurization bacterial solution in continuous bands; The characteristic metabolic band and the first and second reference bands are determined from the continuous bands. The local background baseline is anchored by the original reflectance of each pixel at the first and second reference bands. The ratio bias calculation is performed by combining the original reflectance at the characteristic metabolic band to obtain the metabolic absorption depth. The structure scattering band is determined from the continuous band. The attenuation weight is constructed based on the difference in original reflectance between each pixel and its spatial neighbor pixels in the structure scattering band. The difference in metabolic absorption depth between pixels is weighted and summed using the attenuation weight to obtain the spatial spectral gradient bias. A coefficient of variation modulation term is constructed based on the mean and standard deviation of the spatial spectral gradient bias of all pixels. The metabolic absorption depth is weighted using the spatial spectral gradient bias of each pixel, and the weighted result is fused with the coefficient of variation modulation term to obtain the bacterial activity evaluation parameters. The determination of the characteristic metabolic band and the first and second reference bands includes: comparing samples of highly active desulfurizing bacteria and inactive desulfurizing bacteria, and defining the interval in the spectral curve of the highly active desulfurizing bacteria sample where the original reflectance is minimized due to photon absorption by cytochrome as the empirical range of the characteristic metabolic band; identifying the band intervals on both sides of the characteristic metabolic band that are not significantly affected by the activity state of the desulfurizing bacteria and where the change in original reflectance is the most gradual, and defining them as the empirical ranges of the first and second reference bands respectively; calculating the average original reflectance of each wavelength in the hyperspectral image of the desulfurizing bacterial solution, and determining the wavelength with the smallest average original reflectance within the empirical range of the characteristic metabolic band as the characteristic metabolic band; and defining the wavelengths with the largest average original reflectance within the empirical ranges of the first and second reference bands respectively as the first and second reference bands. Determining the structure scattering band from continuous bands includes: extracting spectral data of desulfurizing bacterial suspensions of different physical densities in the near-infrared band; defining the continuous spectral range where the correlation coefficient between the original reflectance and physical density reaches the maximum positive confidence interval, and where the difference in original reflectance and the gradient of adjacent wavelength changes are the smallest among different active samples at the same physical density, as the empirical range of the structure scattering band; calculating the average original reflectance of each wavelength in the hyperspectral image of the desulfurizing bacterial solution, and determining the wavelength with the largest average original reflectance within the empirical range of the structure scattering band as the structure scattering band.

2. The intelligent detection method for desulfurization bacterial activity parameters based on hyperspectral images according to claim 1, characterized in that, The metabolic absorption depth satisfies the expression: ； In the formula, The first image in the hyperspectral image of the desulfurization bacterial solution is shown. Line number The metabolic absorption depth of the pixels in the column; The first image in the hyperspectral image of the desulfurization bacterial solution is shown. Line number The pixels in the column are in the characteristic metabolic band The original reflectance at that location; The first image in the hyperspectral image of the desulfurization bacterial solution is shown. Line number The pixels of the column in the first reference band The original reflectance at that location; The first image in the hyperspectral image of the desulfurization bacterial solution is shown. Line number The pixels of the column are in the second reference band. The original reflectance at that location; This represents the first minimum constant to prevent division by zero.

3. The intelligent detection method for desulfurization bacterial activity parameters based on hyperspectral images according to claim 1, characterized in that, The spatial spectral gradient bias satisfies the expression: ； In the formula, The first image in the hyperspectral image of the desulfurization bacterial solution is shown. Line number Spatial spectral gradient bias of the pixels in the column; The first image in the hyperspectral image of the desulfurization bacterial solution represents the... Line number The spatial neighborhood centered on the pixel in the column; Representing spatial neighborhood The row and column coordinates of any neighboring pixel within the range; , These represent the first and second images of the desulfurization bacterial solution in the hyperspectral image. Line number Column pixels, neighboring pixels The depth of metabolic absorption; , These represent the first and second images of the desulfurization bacterial solution in the hyperspectral image. Line number Column pixels, neighboring pixels In the structural scattering band The original reflectance at that location; Represents an exponential function with the natural constant as its base; This represents the second minimum constant to prevent division by zero.

4. The intelligent detection method for desulfurization bacterial activity parameters based on hyperspectral images according to claim 1, characterized in that, The bacterial strain activity evaluation parameters satisfy the following expression: ； In the formula, Indicates parameters for evaluating the activity of bacterial strains; and These represent the height and width of the hyperspectral image of the desulfurization bacterial solution, respectively. The first image in the hyperspectral image of the desulfurization bacterial solution is shown. Line number The metabolic absorption depth of the pixels in the column; The first image in the hyperspectral image of the desulfurization bacterial solution is shown. Line number Spatial spectral gradient bias of the pixels in the column; The standard deviation of the spatial spectral gradient bias of all pixels in the hyperspectral image of the desulfurization bacterial solution; This represents the mean of the spatial spectral gradient bias of all pixels in the hyperspectral image of the desulfurization bacterial solution. Represents the natural constant; Represents a logarithmic function with the natural constant as its base; This represents the third minimum constant to prevent division by zero.

5. The intelligent detection method for desulfurization bacterial activity parameters based on hyperspectral images according to claim 1, characterized in that, Also includes: A comprehensive control threshold is generated based on the historical microbial activity evaluation parameters within a time sliding window. The microbial activity evaluation parameters at the current moment are compared with the comprehensive control threshold, and the corresponding process control instructions are output to the automatic control system based on the comparison results.

6. The intelligent detection method for desulfurization bacterial activity parameters based on hyperspectral images according to claim 5, characterized in that, The comprehensive control threshold generated based on historical microbial activity evaluation parameters within a time sliding window includes: Obtain the current hydraulic retention time of the sulfur-containing wastewater bioreactor, and establish a time sliding window with the current hydraulic retention time as the time span; extract all historically continuously output bacterial activity evaluation parameters within the time sliding window, obtain the mean and standard deviation of all bacterial activity evaluation parameters within the time sliding window, calculate the difference between the mean of the bacterial activity evaluation parameters and 3 times the standard deviation of the bacterial activity evaluation parameters, and use the difference as the adaptive activity dynamic threshold; extract the maximum value between the adaptive activity dynamic threshold and the preset absolute activity baseline threshold as the comprehensive control threshold.

7. The intelligent detection method for desulfurization bacterial activity parameters based on hyperspectral images according to claim 5, characterized in that, The step of outputting corresponding process control commands to the automatic control system based on the comparison results includes: If the bacterial activity evaluation parameter at the current moment is greater than or equal to the comprehensive control threshold, it is determined that the current desulfurization bacteria are in a stable and healthy metabolic state, and an instruction to maintain the current operating parameters is output to the automatic control system. If the bacterial activity evaluation parameter at the current moment is less than the comprehensive control threshold, it is determined that the desulfurization bacteria inactivation is abnormal in the current sulfur-containing wastewater bioreactor, a low activity warning signal is generated, and a process control instruction to increase the nutrient dosage and increase the opening frequency of the bottom sludge discharge valve is output to the automatic control system.

8. An intelligent detection system for desulfurization bacterial activity parameters based on hyperspectral imaging, characterized in that, include: The processor and memory, wherein the memory stores computer program instructions that, when executed by the processor, implement the intelligent detection method for desulfurization bacterial activity parameters based on hyperspectral images according to any one of claims 1-7.

Images