A raman spectrum-based hazardous chemical data acquisition and detection method

Abstract

The present application relates to the technical field of chemical detection, in particular to a dangerous chemical data acquisition and detection method based on Raman spectrum. The method comprises the following steps: obtaining target noise degree values of each spectrum data point on a peak signal segment and relative deviation degree values of each spectrum data point on a baseline signal segment, distributing initial weights of the spectrum data points on the original Raman spectrum signal according to the target noise degree values and the relative deviation degree values to obtain initial weights of each spectrum data point on the original Raman spectrum signal, processing the original Raman spectrum signal according to the initial weights and the asymmetric least squares method to obtain a target Raman spectrum signal, and detecting the chemical to be detected according to the target Raman spectrum signal. The present application can improve or ensure the detection effect of the dangerous chemicals.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of chemical detection, and in particular to a dangerous chemical data acquisition and detection method based on Raman spectrum. BACKGROUND

[0002] With the continuous development of industrial production, energy transportation and scientific research activities, the types and usage of dangerous chemicals continue to increase. The flammable, explosive, toxic and strong corrosive characteristics of dangerous chemicals make it easy to cause major safety accidents once they are leaked or improperly contacted. Therefore, rapid, accurate and non-destructive detection of dangerous chemicals is crucial for public safety. Currently, in order to achieve the purpose of rapid, accurate and non-destructive detection of dangerous chemicals, Raman spectrum technology is used. Raman spectrum technology is a spectral analysis method based on the change of molecular vibration and rotation energy level, which has the advantages of fast detection speed, no need for complex pretreatment, in-situ operation and wide application range, and can effectively identify and detect various substances such as liquids, solids and gases. However, when collecting data based on Raman spectrum technology for dangerous chemicals, the Raman scattering signal itself is weak, which makes the collected signal prone to random noise and baseline drift problems due to the influence of environmental light sources, stray light, fluorescence interference, high temperature, high humidity, dust and other complex conditions, resulting in low quality of the collected Raman spectrum signal, which further affects the detection effect of dangerous chemicals, such as false alarms of equipment or missed reports of risk substances.

[0003] In order to reduce or avoid the influence of random noise and baseline drift on chemical detection effect in the prior art, the collected Raman spectrum signal is usually processed using asymmetric least squares method. However, when the asymmetric least squares method is used to process the collected Raman spectrum signal in the prior art, the initial weight is generally set as a fixed value or an empirical value. However, this method of determining the initial weight does not consider the spectral characteristics, which may cause insufficient differentiation between true peaks and noise in the initial fitting stage, resulting in baseline estimation deviation or loss of effective information, and further affecting the detection effect of chemicals based on the processed data. SUMMARY

[0004] To solve the above problems, the present application provides a dangerous chemical data acquisition and detection method based on Raman spectrum, and the technical solution is as follows:

[0005] One embodiment of the present application provides a dangerous chemical data acquisition and detection method based on Raman spectrum, which includes the following steps:

[0006] obtaining an original Raman spectrum signal of a chemical to be detected and peak points on the original Raman spectrum signal;

[0007] obtaining a target similarity sequence of the peak points according to amplitude differences between spectrum data points located on both sides of the peak points and adjacent spectrum data points, and obtaining a baseline signal segment and a peak signal segment according to the target similarity sequence;

[0008] obtaining target noise degree values of spectrum data points on the peak signal segment according to the peak points and end points on the peak signal segment, the target similarity sequence, and peak point differences between the peak signal segment and other peak signal segments;

[0009] obtaining relative deviation degree values of spectrum data points on the baseline signal segment according to differences between amplitudes of the spectrum data points on the baseline signal segment and an overall representative amplitude of the baseline signal segment;

[0010] performing initial weight distribution on spectrum data points on the original Raman spectrum signal according to the target noise degree values and the relative deviation degree values, obtaining initial weights of the spectrum data points on the original Raman spectrum signal, processing the original Raman spectrum signal according to the initial weights and an asymmetric least squares method to obtain a target Raman spectrum signal, and detecting the chemical to be detected according to the target Raman spectrum signal.

[0011] Beneficial effects: the application firstly acquires the peak point on the original Raman spectrum signal, then obtains the target similarity sequence of the peak point according to the amplitude difference between the spectrum data points located on both sides of the peak point and the adjacent spectrum data points, obtains the baseline signal segment and the peak signal segment according to the target similarity sequence, obtains the target noise degree value of each spectrum data point on the peak signal segment according to the peak value point and the end point on the peak signal segment, the target similarity sequence and the peak value point difference between the peak signal segment and other peak signal segments, obtains the relative deviation degree value of each spectrum data point on the baseline signal segment according to the difference between the amplitude of the spectrum data point on the baseline signal segment and the overall representative amplitude of the baseline signal segment; then the initial weight of the spectrum data point on the original Raman spectrum signal is distributed according to the target noise degree value and the relative deviation degree value, the initial weight of each spectrum data point on the original Raman spectrum signal is obtained, finally the target Raman spectrum signal is obtained by processing the original Raman spectrum signal according to the initial weight and the asymmetric least squares method, and the chemical to be detected is detected according to the target Raman spectrum signal. And when the asymmetric least squares method is used to process the original Raman spectrum signal, the initial weight distributed based on the target noise degree value and the relative deviation degree value is applied, which can maximize the retention of true peak information, effectively suppress noise, improve the accuracy and robustness of baseline estimation, and further improve or ensure the sensitivity, reliability, accuracy and the like of the detection of hazardous chemicals. BRIEF DESCRIPTION OF DRAWINGS

[0012] In order to more clearly illustrate the technical solutions and advantages of the embodiments of the present application or the prior art, the drawings needed in the embodiments or the prior art description will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present application, and those skilled in the art can obtain other drawings according to these drawings without creating any creative labor.

