CN-122016584-A - On-line detection and distribution positioning method for airborne particulate matters without lens diffraction

CN122016584ACN 122016584 ACN122016584 ACN 122016584ACN-122016584-A

Abstract

The invention discloses an airborne particulate lens-free diffraction online detection and distribution positioning method, which belongs to the technical field of intelligent environment monitoring, and comprises the steps of collecting aerosol diffraction images and aerial image data, constructing a diffraction spectrum weather space-time alignment set through time stamp alignment, extracting spectrum energy distribution characteristics and annular characteristic input depth inversion models, introducing humidity, wind speed and temperature, combining preset threshold correction weights, inverting equivalent diameter distribution and concentration of particulate matters, generating a particulate matter basic parameter set, deducing a high probability source area based on concentration gradients and weather wind field directions, generating an initial source area estimation map, combining constraint conditions, constructing a constraint type reverse track tracing model, deducing a particulate matter propagation path, constructing a joint loss function iteration optimization path initial weight and source area intensity distribution by adopting a concentration and position error feedback mechanism, correcting the propagation path and positioning, outputting a particulate matter online detection result and a high resolution space distribution positioning map, and realizing accurate tracing and monitoring.

Inventors

PAN XIAOQING
SHAN HAOSHU
YANG NING
WEI XING
ZUO XIN
WANG YING

Assignees

江苏省农业科学院
镇江市动物疫病预防控制中心

Dates

Publication Date: 20260512
Application Date: 20260414

Claims (10)

