CN111896430A

CN111896430A - Pollen monitoring method and device

Info

Publication number: CN111896430A
Application number: CN202010011480.7A
Authority: CN
Inventors: 戎恺
Original assignee: Shanghai Kaiqing Intelligent Technology Co ltd
Current assignee: Shanghai Kaiqing Intelligent Technology Co ltd
Priority date: 2020-01-06
Filing date: 2020-01-06
Publication date: 2020-11-06

Abstract

The invention relates to the technical field of environmental monitoring, in particular to a pollen monitoring method and a pollen monitoring device; comprises air sampling; separating particles in the air according to size; fixing the sampled particles on a glass slide; carrying out microscopic imaging on the sampled particles; analyzing the image; storing the processed slide; promote monitoring efficiency by a wide margin, alleviate intensity of labour, promote the monitoring rate of accuracy, still make the sample can the repetition observation through fixed pollen granule in addition.

Description

Pollen monitoring method and device

Technical Field

The invention relates to the technical field of environmental monitoring, in particular to a pollen monitoring method and a pollen monitoring device.

Background

The conventional collection method at present mainly adopts a gravity sedimentation method to collect pollen, and then adopts a manual microscopic examination method to classify and count the pollen. The manual microscopic examination method is to carefully observe the pollen sample piece amplified by 400 times under a biological optical microscope line by an experienced observer, manually distinguish the pollen particle types, and manually count the number of different pollens and the total number of the pollens. The manual microscopic examination method mainly has the following defects:

1) the amount of labor of the observer is too large, so that physical fatigue is easily caused, and the monitoring accuracy is reduced;

2) the observation technique of the monitor is high in requirement, and long-term continuous technical training is needed to accurately master the microscopic morphological characteristics of different pollen particles so as to correctly identify the pollen. The cost of manpower and material resources is large;

3) the monitoring results are influenced by human factors, and the difference of the observation results obtained by different observers is large;

4) in the pollen propagation peak period, the manual resolution time is prolonged, so that eye fatigue of an observer is easily caused, and the pollen counting working efficiency is influenced.

In addition, the quality guarantee period of the samples prepared by microscopy is short, so that the samples cannot be stored, and the samples cannot be counted again if data is lost. In addition, the problem that the sample cannot be counted for multiple times so as to prevent counting errors and the like is also caused. Due to the defects in the technology, the monitoring and analyzing efficiency of the current manual microscopic examination method is not high, and the monitoring accuracy is difficult to ensure. In addition, because the online monitoring is carried out, the online monitoring is not available, the error of at least 24 hours from the peak time when the data of the pollen scattering peak time is released is avoided, and the real-time prediction cannot be realized, so that people can be reminded of preventing in advance.

Disclosure of Invention

In order to overcome the defects of the prior art, the invention provides a pollen monitoring method and a monitoring device applicable to the pollen monitoring method.

The technical scheme of the invention is as follows:

the pollen monitoring method comprises the following steps:

1) sampling air;

2) separating particles in the air according to size;

3) fixing the sampled particles on a glass slide;

4) carrying out microscopic imaging on the sampled particles;

5) image analysis (including classification and counting);

6) storing the processed slide (containing the analyzed particles);

7) and carrying out next round of sampling, and transmitting the analysis result to the server according to the set time interval.

Further, in step 1, a virtual impactor (also called as an impact sampler) is used for collecting surrounding air and impacting the air on a slide glass, the slide glass is coated with gelatin and glycerol, and pollen particles in the air are adhered by the glycerol, so that sampling is realized.

Further, in the step 2, the air passing through the periphery of the virtual impactor is 60m³The velocity of the/h collects the gas, the virtual impactor taking its central gas flow (about 6 m)³H) the virtual impactor first separates particles of different sizes in the air, and thenParticles of a size consistent with the particle size of the pollen are pushed onto a glass slide.

Furthermore, in the step 3, the gelatin is melted by heating, the pollen particles originally adhered by the glycerin are sunk into the gelatin, and then the pollen is permanently fixed in the gelatin by cooling, so that the subsequent processes are facilitated.