[0013] Figure 1 The flowchart of the present application based on Raman spectrum hazardous chemical data acquisition and detection method. DETAILED DESCRIPTION

[0014] The technical solutions in the embodiments of the present application will be described clearly and completely below in combination with the drawings in the embodiments of the present application. Obviously, the described embodiments are only some embodiments of the present application, not all the embodiments. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art belong to the scope of protection of the embodiments of the present application.

[0015] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the present application belongs.

[0016] This embodiment provides a method for data acquisition and detection of hazardous chemicals based on Raman spectroscopy, detailed below:

[0017] like Figure 1 As shown, this method for acquiring and detecting hazardous chemicals based on Raman spectroscopy includes the following steps:

[0018] Step S001: Obtain the original Raman spectral signal of the chemical to be detected and the peak points on the original Raman spectral signal.

[0019] This embodiment mainly combines spectral features to adaptively allocate initial weights, thereby avoiding the problem of insufficient distinction between true peaks and noise in the initial iteration stage, which leads to poor noise suppression and baseline correction effects, or reduces the probability of inaccurate baseline estimation and poor signal quality in the final output. In other words, this embodiment's adaptive allocation of initial weights based on spectral features can effectively suppress noise, improve the accuracy and robustness of baseline estimation, and thus improve or guarantee the detection effect of hazardous chemicals, such as improving the sensitivity, reliability, and accuracy of hazardous chemical detection.

[0020] Based on the above description, this embodiment is for the detection of hazardous chemicals, and the detection method is the same for different types of chemicals. Therefore, for ease of understanding and analysis, this embodiment will take the detection process of any benzene-based organic solvent to be detected as an example, and refer to it as the chemical to be detected. If there are strong aromatic ring Raman characteristic peaks in the molecular structure of benzene-based organic solvents, they will usually show obvious peaks in Raman spectroscopy detection. Then, the Raman spectroscopy equipment is used to collect data of the chemical to be detected, and the collected signal is recorded as the original Raman spectral signal. The data points on the original Raman spectral signal are called spectral data points. The horizontal axis of the spectral data points is the Raman shift, and the vertical axis is the scattered light intensity. The amplitude of the subsequent spectral data points is the vertical axis value of the corresponding spectral data points.

[0021] Since the initial weight allocation principles for the baseline signal segment and peak signal segment are different in this embodiment, and the acquisition of the baseline signal segment and peak signal segment in this embodiment is achieved by analyzing the peak points, after obtaining the original Raman spectrum signal of the chemical to be detected, this embodiment needs to calculate and obtain the peak points on the original Raman spectrum signal. The specific acquisition process of the peak points is as follows: the maximum points on the original Raman spectrum signal are obtained by using the differentiation method, and all the obtained maximum points are recorded as peak points. The process of obtaining the maximum points by using the differentiation method is well known. Moreover, the peaks on the Raman spectrum signal usually correspond to the specific vibrational modes of the chemical molecules being measured. Their positions and shapes reflect the molecular structural characteristics, while the peak intensity is related to the content of the component in the sample.

[0022] Therefore, the original Raman spectrum signal of the to-be-detected chemical and the peak point on the original Raman spectrum signal are obtained through the above process.

[0023] In step S002, the target similarity sequence of the peak point is obtained according to the amplitude difference between the spectrum data points located on both sides of the peak point and the adjacent spectrum data points, and the baseline signal segment and the peak signal segment are obtained according to the target similarity sequence.

[0024] In order to ensure the effect of noise suppression and baseline correction, or to improve the baseline estimation accuracy and the signal quality after processing, different initial weight allocation principles need to be set for the baseline signal segment and the peak signal segment in the embodiment, so that the baseline signal segment and the peak signal segment need to be distinguished based on the above-obtained peak point before the initial weight adaptive allocation is performed, that is, the baseline signal segment and the peak signal segment on the original Raman spectrum signal need to be obtained in the embodiment, and the specific obtaining process of the baseline signal segment and the peak signal segment is as follows:

[0025] First, the peak signal segment corresponding to each peak point on the original Raman spectrum signal is obtained, and the peak signal segment corresponding to each peak point on the original Raman spectrum signal is the peak signal segment on the original Raman spectrum signal. The peak signal segment corresponding to the peak point is the peak signal segment at the position of the corresponding peak point on the original Raman spectrum signal. Other signal segments except the peak signal segment on the original Raman spectrum signal are obtained and are all recorded as baseline signal segments on the original Raman spectrum signal. For example, if there are other spectrum data points between the first peak signal segment and the second peak signal segment on the original Raman spectrum signal or the first peak signal segment and the second peak signal segment are not adjacent or continuous on the original Raman spectrum signal, then the signal segment between the first peak signal segment and the second peak signal segment is a baseline signal segment.

[0026] For any peak point W on the original Raman spectrum signal, the specific process of obtaining the peak signal segment corresponding to the peak point W is as follows:

[0027] Since the two sides of the peak point in the peak signal segment where the peak point is located usually show a downward trend and the changes on the two sides of the peak point in the peak signal segment have certain similarity characteristics, the embodiment will next obtain the peak signal segment corresponding to the peak point W based on the characteristics of the peak signal segment where the peak point is located, or in other words, the intersection of the peak and the baseline can be obtained based on the characteristics of the peak signal segment where the peak point is located. First, the target similarity sequence of the peak point W is obtained according to the amplitude difference between the spectral data points located on the two sides of the peak point W and the adjacent spectral data points. The target similarity sequence of the peak point W is the basis for subsequently obtaining the intersection of the baseline and the peak, that is, the basis for obtaining the characteristic intersection. The characteristic intersection is the key to obtaining the baseline signal segment and the peak signal segment. The target similarity sequence of the peak point W includes the left target similarity sequence and the right target similarity sequence. Then, the minimum values in the left target similarity sequence and the right target similarity sequence are obtained. The minimum values are obtained by known techniques. Then, the characteristic intersections corresponding to the peak point W are obtained according to the distances between the spectral data points corresponding to all the minimum values in the target similarity sequence of the peak point W and the peak point W. The characteristic intersections are the intersections or boundary points between the baseline and the peak. Since the number of the target similarity sequences of the peak point W is 2, the number of the characteristic intersections corresponding to the peak point W is also 2. The peak signal segment corresponding to the peak point W is obtained according to the two characteristic intersections corresponding to the peak point W. The peak signal segment corresponding to the peak point W is the signal segment between the two characteristic intersections corresponding to the peak point W. The peak signal segment corresponding to the peak point W contains the characteristic intersections corresponding to the peak point W. The above process not only preserves the peak shape characteristics, but also accurately determines the boundary positions of the peak and the baseline.