1. The method for on-line detection and distribution positioning of airborne particulate matters without lens diffraction is characterized by comprising the following steps: collecting aerosol diffraction images and meteorological data, preprocessing, aligning according to a uniform time stamp, and constructing a diffraction spectrum meteorological space-time alignment set; Extracting spectral energy distribution characteristics and annular characteristics, inputting the spectral energy distribution characteristics and the annular characteristics into a depth inversion model, introducing humidity, wind speed and temperature in meteorological data, correcting weights by combining preset thresholds, inverting equivalent diameter distribution and concentration of particles, and outputting a particle basic parameter set; based on the initial source region estimation diagram and the particulate matter basic parameter set, introducing constraint conditions, constructing a constraint type reverse track tracing model, and deducing a particulate matter propagation path by adopting a Lagrange reverse track method; And correcting the propagation path and positioning distribution by adopting a concentration and position double-error feedback mechanism, constructing a joint loss function, reversely adjusting path weight and source region intensity distribution, and iteratively optimizing to generate a particulate matter online detection result and a high-resolution space distribution positioning map.
2. The method for on-line detection and distribution positioning of airborne particulate lens-free diffraction of claim 1, wherein the specific step of inverting the equivalent diameter distribution and concentration of the particulate comprises: acquiring spectrum energy distribution characteristics, and generating a spectrum characteristic vector by a local maximum method; extracting a diffraction ring intensity curve, generating an annular feature vector by combining Hough circle transformation, splicing the annular feature vector with the spectrum feature vector, and weighting and calculating an enhanced image feature vector by combining linear transformation; Normalizing the meteorological data into meteorological feature vectors, and calculating calibration weights by a cross attention mechanism with the meteorological feature vectors as a query matrix; introducing humidity, wind speed and temperature in meteorological conditions, and correcting the calibration weight by combining a humidity threshold value, a wind speed threshold value and a temperature threshold value; Multiplying the corrected calibration weight by the enhanced image feature vector to obtain weather correction image features, and generating a depth fusion feature vector by combining the original weather features; And inputting the depth fusion feature vector into a depth inversion model, and outputting a preliminary inversion result of the equivalent diameter distribution PSD of the particulate matters and the particulate matter mass concentration MC through feature dimension reduction and nonlinear transformation.
3. The method for on-line detection and distribution positioning of airborne particulate lensless diffraction of claim 2, wherein the specific step of inverting the equivalent diameter distribution and concentration of particulate further comprises: setting a particle size threshold range and a mass concentration threshold of equivalent diameter distribution; constructing a particle size constraint rule, if the PSD peak particle size exceeds a particle size threshold range, judging that the particle size constraint condition is not met, and marking the abnormal particle size, otherwise, judging that the particle size constraint condition is met, and marking the normal particle size; Constructing a concentration constraint rule, if MC is larger than a mass concentration threshold, judging that the concentration constraint rule is not satisfied and marking the concentration abnormality; if the particle size constraint rule and the concentration constraint rule are met at the same time, judging that the preliminary inversion result is reasonable, otherwise, judging that the preliminary inversion result is unreasonable; And verifying all detection points through time sequence and space grid inversion, integrating effective inversion results, and generating a particulate matter basic parameter set.
4. The method for on-line detection and distribution positioning of airborne particulate lens-free diffraction of claim 3, wherein the specific step of generating an initial source region estimate map comprises: based on the particulate matter basic parameter set, generating a concentration grid by spatially interpolating particulate matter concentration data, generating a wind field grid by spatially interpolating wind field data, and spatially aligning the wind field grid with the concentration grid; Extracting the duty ratio of fine particles and coarse particles of each grid unit, and setting a first threshold value and a second threshold value; For the fine particles, if the ratio of the fine particles exceeds a first threshold, judging that the fine particles have dominant positions; For coarse particles, if the ratio of coarse particles exceeds a second threshold, judging that the coarse particles have dominant positions, otherwise, judging that the coarse particles have no dominant positions; If only one dominant position exists in each grid unit, judging that the particle dominant type exists, wherein the particle dominant type comprises fine particle dominant and coarse particle dominant, and marking the particle dominant type as a target grid; Otherwise, judging that the particle leading type does not exist, and marking the particle leading type as a non-target grid; And calculating the concentration difference value and the actual space distance between the target grid and each neighborhood grid, dividing the concentration difference value by the actual space distance to obtain a concentration gradient amplitude value, and determining the gradient direction according to the positive and negative of the concentration difference value.
5. The method for on-line detection and distribution positioning of airborne particulate lens-free diffraction of claim 4, wherein the specific step of generating an initial source region estimate map further comprises: The direction opposite to the gradient is selected, calculating included angle by combining wind field and wind direction ; If it is Judging that the wind field is consistent with the gradient direction, and setting a first cooperative weight; If it is Judging that the direction of the wind field is consistent with the direction of the gradient part, and setting a second cooperative weight; If it is Judging that the wind field collides with the gradient direction, and setting a third cooperative weight; Combining wind field direction weighting correction to obtain cooperative directional vectors, dividing candidate source units, counting the quantity and intensity of the cooperative directional vectors, and calculating the product of the quantity proportion and the intensity proportion to obtain convergence; Setting a convergence threshold, and reserving candidate source units exceeding the convergence threshold to form a primary screening high-probability source area; Calculating theoretical distances from the high probability source region of the primary screening to the high-concentration grid, and eliminating regions exceeding the theoretical distances; When the primary screening high-probability source region with coarse particle dominant and fine particle dominant is spatially overlapped, taking the overlapped part as a core source region, calculating the comprehensive source probability, and generating an initial source region estimation graph by combining geographic information.