Further, in the step 4, the sample on the slide glass is scanned and photographed under an electron microscope, and an image stack is obtained and stored in a hard disk. Specifically, the method comprises the following steps:

the slide glass and the electron microscope can rotate to a certain degree to facilitate shooting of samples at multiple angles, the electron microscope selects 120 positions for shooting of the samples on each slide glass, and by means of image preprocessing (mainly including image transformation, image enhancement, edge detection and segmentation methods), shooting results are spliced again into image stacks which facilitate subsequent analysis and processing after being separated according to target particles and stored in a hard disk.

Further, in the step 5, the image stack obtained in the step 4 is exported from a hard disk for image analysis, and the image stack is further classified, identified and counted by pollen through big data and a machine learning algorithm. The method mainly comprises the following steps:

1) and (5) feature extraction. And (3) performing feature extraction on the image stack of the target particle (mainly comprising a pollen outline feature, a pollen structural feature and a pollen texture feature), judging that the particle is not pollen if a certain feature is unsuccessfully extracted, and forming a feature vector set (used for automatic classification of the next step) of each particle object by using the extracted parameter values if all the feature is successfully extracted.

2) Automatic (recognition) classification. The system uses a multi-stage classifier (comprising three sub-classifiers, namely an outer contour identification module, an inner structure identification module based on membership and a texture identification module) to classify the pollen particles of the target (determine the plant type corresponding to the pollen).

3) And counting. And repeating the steps to process all the shooting results of the electron microscope in a period of time to obtain the number of the pollen of different plant species.

Pollen monitoring devices, it includes rotating device, rotatory sample platform, slide glass and heating module, rotating device installs on a base, the vertical upwards setting of rotating device's pivot, the pivot top is connected with rotatory sample platform, be provided with a plurality of recesses on the rotatory sample platform, the recess is circular and its bottom is provided with the through-hole, detachably places the slide glass in the recess, be provided with the gelatin layer on the slide glass, be provided with the glycerine layer on the gelatin layer, still install heating module on the base, the heating terminal of heating module corresponds the recess setting of rotatory sample platform for the slide glass can be heated by the heating terminal of heating module when the recess rotates to the top of heating module. During the use, through equipment such as striking sample thief with pollen striking on the slide, the pollen is stuck by the glycerine layer this moment, then rotates rotatory sample platform for the slide in the recess is located heating module top, and heats. After heating, the pollen is sunk into the melted gelatin layer, and after cooling, the fixation of the pollen can be realized. And then rotating the rotary sample table, and observing the pollen sample by using an electron microscope.

In some embodiments, the rotating device employs an indexing disk. The angle of each rotation can be accurately controlled.

In some embodiments, the material of the rotary sample stage is one of a cast iron plate, a stainless steel plate or a high temperature resistant resin plate. Has enough hardness and is not afraid of high temperature.

In some embodiments, the heating module is a far infrared heater, and the heating direction of the heating module is vertically upward. As is known, the far infrared heater has a limited heating effect on an air medium, so most of heat can directly reach the glass slide, the heating effect is improved, the heat loss is reduced, and the air around the device cannot be greatly heated.

The invention has the beneficial effects that: promote monitoring efficiency by a wide margin, alleviate intensity of labour, promote the monitoring rate of accuracy, still make the sample can the repetition observation through fixed pollen granule in addition.

Drawings

FIG. 1 is a schematic diagram of the principles of the present invention;

FIGS. 2-3 are schematic diagrams of step 4;

FIG. 4 is a diagram of a multi-stage classifier.