[0028] The specific obtaining process of the target similarity sequence of the peak point W is as follows: firstly, on the original Raman spectrum signal, a peak point located on the left side of the peak point W and closest to the peak point W is obtained and recorded as a left adjacent peak point, a peak point located on the right side of the peak point W and closest to the peak point W is obtained and recorded as a right adjacent peak point, a signal segment from the left adjacent peak point to the peak point W is recorded as a left side to-be-analyzed signal segment of the peak point W, and a signal segment from the peak point W to the right adjacent peak point is recorded as a right side to-be-analyzed signal segment of the peak point W. The left side to-be-analyzed signal segment and the right side to-be-analyzed signal segment do not contain peak points, and both belong to the to-be-analyzed signal segment of the peak point W. The left adjacent peak point has the smallest difference in the horizontal coordinate between the left adjacent peak point and the peak point W, compared with other peak points on the left side of the peak point W, and the right adjacent peak point is the same. In addition, if there is no other peak point on one side of the peak point W, the end point on the side is selected as the adjacent peak point. If there is no other peak point on the left side of the peak point W, the left end point on the original Raman spectrum signal is selected as the left adjacent peak point. Then, according to the amplitude difference between each spectrum data point on the to-be-analyzed signal segment of the peak point W and the adjacent spectrum data point of the corresponding spectrum data point, the local similarity of each spectrum data point on the to-be-analyzed signal segment of the peak point W is obtained. The sequence formed by the local similarity of all spectrum data points on the left side to-be-analyzed signal segment of the peak point W is smoothed, and the new sequence obtained after smoothing is recorded as the left side target similarity sequence of the peak point W. The sequence formed by the local similarity of all spectrum data points on the right side to-be-analyzed signal segment of the peak point W is smoothed, and the new sequence obtained after smoothing is recorded as the right side target similarity sequence of the peak point W. The left side target similarity sequence and the right side target similarity sequence both belong to the target similarity sequence of the peak point W. The target similarity of a data point is the result obtained by smoothing the local similarity of the data point. The target similarity in the target similarity sequence is the result obtained by smoothing the local similarity of the data point in the to-be-analyzed signal segment. Here, the median filtering is used for smoothing, and the smoothing is used to reduce the interference of fluctuations on the feature intersection recognition.

[0029] The obtaining process of the local similarity of each spectral data point on the signal segment to be analyzed is as follows: for any spectral data point s on the signal segment to be analyzed, the spectral data points located on both sides of the spectral data point s and adjacent to the spectral data point s are obtained on the original Raman spectrum signal, and are respectively recorded as the left adjacent data point and the right adjacent data point of the spectral data point s. The mean value of the amplitudes of the left adjacent data point and the right adjacent data point is recorded as the comprehensive neighborhood amplitude of the spectral data point s. If there is no other data point on the left side of the spectral data point s on the original Raman spectrum signal, the amplitude of the right adjacent data point of the spectral data point s is directly recorded as the comprehensive neighborhood amplitude of the spectral data point s. If there is no other data point on the right side of the spectral data point s, the same is true. The absolute value of the difference between the amplitude of the spectral data point s and the comprehensive neighborhood amplitude of the spectral data point s is calculated and recorded as the neighborhood difference of the spectral data point s. The neighborhood difference of the spectral data point s is negatively correlated mapped by using the negative exponential function with constant e as the base. The mapping result is recorded as the local similarity of the spectral data point s. The expression of the local similarity of the spectral data point s is wherein exp() is the exponential function with constant e as the base, is the amplitude of the spectral data point s, is the comprehensive neighborhood amplitude of the spectral data point s, The greater the neighborhood difference of the spectral data point s or the greater the local similarity of the spectral data point s, the smoother the change of the spectral data point s between the data points before and after the spectral data point s, and the greater the local similarity of the spectral data point s and the neighborhood data points before and after the spectral data point s.

[0030] The specific process of obtaining the characteristic intersection point corresponding to the peak point W according to the distance between the spectral data point corresponding to all minima in the target similarity sequence of the peak point W and the peak point W is as follows: in the spectral data points corresponding to all minima in the left side of the target similarity sequence of the peak point W, the spectral data point closest to the peak point W is selected as the characteristic intersection point corresponding to the peak point W; in the spectral data points corresponding to all minima in the right side of the target similarity sequence of the peak point W, the spectral data point closest to the peak point W is selected as the characteristic intersection point corresponding to the peak point W, that is, if the horizontal coordinate difference between the spectral data point r and the peak point W is the smallest among the spectral data points corresponding to all minima in the left side of the target similarity sequence of the peak point W, then the spectral data point r is the characteristic intersection point corresponding to the peak point W, and if the target similarity of the target similarity sequence is the target similarity of the spectral data point r, then the spectral data point corresponding to the target similarity is the spectral data point r; and generally, the spectral data point corresponding to the rightmost minimum in the left side of the target similarity sequence of the peak point W is the characteristic intersection point corresponding to the peak point W, and the spectral data point corresponding to the leftmost minimum in the right side of the target similarity sequence of the peak point W is the characteristic intersection point corresponding to the peak point W; in addition, if there is no minimum in the target similarity sequence of the peak point W, then the end point of the target similarity sequence is selected as the characteristic intersection point corresponding to the peak point W, for example, if there is no minimum in the left side of the target similarity sequence of the peak point W, then the spectral data point corresponding to the initial data in the left side of the target similarity sequence is selected as the characteristic intersection point corresponding to the peak point W, and if there is no minimum in the right side of the target similarity sequence of the peak point W, then the spectral data point corresponding to the end data in the right side of the target similarity sequence is selected as the characteristic intersection point corresponding to the peak point W.