6. The method for on-line detection and distribution positioning of airborne particulate matters without lenses according to claim 5, wherein the specific steps of constructing a constraint type reverse track traceability model comprise: Introducing constraint conditions, and constructing a constraint type reverse track tracing model, wherein the constraint type reverse track tracing model comprises an input layer, a core calculation layer, a constraint correction layer and an output layer; An input layer receives the initial source region estimation graph and the particulate matter basic parameter set; The core calculation layer takes the detection point as an initial position and calculates the space-time coordinates of the track foundation by combining the reverse wind speed vector and the time step; The constraint correction layer converts constraint conditions into a precipitation removal function, a radiation attenuation function, a temperature and humidity sedimentation function and a wind speed guide function based on the track basic space-time coordinates, and sequentially corrects the constraint conditions by combining with a constraint priority controller to obtain corrected track basic space-time coordinates; The output layer converts the corrected track basic space-time coordinates into a multi-dimensional result; And calling historical particulate matter tracing cases, constructing a training set, and calibrating key parameters of the constraint correction layer by taking the minimum error of the multidimensional result, the real source region coordinates and the contribution degree as a target to obtain an optimized constraint type reverse track tracing model.
7. The method for on-line detection and distribution positioning of airborne particulate lens-free diffraction of claim 6, wherein the step of deriving the propagation path of the particulate comprises: setting a reverse deduction period, backtracking from the sampling time, setting an initial height by taking a detection point as an end point and combining the particle size and the particle density, and setting a first-level weight for each track; Moving the particles along the reverse direction of the wind speed to form an initial track cluster; Counting the precipitation amount based on the precipitation clearing function, and setting a medium precipitation threshold and a strong precipitation threshold to divide a strong clearing area and a weak clearing area; dividing the particles with strong hygroscopicity, the particles with weak hygroscopicity and the biological particles by combining the basic parameter set of the particles, and matching the clearance rates of the strong clearance area and the weak clearance area, and removing the tracks affected by precipitation from the initial track cluster to form a first-stage correction track cluster; Dividing the solar radiation into a strong inactivation area and a weak inactivation area according to the intensity and duration of the solar radiation based on the radiation attenuation function; Dividing the radiation resistance type by combining the radiation resistance performance of the particles, matching the radiation attenuation coefficient from a preset lookup table, and calculating a secondary weight by combining the primary weight; and if the secondary weight is lower than a preset weight threshold, removing the track, otherwise, reserving the track to form a secondary correction track cluster.
8. The method for on-line detection and distribution positioning of airborne particulate lens-free diffraction of claim 7, further comprising the specific step of deriving a particulate propagation path comprising: Calculating a sedimentation speed based on the temperature and humidity sedimentation function and combining the particle size and the particle density; Extracting the current track node height, and calculating the corrected track node height by combining the sedimentation speed and the backtracking time length; calling the range of an atmospheric boundary layer, and adjusting the secondary weight according to the height of the corrected track node to obtain a tertiary weight; Calculating the deviation of the direction and the size of the node based on the wind speed guiding function, dynamically setting a deviation threshold value, and counting the proportion of abnormal nodes; if the abnormal node duty ratio is lower than a preset duty ratio threshold value, correcting the track direction, otherwise, eliminating the corresponding track to obtain a four-level correction track cluster; setting an initial high-weight source region, calculating an included angle with the track direction, adjusting the three-level weights to obtain four-level weights, clustering the density of reverse track nodes, calculating the total weight of each type of nodes, screening a central source region in a descending order, and calculating the overlapping rate with the initial source region estimation graph; when the overlapping rate is lower than a preset overlapping rate threshold value, correcting to obtain a high-resolution source region thermodynamic diagram, and screening the thermodynamic diagram before screening The class node trajectory is a propagation path.
9. The method for on-line detection and distribution positioning of airborne particulate lens-free diffraction of claim 8, wherein the specific step of generating a high resolution spatial distribution positioning map comprises: Dividing propagation path nodes into source region nodes and non-source region nodes based on high-resolution source region thermodynamic diagram boundaries; counting the residence time, acquiring the initial source area intensity, calculating the total weight of the source area node and the non-source area node, and summing to obtain the path initial weight; setting a transmission coefficient, calculating a predicted concentration based on the path initial weight, and calculating a concentration error by combining the actual concentration; Calculating the deviation of the source region node according to the average Euclidean distance between the source region node and the center of the high-resolution source region thermodynamic diagram region, calculating the deviation of the end point according to the Euclidean distance between the theoretical propagation path end point and the actual propagation path end point, and weighting to calculate the position error; and dynamically adjusting the error reliability by combining the actual concentration and the fluctuation amplitude of the wind speed to generate an error reliability matrix.
10. The method for on-line detection and distribution positioning of airborne particulate lensless diffraction of claim 9, wherein the specific step of generating a high resolution spatial distribution positioning map further comprises: Constructing a joint loss function to calculate joint loss based on the error credibility matrix, and optimizing the initial weight of the path by adopting a gradient descent method to obtain an optimized path weight; Adjusting the source region intensity according to the concentration error to obtain an optimized source region intensity, calculating an optimized predicted concentration by combining a transmission coefficient, and weighting to calculate a corrected detection value to generate an online particulate matter detection result; Calculating an optimized concentration value and spatial gradient distribution based on the optimized source region intensity and in combination with a diffusion rule; Introducing the optimized concentration value into the high-resolution source region thermodynamic diagram to generate a three-dimensional coordinate diagram, adding the propagation path corresponding to the optimized path weight and the spatial gradient distribution, and generating a high-resolution spatial distribution positioning diagram.