Detailed Description

The following further describes embodiments of the present invention with reference to the accompanying drawings:

example 1

As shown in fig. 1-3, 60m from the ambient air³The velocity of the gas/h is such that the gas is collected, taking the central gas flow (about 6 m)³And h) entering a virtual impactor for later sampling analysis, wherein the virtual impactor firstly separates particles with different sizes in air, then pushes the particles which accord with the particle size of the pollen to a glass slide (the glass slide is covered with gelatin and glycerin), the virtual impactor continuously works according to set intervals until one sampling period is finished (the sampling period can be set, for example, 3 hours), a rotating sample table rotates to a heating module, the glass slide is heated to 90 ℃, the viscous surface on the glass slide is melted at high temperature, so that the particles are sunk into a medium, the position of the particles is fixed and adjusted, and later electron microscope observation focusing is facilitated. Then, the sample stage is rotated to enable the glass slide to rotate to a position above the electron microscope, then the sample is scanned and shot under the electron microscope, the glass slide and the electron microscope can rotate to a certain degree, so that the sample can be shot at a plurality of angles, and the shot results are combined to form a pollen image stack and stored in a hard disk. After the scan is complete, the handling system places the sample slides into the storage cassette. This cartridge has sufficient capacity to hold a full month of sample.

The computer system loads the image stack from the hard disk for image analysis. The pollen and other aerosol particles are separated, and the pollen type is identified according to the morphological characteristics of the pollen.

The sampling period can be set by the operator, typically 1-6 hours per day, with one slide being consumed per sampling period, and the slides being stored in a reusable storage cassette.

The electron microscope selected 120 positions of the specimen on each slide for photographing, and the photographed results were separated according to the target particles by image preprocessing and composed into an image stack. Then the image stack is analyzed by the image analysis module, a plurality of automatic classification condition combinations (including length, width, radius, number of gaps, texture and the like) are used in the analysis process, and finally the image stack and the analysis result are stored in a local hard disk together.

The image analysis processing flow is as follows:

1) and (5) image preprocessing. The method comprises the steps of shooting results of samples from a plurality of angles by an electron microscope, and splicing the shot results into image stacks convenient for subsequent analysis and processing after separating the shot results according to target particles by utilizing image preprocessing (mainly comprising image transformation, image enhancement, edge detection and segmentation methods).

2) And (5) feature extraction. And (3) performing feature extraction on the image stack of the target particle (mainly comprising a pollen outline feature, a pollen structural feature and a pollen texture feature), judging that the particle is not pollen if a certain feature is unsuccessfully extracted, and forming a feature vector set (used for automatic classification of the next step) of each particle object by using the extracted parameter values if all the feature is successfully extracted.

3) Automatic (recognition) classification. The system uses a multi-stage classifier (comprising three sub-classifiers, namely an outer contour identification module, an inner structure identification module based on membership and a texture identification module) to classify the pollen particles of the target (determine the plant type corresponding to the pollen).

4) And counting. And repeating the steps to process all the shooting results of the electron microscope in a period of time to obtain the number of the pollen of different plant species.

Remarking:

when the pollen plant species cannot be judged according to the automatic (identification) classification result, the images can be manually read for judgment in the later period, the judged plant species are input into the system, and then the pollen of the species can be automatically identified when the system encounters.

The following describes in detail the image preprocessing, feature extraction, and automatic (recognition) classification, which are key steps in the image analysis processing flow:

1. image pre-processing

1.1. Image transformation

The image transformation referred to herein is a two-dimensional orthogonal transformation, which plays an important role in image processing. The image transformation is mainly a further image processing consideration, such as the average value after fourier transformation is proportional to the average value of the image gray scale, and the high frequency components indicate the edge information of the image, and the characteristic value can be extracted from the image by utilizing the properties. The system designs common image transformation methods such as Fourier transformation, cosine transformation, KL transformation, wavelet transformation and the like, and uses proper transformation methods according to different image characteristics.

1.2. Image enhancement

The purpose of image enhancement is to use a series of techniques to improve the visual effect of the image, to increase the sharpness of the image, or to convert the image into a form more suitable for human eye observation and automatic machine analysis. If sharpening can highlight the edge contour of the image, the computer can be programmed to carry out tracking, and various feature analyses can be carried out. The plant pollen image enhancement method comprises the steps of trimming image gray level, smoothing image, sharpening image and the like.