[0031] Therefore, the baseline signal segment and the peak signal segment on the original Raman spectrum signal can be obtained by the above process, and the baseline signal segment and the peak signal segment are separated.

[0032] In step S003, the target noise degree value of each spectral data point on the peak signal segment is obtained according to the peak value point and the end point on the peak signal segment, the target similarity sequence, and the difference between the peak value points of the peak signal segment and other peak signal segments.

[0033] Due to the fluorescence background effect of benzene substances, environmental sensitivity is strong, the Raman scattering signal itself is weak and other characteristics, it will be affected by the environment light, stray light, fluorescence interference and complex conditions such as high temperature, high humidity, dust during data acquisition, so there will be random noise in the collected signal, the existence of random noise will lead to the existence of noise components or noise information in all the peak signal segments obtained above, or the existence of random noise not only covers part of the effective peak, but also may lead to peak distortion, if the subsequent initial weight distribution is still in accordance with the existing distribution mode, it will lead to the problem that the real peak and noise are not fully distinguished in the initial fitting stage, or the baseline correction or peak position estimation deviation, if the noise may be misjudged as a real peak, which will lead to baseline estimation deviation or effective information loss, so that the accuracy and reliability of chemical component identification and quantitative analysis are low; in order to avoid the problem of baseline estimation deviation or effective information loss, or to maximize the retention of real peak information while effectively suppressing noise and improving the accuracy and robustness of baseline estimation, the embodiment will analyze the noise degree of the data points on the peak signal segment based on the peak intensity, peak width and distribution density and other related characteristics of the peak signal segment, and the subsequent weight distribution based on the noise degree of the data points can not only maximize the retention of real peak information, but also effectively suppress noise and improve the accuracy and robustness of baseline estimation, so as to ensure that the spectrum detection result of dangerous chemicals is more accurate.

[0034] Based on the above description, the peak signal segment needs to be analyzed to obtain the target noise degree value of each spectrum data point in the peak signal segment, and the specific process of obtaining the target noise degree value of each spectrum data point in each peak signal segment on the original Raman spectrum signal is as follows: for any peak signal segment G:

[0035] Firstly, the real peak possibility characteristic value of the peak signal segment G is obtained according to the amplitude of the peak value point on the peak signal segment G, the abscissa value of the data points at both ends of the peak signal segment G and the target similarity sequence of the peak value point on the peak signal segment G, since the greater the possibility of the peak signal segment G being a real peak, the more likely the data points in the peak signal segment G are disturbed or have lower noise degree, so the real peak possibility characteristic value of the peak signal segment G is an important parameter for determining the target noise degree value of each spectrum data point in the peak signal segment G; the specific process of obtaining the real peak possibility characteristic value of the peak signal segment G is as follows:

[0036] The difference between the abscissa value of the right end data point of the peak signal segment G and the abscissa value of the left end data point of the peak signal segment G is calculated, and is recorded as the abscissa difference value of the peak signal segment G. The ratio between the amplitude of the peak point on the peak signal segment G and the abscissa difference value is calculated, and is recorded as the energy concentration characteristic value of the peak signal segment G. A new set composed of all data in the left target similarity sequence of the peak signal segment G and the right target similarity sequence of the peak signal segment G is recorded as the comprehensive set of the peak signal segment G, and the coefficient of variation of the comprehensive set of the peak signal segment G is calculated. The coefficient of variation of a set is generally the ratio between the standard deviation of the corresponding set and the corresponding set multiplied by 100%. If the peak signal segment G is the peak signal segment corresponding to the peak point G0, then the peak point on the peak signal segment G refers to the peak point G0. The negative correlation mapping results of the amplitude of the peak point on the peak signal segment G, the energy concentration characteristic value of the peak signal segment G, and the coefficient of variation of the comprehensive set of the peak signal segment G are multiplied, and the multiplication result is recorded as the real peak possibility characteristic value of the peak signal segment G. Here, the negative correlation mapping is performed by using the negative exponential function with constant e as the base. The specific calculation expression of the real peak possibility characteristic value of the peak signal segment G is:

[0037]

[0038] wherein, is the real peak possibility characteristic value of the peak signal segment G, is the amplitude of the peak point on the peak signal segment G, is the abscissa value of the right end data point of the peak signal segment G, is the abscissa value of the left end data point of the peak signal segment G, exp() is the exponential function with constant e as the base, is the coefficient of variation of the comprehensive set of the peak signal segment G.