Description

On-line detection and distribution positioning method for airborne particulate matters without lens diffraction Technical Field The invention relates to a lens-free diffraction online detection and distribution positioning method for airborne particles, and belongs to the technical field of intelligent environment monitoring. Background In the breeding house, airborne particles can carry harmful microorganisms, ammonia and other pollutants, so that not only the breeding ecological environment is destroyed, but also respiratory mucosa of livestock can be stimulated, and the immunity and the production performance of the livestock are reduced. The method for accurately detecting and positioning the airborne particulate matters in the cultivation house is of great importance to tracing pollution sources, evaluating cultivation environment risks and formulating a scientific house environment treatment strategy. The traditional particle detection method has the limitations of low resolution, low detection speed and difficulty in real-time positioning under a complex cultivation environment, and the lens-free diffraction imaging technology becomes a research hot spot for realizing rapid, online and real-time detection of airborne particles in a cultivation house due to the technical advantages of high resolution, wide view field, non-marking and high flux. The prior Chinese patent application with publication number of CN119845904A discloses a diffraction optics-based virus infection positioning detection method and system, which uses a diffraction imaging principle to collect diffraction fingerprints corresponding to all cells to be detected in an imaging area, uses a gray level symbiotic matrix method to extract texture characteristics of each diffraction fingerprint in the imaging area, compares the extracted texture characteristics with standard texture characteristics pre-stored in a relation model of cell morphological change and diffraction fingerprints, thereby obtaining the cell state of each cell to be detected in the imaging area, and positioning and identifying the virus infection state of each cell in the imaging area. The high-efficiency optical diffraction spectrum system is used for detecting virus infection, and the requirements of professionals and expensive and complex professional field detection equipment are eliminated. The method meets the continuous and high-throughput evaluation of the cell condition after virus infection, and lays a foundation for popularization and application of a virus infection instant detection platform. Although the existing virus infection detection technology based on diffraction optics can realize virus infection positioning at the cell level, the technology is limited to a microscopic imaging area, meteorological data are not fused to track the track of particles in the atmosphere, and precise positioning of the particles such as airborne viruses in a complex space is difficult to realize. Disclosure of Invention Aiming at the defects of the prior art, the invention aims to provide the lens-free diffraction on-line detection and distribution positioning method for the airborne particles, which realizes the precise inversion of the equivalent diameter distribution and concentration of the particles, the efficient positioning of a source region and the clear tracing of a path through the space-time alignment fusion, the dynamic weight correction, the multi-constraint reverse track tracing and the double-error optimization of the aerosol diffraction image and the meteorological data, and improves the reliability and the practicability of the on-line detection and the distribution positioning of the airborne particles. In order to achieve the above purpose, the present invention provides the following technical solutions: The method for on-line detection and distribution positioning of airborne particulate matters without lens diffraction comprises the following steps: collecting aerosol diffraction images and meteorological data, preprocessing, aligning according to a uniform time stamp, and constructing a diffraction spectrum meteorological space-time alignment set; Extracting spectral energy distribution characteristics and annular characteristics, inputting the spectral energy distribution characteristics and the annular characteristics into a depth inversion model, introducing humidity, wind speed and temperature in meteorological data, correcting weights by combining preset thresholds, inverting equivalent diameter distribution and concentration of particles, and outputting a particle basic parameter set; based on the initial source region estimation diagram and the particulate matter basic parameter set, introducing constraint conditions, constructing a constraint type reverse track tracing model, and deducing a particulate matter propagation path by adopting a Lagrange reverse track method; And correcting the propagation path and positioning distribution by