1.3. Edge detection and segmentation

The edge of the image is the most basic feature of the image, and the edge can delineate a target object so as to be clear to an observer. The edge contains rich intrinsic information such as the direction, the step property, the shape and the like of the image, and is an important attribute for extracting image features in image recognition. Essentially, an image edge is a reflection of a discontinuity in the local characteristics of the image (abrupt change in gray level, abrupt change in color, etc.) that marks the end of one region and the beginning of another region. The edge extraction firstly detects the discontinuity of the local characteristics of the image and then connects the discontinuous edge pixels into a complete boundary. The edge is characterized by a gradual change of pixels along the edge, and a sharp change of pixels perpendicular to the edge.

Image segmentation is a processing technique that divides an image into several meaningful regions. "significant" herein broadly refers to "corresponding to a target" or "a function of a problem under study". For example, if a topographic aerial photograph or a topographic remote sensing image is input, it is necessary to segment a mountain area, a plain, a water area, a forest, a city, a road, and the like. When processing a plant pollen image, it is necessary to detect identification features such as holes and furrows of the pollen. These "objects" that are separated from the image domain are the objects of the image segmentation.

The basis of image segmentation is the similarity and the saltation between pixels. By "similarity" is meant that the pixels in a certain region have certain similar characteristics, such as consistent gray levels; by "jump" is meant a discontinuity in a property, such as an abrupt change in gray level value. There are various image segmentation methods, including a threshold method, a boundary detection method, a matching method, and a tracking method. In the system, the segmentation of the image is researched by adopting edge extraction based on pixel gray level mutation, a region generation method based on characteristic similarity and texture analysis.

2. Feature extraction

In order to create a system for identifying objects of different kinds, it is first necessary to determine which properties of the object should be measured as description parameters. These particular properties that are measured are called the features of the objects, and the resulting parameter values constitute a feature vector for each object. It is very important to select the features appropriately because it is the only basis when identifying the object.

In the image recognition problem, a small number of features with good distinguishability, reliability and independence are required to be extracted. The system selects 3 aspects of the pollen, such as outline, structure, texture and the like.

2.1. Pollen contour feature extraction

The pollen contour is circular, oval, triangular, pillow-shaped and petal-shaped, the gray level difference between the pollen and the background is large, and simultaneously due to noise formed by small impurities around the pollen, Gaussian filtering is used for amplitude filtering to remove the noise, the obtained contour is more regular, and edge tracking is carried out after binarization. The edge tracking obtains a closed curve, and then obtains the chain code of the curve. On the basis of the chain code, characteristic values such as perimeter, area and region roundness of the pollen outline are extracted, the distance from a point on the closed curve to the center of mass is calculated to obtain another curve, and the characteristics (such as whether a plurality of peaks are close to a straight line or not) of the curve are taken as the characteristics of the pollen outline.

2.2. Pollen structural feature extraction

Since pollen of the same family or genus may have different profiles, pollen of different families may have similar profiles, and the use of profile features alone is not sufficient. The next step requires finding structural features. Pollen structures are mainly represented by pores and grooves in the interior. The grooves are thin and long and the holes are eccentric. It would be an important feature if the information of the holes and grooves could be extracted and described. Because the holes and the grooves are concave in three-dimensional, the gray level abrupt change positions are shot under an electron microscope. The gradient operator is used to find its edges.

2.3. Pollen texture feature extraction

Texture reflects some change in color and intensity of the object surface, which in turn is related to the properties of the object itself. For example, the wood of the same species has the same or similar texture, and people can identify the species and the material of the wood by identifying the type of the wood grain. The pollen is smooth or wavy on the surface, and some pollen is provided with various sculptures such as spurs, tumors, particles, stripes, nets and the like to form various textures. These textures are important criteria for identifying pollen species.