[0039] When the vibration energy level distribution of the peak signal segment is concentrated, the signal source is single, and the characteristics are clear, the peak value of the peak signal segment is higher and narrower at this time, so the corresponding peak signal segment is more likely to be a real peak. When the vibration energy level distribution of the peak signal segment is less concentrated, the signal source is more, and the peak value of the peak signal segment is lower and wider at this time, so the corresponding peak signal segment is more likely to be a pseudo peak. The pseudo peak signal may be mixed with scattered components from noise, background interference or multi-component superposition, or the pseudo peak signal is greatly disturbed by complex conditions or the noise characteristics are obvious. When the peak value of the peak signal segment changes more smoothly on both sides, it generally indicates that the overall signal quality of the peak signal segment is better, the spectral curve continuity is strong, the baseline interference is smaller, and the peak shape is closer to the inherent characteristics of the real molecular vibration, that is, when the peak value of the peak signal segment changes more smoothly on both sides, the corresponding peak signal segment is more likely to be a real peak. Therefore, based on the above description, when is larger, it indicates that the peak signal segment G is more likely to be a real spectrum or a real peak, and when The smaller the value is, the more smoothly and stably the peak point on the peak signal segment G changes on both sides, and the greater the possibility that the peak signal segment G is a real spectrum or a real peak is. When the value is greater than a threshold, the peak signal segment G is a real spectrum or a real peak. The greater the value is, the more concentrated the energy of the peak signal segment G is, and the greater the possibility that the peak signal segment G is a real spectrum or a real peak is. When the value is greater than a threshold, the peak signal segment G is a real spectrum or a real peak. The greater the value is, The smaller the value is, The greater the value is, The greater the value is, and when the value is greater than a threshold, the peak signal segment G is a real spectrum or a real peak. The greater the value is, and the greater the possibility that the peak signal segment G is a real spectrum or a real peak is. When the value is less than a threshold, the peak signal segment G is not a real spectrum or a real peak. The smaller the value is, and the smaller the possibility that the peak signal segment G is a real spectrum or a real peak is.

[0040] After obtaining the real peak possibility feature value, in order to more fully and completely distinguish the real peak and the false peak, further maximize the retention of real peak information, effectively suppress noise, and improve the accuracy and robustness of baseline estimation, the local density feature value of the peak signal segment G is obtained according to the horizontal coordinate difference between the peak value point on the peak signal segment G and the peak value points on each adjacent signal segment in the adjacent signal segment set of the peak signal segment G and the distance factor of each adjacent signal segment in the adjacent signal segment set of the peak signal segment G. Since the local density feature value can reflect the possibility that the peak signal segment G is a real peak, the degree of interference of the peak signal segment G by complex conditions, and the possibility that the peak signal segment G is caused by noise or baseline disturbance, the local density feature value is also an important parameter for determining the target noise degree value of each spectrum data point in the peak signal segment G. The specific process of obtaining the local density feature value of the peak signal segment G is as follows:

[0041] From the original Raman spectral signal, obtain the set of all peak signal segments except for peak signal segment G, and denote it as the set of neighboring signal segments of peak signal segment G. Denote the signal segments in the set as neighboring signal segments. Arrange all neighboring signal segments to the left of peak signal segment G in order of distance from peak signal segment G from closest to farthest, and denote the arrangement as the left neighboring signal segment sequence. Similarly, arrange all neighboring signal segments to the right of peak signal segment G in order of distance from peak signal segment G from closest to farthest, and denote the arrangement as the right neighboring signal segment sequence. Based on the order of the neighboring signal segments in the left and right neighboring signal segment sequences, respectively... The neighboring signal segments in the neighboring signal segment sequence are labeled to obtain the label values ​​of each neighboring signal segment in the left and right neighboring signal segment sequences. The label value of the f-th neighboring signal segment in the left and right neighboring signal segment sequences is f, and the label value of the f-th neighboring signal segment in the right neighboring signal segment sequence is also f. Since the left and right neighboring signal segment sequences belong to the neighboring signal segment set, the label values ​​of each neighboring signal segment in the neighboring signal segment set can be obtained. A negative correlation mapping is performed on the label values ​​of the neighboring signal segments using a negative exponential function with a base of e. The mapping result is denoted as the normalization factor to be normalized for the corresponding neighboring signal segment. The normalization factor to be normalized for the a-th neighboring signal segment in the neighboring signal segment set is... , Let be the normalization factor for the 'a'-th neighboring signal segment. Perform proportional normalization on the normalization factor for each neighboring signal segment in the set of neighboring signal segments. Record the normalization result as the distance factor of the corresponding neighboring signal segment. That is, the ratio of the normalization factor for the 'a'-th neighboring signal segment to the sum of the normalization factors for all neighboring signal segments in the set of neighboring signal segments is the distance factor for the 'a'-th neighboring signal segment. Calculate the absolute value of the difference in abscissa between the peak point on peak signal segment G and the peak points on each neighboring signal segment in the set of neighboring signal segments of peak signal segment G, and then multiply this by the distance factor of the corresponding neighboring signal segment. Record the result as the distance factor for the 'a'-th neighboring signal segment. The set of weighted differences of neighboring signal segments in the set of neighboring signal segments of peak signal segment G is denoted as the set of weighted differences corresponding to peak signal segment G. Specifically, the a-th weighted difference in the set is the product of the absolute value of the difference in the abscissa between the peak point on peak signal segment G and the peak point on the a-th neighboring signal segment in the set of neighboring signal segments, multiplied by the distance factor of the a-th neighboring signal segment. All weighted differences in the set are accumulated, and the accumulated result is negatively correlated using a negative exponential function with base e. The mapping result is denoted as the local density feature value of peak signal segment G. The specific calculation expression for the local density feature value of peak signal segment G is as follows:

[0042]

[0043] wherein, is the local density characteristic value of the peak signal segment G, M is the total number of the neighborhood signal segments in the neighborhood signal segment set of the peak signal segment G, is the abscissa value of the peak point on the peak signal segment G, is the abscissa value of the peak point on the a-th neighborhood signal segment in the neighborhood signal segment set, is the distance factor of the a-th neighborhood signal segment in the neighborhood signal segment set. Since the molecular structures and vibration modes of different chemical components are different, the corresponding vibration energy levels in the Raman spectrum are different, and thus the peak positions and shapes generated often have strong specificity, which makes the spectral characteristics of different components usually separate from each other, and the difference is obvious in the main peak region, that is, the greater the difference between the abscissa value of the peak point on the peak signal segment G and the abscissa value of the peak point on the other peak signal segment adjacent to the peak signal segment G, the greater the probability that the peak signal segment G is a true peak. In contrast, the pseudo-peak generated by noise or baseline disturbance lacks such structural regularity and often appears as isolated or locally dense fragments, that is, the smaller the difference between the abscissa value of the peak point on the peak signal segment G and the abscissa value of the peak point on the other peak signal segment adjacent to the peak signal segment G, the greater the probability that the peak signal segment G is a pseudo-peak or the peak signal segment G is more likely to be caused by noise or baseline disturbance. Therefore, the local density of the peak signal segment G is measured based on the difference between the abscissa values of the peak points on the peak signal segment G and the other peak signal segments, and since the distance between the peak signal segment closer to the peak signal segment G and the peak signal segment G can represent the local density of the peak signal segment G, in order to ensure the reliability of the obtained local density, a distance factor is introduced, which is determined by the proximity of the other peak signal segments to the peak signal segment G. The greater the distance factor of the peak signal segment closer to the peak signal segment G, the smaller the difference between the abscissa value of the peak point on the other peak signal segment adjacent to the peak signal segment G and the abscissa value of the peak point on the peak signal segment G, that is, the smaller or the greater, the smaller the local density of the peak signal segment G, which indicates that the possibility of the peak signal segment G being a true spectrum or a true peak is smaller or the degree of influence of the peak signal segment G by noise such as environmental light source, stray light, fluorescence interference is greater. When the difference between the abscissa value of the peak point on the other peak signal segment adjacent to the peak signal segment G and the abscissa value of the peak point on the peak signal segment G is greater, that is, the greater or The smaller the peak signal segment G is, the greater the local density of the peak signal segment G is. The greater the local density of the peak signal segment G is, the greater the possibility that the peak signal segment G is a real spectrum or a real peak is, or the smaller the influence of environmental light sources, stray light, fluorescence interference and other noises on the peak signal segment G is.

[0044] After the local density characteristic value of the peak signal segment G is obtained, the local density characteristic value of the peak signal segment G and the real peak possibility characteristic value are fused and analyzed, and the fusion result is recorded as the noise characteristic value of the peak signal segment G. The calculation process of the noise characteristic value of the peak signal segment G is: the real peak possibility characteristic value of the peak signal segment G is negatively correlated and mapped by using the negative exponential function with the constant e as the base, and then multiplied by the local density characteristic value of the peak signal segment G. The result of the multiplication is recorded as the noise characteristic value of the peak signal segment G, that is, The greater the noise characteristic value of the peak signal segment G is, the greater the influence of environmental light sources, stray light, fluorescence interference and other complex conditions or noises on the peak signal segment G is, or the greater the noise degree of the peak signal segment G is. The smaller the noise characteristic value of the peak signal segment G is, the smaller the influence of environmental light sources, stray light, fluorescence interference and other complex conditions or noises on the peak signal segment G is, or the smaller the noise degree of the peak signal segment G is.

[0045] Since the signal strength of the data point with a low amplitude in the peak signal segment is small, it is more easily covered by baseline drift or background noise, or the data point with a low amplitude in the peak signal segment is more easily affected by complex conditions or noises. Therefore, after the noise characteristic value of the peak signal segment G is obtained, the target noise degree value of each spectral data point in the peak signal segment G is obtained in combination with the degree of the amplitude of the data point in the peak signal segment G relative to the amplitude of the peak value point. That is, the target noise degree value of each spectral data point in the peak signal segment G is obtained according to the ratio of the amplitude of each spectral data point in the peak signal segment G to the amplitude of the peak value point on the peak signal segment G and the noise characteristic value of the peak signal segment G. The acquisition process of the target noise degree value of the u-th spectral data point in the peak signal segment G is: the negative correlation mapping result of the ratio of the amplitude of the u-th spectral data point to the amplitude of the peak value point on the peak signal segment G is multiplied by the normalized result of the noise characteristic value of the peak signal segment G. The target noise degree value of the u-th spectral data point is recorded as the target noise degree value of the u-th spectral data point. The expression of the target noise degree value of the u-th spectral data point is:

[0046]

[0047] wherein, is the target noise degree value of the u-th spectral data point in the peak signal segment G, and Norm() is a normalization function. an amplitude of the u-th spectral data point in the peak signal segment G, an amplitude of the peak value point on the peak signal segment G, a noise representation value of the peak signal segment G; and the smaller the value of the u-th spectral data point in the peak signal segment G is, and the larger the value of the u-th spectral data point in the peak signal segment G is, the greater the influence of the complex conditions or noises such as the environmental light source, stray light, fluorescent interference, etc. on the u-th spectral data point in the peak signal segment G is, and the greater the target noise degree value of the u-th spectral data point in the peak signal segment G is. the greater the target noise degree value of the u-th spectral data point in the peak signal segment G is, the smaller the value of the u-th spectral data point in the peak signal segment G is, and the larger the value of the u-th spectral data point in the peak signal segment G is, the smaller the influence of the complex conditions or noises such as the environmental light source, stray light, fluorescent interference, etc. on the u-th spectral data point in the peak signal segment G is, and the smaller the target noise degree value of the u-th spectral data point in the peak signal segment G is. the smaller the target noise degree value of the u-th spectral data point in the peak signal segment G is.

[0048] Therefore, the embodiment can obtain the target noise degree value of each spectral data point in the peak signal segment through the above process. Since the greater the target noise degree value of the spectral data point is, the greater the noise degree of the spectral data point is, and the embodiment can effectively suppress the noise and improve the precision and robustness of the baseline estimation while retaining the real peak information to the maximum, the initial weight of the spectral data point with the greater noise degree needs to be greater when the initial weight is allocated to the spectral data point in the peak signal segment subsequently, that is, the target noise degree value is the basis for the initial weight allocation to the spectral data point in the peak signal segment subsequently.

[0049] In step S004, the relative deviation degree value of each spectral data point on the baseline signal segment is obtained according to the difference between the amplitude of the spectral data point on the baseline signal segment and the overall representative amplitude of the baseline signal segment.