In image recognition, texture is an attribute that reflects the spatial distribution of pixel gray levels in a region. Therefore, what is of interest when extracting texture features is a texture measure of an object in an image. If the gray level is a constant, or nearly constant, throughout the image, the object is not textured. If the gray level of the image changes significantly but not simply shade, then the object has texture. In order to measure texture, one tries to quantify the nature of the image gray level variation. The texture feature is a value calculated from an image of an object, which quantifies a characteristic of a gray level change of the image of the object. Firstly, the thickness and the fineness of the texture are judged by using a gray level co-occurrence matrix. And extracting statistical parameters such as energy, entropy, contrast, correlation and the like from the gray level co-occurrence matrix to be used as texture characteristic parameters, and respectively reflecting the uniformity of the gray level distribution of the image, the information content of the image, the definition of the image and the similarity degree of gray level co-occurrence matrix elements in the row direction or the column direction.

3. Automatic (recognition) classification

3.1. Principle of multi-stage classifier

As shown in fig. 4, the classifier can be viewed as a "machine" consisting of hardware or software. Its function is to calculate c discriminant functions gi first, and then select the class corresponding to the discriminant function as the maximum value as the decision result.

A multi-stage classifier is constructed in the present system. The multi-stage classifier applies different identification technologies to pollen images with different characteristics, and makes full use of the advantages of different identification methods, so that the performance of the multi-stage classifier is superior to that of a single-stage identification system.

In recent years, the combination of multiple classifiers has been the leading research topic in the field of pattern recognition, and has achieved better application effects in many applications of pattern recognition, such as character recognition, object recognition, and the like. Many approaches have been proposed for multi-classifier combinations. The common methods include a majority decision method, a linear weighting method, Bayesian estimation and the like.

Aiming at the particularity of the system, a dynamic weight method for multi-classifier result combination is designed. The method does not fix the weight of each classifier, and outputs a candidate set sequence and corresponding credibility. The confidence level of a candidate set refers to the degree of reliability with which the input pattern belongs to this pattern (or classification). If the confidence of the classification result a is much higher than that of other classification results in the classifier x, a higher weight is assigned to the classifier, i.e., for a single classifier x, the higher the separation of the confidence of its classification results, the higher the weight of the classifier is. The benefits of this algorithm are: when the classification result can be basically determined by a single classifier, the classifier is dominant in the multi-classifier combination. When the separation of the credibility of the classification results of all the classifiers is similar, each classifier plays a role in approximate decision making.

3.2. Multi-stage classifier implementation

Consists of three sub-identification modules (sub-classifiers):

(1) outer contour recognition module

(2) Internal structure identification module based on membership degree

(3) Texture recognition module

The correlation between classifiers at different levels can be divided into two cases: 1. the same aspects of the same object are described to a different degree only. 2, different aspects of the same object are described. Case 1 reflects redundant information of the classifier and case 2 reflects complementary information of the classifier. In the fusion of sensor data, which is a low-level information fusion, the situation that a certain number of sensors fail due to the influence of strong noise needs to be considered, and redundant information can be utilized to maintain the stability of a fusion system. In the fusion of classifiers, which is a high layer of information fusion, redundant information leads to redundant estimation of the combined result, so that the redundant information needs to be removed.

Complementary information, while describing different aspects of the same object, is not beneficial to the combination of classifiers, only complementary information that improves the effectiveness of the combination is useful. The key of the classifier combination is how to make up for the deficiencies of each classifier, so that the combination effect is better than that of each individual classifier. The partitioning of the three sub-classifiers of the present system is also based on the consideration that complementarity is maximized and redundancy is minimized.

3.2.1. Outer contour recognition module

And extracting the outer contour features, including the perimeter, the area, the region roundness, the number of angle points, the number of wave peaks of the centroid profile and the variance of the distance from the centroid. First, the sample images can be classified into two categories according to the region circularity: round and non-round. When the outer edge is roughly judged to be a circle, whether the edge curve is regular or not can be judged according to the variance of the section curve of the centroid. If the pattern is non-circular, the number of sides of the pollen image can be roughly determined by the number of peaks of the area centroid cross-section curve. When the number of peaks is 2, there may be two cases: oval and pillow-shaped. Since they represent two broad classes of pollen, it is necessary to distinguish them. Experiments prove that the variance of the area centroid profile curve has a good classification effect when the two types are distinguished. If the variance is large, the pillow shape can be considered. The variance is small and is elliptical. However, the ellipse is an ellipse whose major axis is largely different from its minor axis because a circle-like ellipse whose major and minor axes are close to each other has been put into a circle. When the number of wave peaks is larger than or equal to 3, the outline shape of the pollen image is not easy to determine, and the number of nodes on the edge is considered. The method has the advantages that the edges are smooth when the number of the angular points is small, and the number of wave crests is used as the number of edges of the outer contour; when the number of the angular points is more, the shape is judged to be irregular.

One of the more difficult points is the determination of the threshold. Such as the roundness of the region, the variance of the cross-sectional curve of the centroid, the number of the angle points, etc. In the system, the adopted method is mainly determined through empirical values, and the classification interval is determined while determining the threshold values. The confidence that the representative sample belongs to each possible classification is determined by the distance between these thresholds and the characteristic values of the representative sample.

3.2.2. Internal structure identification module based on membership degree

(1) Degree of membership and membership function

The concept of no explicit extension is called fuzzy concept and the totality of objects in question is called domain or space. Fuzzy concepts are not extended by common sets, and elements in a domain of discourse are not absolutely 0 or 1 to the extent they conform to the concept, which may be between 0 and 1. In fuzzy mathematics, the absolute membership of elements to common sets is activated and the concept of membership is proposed. The degree of membership may be described by a membership function.

Fuzzy set a on any domain S ═ { X } refers to a well-defined whole of elements of some nature in X, and a membership function F can be used_A(x) To characterize, F_A(x) The value of (a) reflects the membership degree of x to the fuzzy set A, and the value of x is in a closed interval [0, 1 ]]In (1). If F_A(x) A value of "1" indicates that x is highly dependent on A; if F_A(x) A value of (a) close to 0 indicates that the degree to which x depends on a is low. For the domain, the domain elements are always distinct, and only the subset A, B in S is ambiguous, so the fuzzy set is usually referred to as a fuzzy subset. In the situation of not easy to be confused, the fuzzy subset is simply called fuzzy set. For example, S is a number domain, and if the fuzzy set a represents a real number far greater than 0, i.e., a ═ x > 0, then a is subject to membershipThe function can be written as:

(2) internal structure classifier design

The method of directly determining the membership of the sample by calculating the membership degree of the sample is called a membership principle of pattern classification, and is also called a direct method of fuzzy pattern classification.

Membership principle:

there are n fuzzy subsets in the set domain U

And each pair of

All have membership functions

Then consider x₀Belonging to A_i。

The membership principle is obvious and easy to recognize, but how effective it is depends on the skill of establishing membership functions of known pattern classes.

In the present system, Gaussian-type membership functions are used

The main reason is that this function has several better characteristics: the expression form is simple, and the complexity is not increased too much even for the multi-input variable. ② the product is radially symmetrical and accords with the property of common objective things. Third, nature-checking.

Obtaining the clustering center m of each category by clustering analysis_iAnd standard deviation σ_iObtaining membership functions of each class

Extracting characteristic vector of sample to be detected

Substituting into membership functions of each class. And obtaining the category of the maximum membership degree as the category of the sample to be detected according to the membership principle. In this classifier, the confidence level can be simply considered as a degree of membership.

3.2.3. Texture recognition module

(1) Texture feature selection by entropy minimization

Entropy (Entropy) is a statistical measure of uncertainty. For a given population of mode vectors, the intra-class dispersion is measured by the overall entropy, i.e.

H＝-E_p{lnp(x)}

Where p (x) is the probability density of the population of modes, E_pIs the desired operation for p (x). Features that reduce uncertainty are considered to have a greater amount of information when considering the best feature selection. Therefore, when entropy is used as the uncertainty measure, selecting features that minimize the entropy of the pattern class is a reasonable feature selection criterion.