[0050] In the baseline correction process, the baseline signal segment is the focus of correction, but due to the randomness of noise, different data points on the baseline signal segment may have different deviations relative to the overall baseline signal segment, so in the subsequent initial weight allocation of the data points on the baseline signal segment, the relative deviation degree of the corresponding data point can be allocated based on the deviation of the different data points on the baseline signal segment relative to the overall baseline signal segment, that is, the initial weight of the data points on the baseline signal segment can be allocated based on the relative deviation degree in the baseline correction, and the initial weight allocated can more accurately recover the baseline form of the true spectrum while effectively suppressing random noise, or effectively suppress noise and improve the accuracy and robustness of baseline estimation, so based on the above description, the relative deviation degree value of each spectral data point on the baseline signal segment can be obtained based on the difference between the amplitude of the spectral data point on the baseline signal segment of the original Raman spectrum signal and the overall representative amplitude of the baseline signal segment, and the subsequent will be described with an example of the acquisition process of the relative deviation degree value of the nth spectral data point on any baseline signal segment N of the original Raman spectrum signal, that is, the specific acquisition process of the relative deviation degree value of the nth spectral data point on the baseline signal segment N is as follows:

[0051] The set of amplitudes of all spectral data points on all baseline signal segments of the original Raman spectrum signal is denoted as a comprehensive amplitude set, the median amplitude of the comprehensive amplitude set is obtained, and is denoted as the overall representative amplitude of the baseline signal segment, and since the median is not easily affected by extreme values, the median amplitude of the amplitudes of all spectral data points on all baseline signal segments can better represent the overall level of the baseline signal segment, the absolute value of the difference between the amplitude of the nth spectral data point and the overall representative amplitude of the baseline signal segment is calculated, the absolute value of the difference between the amplitude of the nth spectral data point and the overall representative amplitude of the baseline signal segment is normalized, and the normalized result is used as the relative deviation degree value of the nth spectral data point. Here, the normalization function Norm() is used for normalization, and the smaller the relative deviation degree value of the nth spectral data point, the greater the possibility that the nth spectral data point is a baseline data point, and the greater the relative deviation degree value of the nth spectral data point, the smaller the possibility that the nth spectral data point is a baseline data point.

[0052] Therefore, the embodiment can obtain the relative deviation degree value of each spectral data point on each baseline signal segment on the original Raman spectrum signal through the above process. In order to effectively suppress noise and improve the accuracy and robustness of baseline estimation, the initial weight of the spectral data point with a larger relative deviation degree value is relatively smaller, and the initial weight of the spectral data point with a smaller relative deviation degree value is relatively larger when the initial weight of the spectral data point on the baseline signal segment is allocated subsequently, that is, the relative deviation degree value is the basis for the subsequent initial weight allocation of the spectral data point in the baseline signal segment.

[0053] In step S005, the initial weight of the spectral data point on the original Raman spectrum signal is allocated according to the target noise degree value and the relative deviation degree value, and the initial weight of each spectral data point on the original Raman spectrum signal is obtained. The original Raman spectrum signal is processed according to the initial weight and the asymmetric least squares method to obtain a target Raman spectrum signal, and the target Raman spectrum signal is used to detect the chemical to be detected.

[0054] After obtaining the target noise degree value and the relative deviation degree value as the basis for the initial weight allocation, the initial weight of the spectral data point on the original Raman spectrum signal is allocated according to the relative deviation degree value of each spectral data point on the baseline signal segment and the target noise degree value of each spectral data point in the peak signal segment on the original Raman spectrum signal, and the initial weight of each spectral data point on the original Raman spectrum signal is obtained. The initial weight of each spectral data point on the original Raman spectrum signal is allocated according to the relative deviation degree value of each spectral data point on the baseline signal segment and the target noise degree value of each spectral data point in the peak signal segment on the original Raman spectrum signal, and the specific process is as follows:

[0055] For the vth spectral data point on the original Raman spectrum signal, if it is judged that the vth spectral data point belongs to the baseline signal segment, then the normalized result of the negative correlation mapping result of the relative deviation degree value of the vth spectral data point and the preset increasing coefficient is added as the to-be-processed weight of the vth spectral data point, and the to-be-processed weight expression of the vth spectral data point is , is the relative deviation degree value of the vth spectral data point on the original Raman spectrum signal, c is the preset increasing coefficient, and the constant 1 is subtracted is to realize the negative correlation mapping of the relative deviation degree value, and since the weight of the baseline data point is required to be large in order to maximize the retention of the true peak information, effectively suppress the noise, and improve the accuracy and robustness of the baseline estimation, the initial weight of the spectral data point belonging to the baseline signal segment is assigned by referring to the preset increasing coefficient, and the preset increasing coefficient can be set by the implementer according to the actual situation in specific application, but the preset increasing coefficient is required to be a positive number, such as setting the preset increasing coefficient to 1 in the embodiment; and if it is judged that the vth spectral data point belongs to the peak signal segment, then the target noise degree value of the vth spectral data point is taken as the to-be-processed weight of the vth spectral data point; the greater the to-be-processed weight, the greater the contribution degree of the corresponding data point when participating in the baseline correction, and the contribution degree of the data point belonging to the baseline signal segment is relatively large when participating in the baseline correction compared with the data point belonging to the peak signal segment, and not only the effective protection of the true peak when correcting the spectral signal can be improved, but also the noise interference can be effectively removed or suppressed.

[0056] After that, the to-be-processed weights of all the spectral data points on the original Raman spectrum signal are proportionally normalized, and the normalization result is taken as the initial weight of the corresponding spectral data point, that is, the initial weight of the vth spectral data point is the ratio of the to-be-processed weight of the vth spectral data point to the cumulative result of the to-be-processed weights of all the spectral data points on the original Raman spectrum signal, and the greater the initial weight, the greater the contribution degree of the corresponding data point when participating in the baseline correction.