Consider that there are M pattern classes with probability densities of p (x | ω), respectively₁)，p(x|ω₂)，…，p(x|ω_M) Then the entropy of the ith pattern class is

H_i＝-∫_xp(x|ω_i)ln[p(x|ω_i)]dx

The integral domain of which is the entire mode space. Obviously, if p (x | ω [ ])_i) When 1, then H_iAt this point there is no uncertainty. It follows that the smaller the entropy, the smaller the uncertainty, and the better the separability.

In the texture feature extraction, various texture information features are obtained by a gray level co-occurrence matrix and a Fourier transform texture analysis method. Due to the fact that redundancy among features is large, entropy of the whole mode class is large, and separability is poor. And the dimensionality of the feature vector is high, and the computational complexity is greatly increased. Therefore, the dimension of the feature vector must be reduced on the basis of the principle of entropy reduction.

(2) Detailed description of the invention

In the system, the same type of characteristics are firstly classified into one type, and a characteristic quantity describing each type of characteristics is obtained through a weighted average method according to the gradeability of each characteristic. In this way, redundancy between like features is reduced. Experiments prove that the entropy value of the mode class after feature selection is greatly reduced compared with that before feature selection.

Texture roughness is described by: an angular second-order distance f1 extracted from the gray level co-occurrence matrix; gray difference probability P (d) and entropy f₄. R obtained by Fourier transform_max. According to the value of each characteristic, estimating the roughness expressed by each characteristic, and then weighting and averaging to obtain uniform roughness C₀. Due to the angular second order distance f1 and r of the Fourier transform_maxIs better, a larger weight is assigned. The grayscale difference probability p (d) entropy f4 assigns a small weight.

Because the similar pollen images have rotation in different directions, the system extracts the strong and weak characteristics of the pollen texture direction rather than the specific direction emphatically when extracting the texture direction characteristics. There are three features employed in the system: MAX P_i(r) shows the Ring characteristic P_i(r) whether the upper peak is significant, the number of peaks N and the direction characteristics DIR extracted from the power spectrum matrix. Mainly composed of MAX P_i(r) it was decided that N and DIR served as assistance. Obtaining the directional strength I_d。

In experiments, some other texture characteristics are found to play a good role in classifying pollen textures:

contrast of gray level co-occurrence matrix

Reflecting the clarity of the texture.

Correlation of gray level co-occurrence matrix

The degree of similarity of the elements of the gray level co-occurrence matrix in the row direction or the column direction is reflected.

After the characteristics are selected, the characteristics of each pollen in the pollen image library are averaged, the characteristic value of the sample to be detected is compared with the average value, and the sample to be detected is classified into the pollen class with the minimum distance. The distance is normalized as the confidence level for each class.

Example 2

Use pollen monitoring devices, it includes rotating device, rotatory sample platform, slide glass and heating module, rotating device installs on a base, the vertical upwards setting of rotating device's pivot, the pivot top is connected with rotatory sample platform, be provided with a plurality of recesses on the rotatory sample platform, the recess is circular and its bottom is provided with the through-hole, detachably places the slide glass in the recess, be provided with the gelatin layer on the slide glass, be provided with the glycerine layer on the gelatin layer, still install heating module on the base, the heating terminal of heating module corresponds the recess setting of rotatory sample platform for the slide glass can be heated by the heating terminal of heating module when the recess rotates the top of heating module. During the use, through equipment such as striking sample thief with pollen striking on the slide, the pollen is stuck by the glycerine layer this moment, then rotates rotatory sample platform for the slide in the recess is located heating module top, and heats. After heating, the pollen is sunk into the melted gelatin layer, and after cooling, the fixation of the pollen can be realized. And then rotating the rotary sample table, and observing the pollen sample by using an electron microscope.

The rotating device adopts an index plate. The angle of each rotation can be accurately controlled.

The material of the rotary sample table is one of a cast iron plate, a stainless steel plate or a high-temperature-resistant resin plate. Has enough hardness and is not afraid of high temperature.

The heating module is a far infrared heater, and the heating direction of the heating module is vertical upwards. As is known, the far infrared heater has a limited heating effect on an air medium, so most of heat can directly reach the glass slide, the heating effect is improved, the heat loss is reduced, and the air around the device cannot be greatly heated.

One of the use modes is as follows:

60m from the surrounding air by means of impact collectors³Velocity collection/hThe gas thus impinges pollen on the slide (of course, pollen may not be present in the air, and thus on the slide), and the collection cycle should be fixed, for example, 3 hours.

The rotating sample table is rotated to the heating module, the glass slide is heated to 90 ℃, the viscous surface on the glass slide is melted at high temperature, so that the particles are sunk into the medium, the position of the particles is fixed and adjusted, and observation and focusing of a later electron microscope are facilitated. The rotation can be manually controlled or can be automatically controlled by control equipment such as a PLC (programmable logic controller) and the like (a user installs the heating device by himself), and the heating temperature needs the user to obtain the optimal heating module parameters by himself through limited experiments.

The heated slide glass of the rotary sample stage is rotated to reach the observation position after heating, and observation and recording can be carried out through an electron microscope at the moment. After the recording is completed, the slide with the pollen sample fixed thereon can be taken out and put into a storage box. And then carrying out graph stack analysis, thereby realizing pollen classification.

The foregoing embodiments and description have been presented only to illustrate the principles and preferred embodiments of the invention, and various changes and modifications may be made therein without departing from the spirit and scope of the invention as hereinafter claimed.

Claims

1. The pollen monitoring method comprises the following steps:

1) sampling air;

2) separating particles in the air according to size;

3) fixing the sampled particles on a glass slide;

4) carrying out microscopic imaging on the sampled particles;

5) analyzing the image;

6) storing the processed slide (containing the analyzed particles);

2. The pollen monitoring method as claimed in claim 1, wherein: in the step 1, a virtual impactor is used for collecting surrounding air and enabling the air to impact a glass slide, gelatin and glycerol are coated on the glass slide, and pollen particles in the air are adhered by the glycerol, so that sampling is achieved.

3. The pollen monitoring method as claimed in claim 2, wherein: in the step 2, the air passing through the periphery of the virtual impactor is 60m³The speed of/h collects the gas, the central air flow is taken by the virtual impactor, the virtual impactor first separates particles of different sizes in the air, and then pushes the particles according with the particle size of the pollen to the glass slide.

4. The pollen monitoring method as claimed in claim 3, wherein: in the step 3, the gelatin is melted by heating, the pollen particles originally adhered by the glycerin are sunk into the gelatin, and then the pollen is permanently fixed in the gelatin by cooling.

5. The pollen monitoring method as claimed in claim 4, wherein: scanning and shooting a sample on the glass slide by using an electron microscope to obtain an image stack and storing the image stack in a hard disk;

the glass slides and the electron microscope can rotate to a certain degree so as to be convenient for shooting samples at multiple angles, the electron microscope shoots the samples on each glass slide at 120 positions, the shooting results are separated again and spliced into image stacks according to target particles by using image preprocessing, and then the image stacks are stored in a hard disk.

6. The pollen monitoring method as claimed in claim 5, wherein: and 5, exporting the image stack obtained in the step 4 from a hard disk for image analysis, further carrying out pollen classification identification on the image stack through big data and a machine learning algorithm, and realizing pollen classification statistical counting on the basis.

7. Pollen monitoring devices, its characterized in that: it includes rotating device, rotatory sample platform, slide glass and heating module, rotating device installs on a base, rotating device's the vertical upwards setting of pivot, the pivot top is connected with rotatory sample platform, rotatory sample bench is provided with a plurality of recesses, the recess is circular and its bottom is provided with the through-hole, detachably places slide glass in the recess, be provided with the gelatin layer on the slide glass, be provided with the glycerine layer on the gelatin layer, still install heating module on the base, the heating end of heating module corresponds the recess setting of rotatory sample platform for slide glass can be heated by heating module's heating end when the recess rotates the top to heating module.

8. The pollen monitoring device of claim 7, wherein: the rotating device adopts an index plate.

9. The pollen monitoring device of claim 8, wherein: the heating module is a far infrared heater, and the heating direction of the heating module is vertical upwards.