[0057] After obtaining the initial weights of all the spectral data points on the original Raman spectrum signal, the original Raman spectrum signal is processed by using the asymmetric least squares method, and the final result output is taken as the target Raman spectrum signal, that is, the initial weight used when the original Raman spectrum signal is processed by using the asymmetric least squares method is the initial weight assigned by the embodiment, and the specific process of processing the original Raman spectrum signal based on the asymmetric least squares method in the embodiment is the same as the existing process. After obtaining the target Raman spectrum signal, the detection of the to-be-detected chemical is realized based on the obtained target Raman spectrum signal of the to-be-detected chemical, and how to collect and detect the component information in the detected chemical based on the Raman spectrum signal of the to-be-detected chemical is a known technology.

[0058] Thus, the embodiment completes detection of the to-be-detected chemical product, and when the original Raman spectrum signal is processed by using the asymmetric least squares method, the initial weight based on the target noise degree value and the relative deviation degree value is applied, so that the real peak information can be retained to the maximum, noise can be effectively suppressed, the precision and robustness of the baseline estimation can be improved, and the detection effect of the dangerous chemical product is improved or guaranteed, such as the sensitivity, reliability, and accuracy of the detection of the dangerous chemical product.

[0059] In summary, the embodiment first acquires the peak points on the original Raman spectrum signal, then obtains the target similarity sequence of the peak points according to the amplitude difference between the spectrum data points located on both sides of the peak points and the adjacent spectrum data points, obtains the baseline signal segment and the peak signal segment according to the target similarity sequence, obtains the target noise degree value of each spectrum data point on the peak signal segment according to the peak points and the end points on the peak signal segment, the target similarity sequence, and the peak point difference between the peak signal segment and other peak signal segments, and obtains the relative deviation degree value of each spectrum data point on the baseline signal segment according to the difference between the amplitude of the spectrum data point on the baseline signal segment and the overall representative amplitude of the baseline signal segment; then the initial weight of the spectrum data point on the original Raman spectrum signal is distributed according to the target noise degree value and the relative deviation degree value, the initial weight of each spectrum data point on the original Raman spectrum signal is obtained, and finally the original Raman spectrum signal is processed to obtain the target Raman spectrum signal according to the initial weight and the asymmetric least squares method, and the to-be-detected chemical product is detected according to the target Raman spectrum signal. When the original Raman spectrum signal is processed by using the asymmetric least squares method, the initial weight based on the target noise degree value and the relative deviation degree value is applied, so that the real peak information can be retained to the maximum, noise can be effectively suppressed, the precision and robustness of the baseline estimation can be improved, and the sensitivity, reliability, and accuracy of the detection of the dangerous chemical product are improved or guaranteed.

[0060] The above-described embodiments are only used to illustrate the technical solutions of the present application, rather than limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that they can modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement for part of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application, and should be included in the protection scope of the present application.

Claims

1. A method for data acquisition and detection of hazardous chemicals based on Raman spectroscopy, characterized in that, The method includes the following steps: Obtain the raw Raman spectral signal of the chemical to be detected and the peak points on the raw Raman spectral signal; Based on the amplitude difference between the spectral data points located on both sides of the peak point and the adjacent spectral data points, a target similarity sequence of the peak point is obtained, and a baseline signal segment and a peak signal segment are obtained based on the target similarity sequence. Based on the peak points and endpoints of the peak signal segment, the target similarity sequence, and the peak point differences between the peak signal segment and other peak signal segments, the target noise level value of each spectral data point on the peak signal segment is obtained; Based on the difference between the amplitude of the spectral data points on the baseline signal segment and the overall representative amplitude of the baseline signal segment, the relative deviation value of each spectral data point on the baseline signal segment is obtained. The relative deviation value of each spectral data point on the baseline signal segment is the normalized result of the absolute value of the difference between the amplitude of the corresponding spectral data point and the overall representative amplitude of the baseline signal segment. The overall representative amplitude of the baseline signal segment is the median amplitude of the set formed by the amplitudes of all spectral data points on all baseline signal segments. The initial weights of the spectral data points on the original Raman spectral signal are assigned based on the target noise level and the relative deviation level, thereby obtaining the initial weights of each spectral data point on the original Raman spectral signal. The original Raman spectral signal is then processed using the initial weights and the asymmetric least squares method to obtain the target Raman spectral signal. The chemical to be detected is then detected based on the target Raman spectral signal. The method for obtaining the target noise level value of each spectral data point on the peak signal segment includes: Based on the amplitude of the peak point on the peak signal segment, the abscissa values ​​of the data points at both ends of the peak signal segment, and the target similarity sequence of the peak point on the peak signal segment, the true peak probability feature value of the peak signal segment is obtained. The set of peak signal segments other than the peak signal segment on the original Raman spectrum signal is denoted as the neighborhood signal segment set of the peak signal segment. The neighborhood signal segments located on both sides of the peak signal segment in the neighborhood signal segment set are labeled in order of distance from the peak signal segment from near to far, so as to obtain the label value of each neighborhood signal segment in the neighborhood signal segment set. The normalized result of the negative correlation mapping result of the label value of the neighborhood signal segment is used as the distance factor of the corresponding neighborhood signal segment. The local density feature value of the peak signal segment is obtained based on the difference in abscissa between the peak point on the peak signal segment and the peak point on the neighboring signal segment, as well as the distance factor of the neighboring signal segment. The product of the negative correlation mapping result of the true peak probability feature value of the peak signal segment and the local density feature value of the corresponding peak signal segment is used as the noise characterization value of the corresponding peak signal segment. The target noise level value of each spectral data point in the peak signal segment is obtained based on the ratio of the amplitude of each spectral data point in the peak signal segment to the amplitude of the peak point in the corresponding peak signal segment and the noise characterization value. The method for obtaining the initial weights of each spectral data point on the original Raman spectral signal includes: For any spectral data point on the original Raman spectral signal, if the spectral data point belongs to the baseline signal segment, the normalized result of adding the negative correlation mapping result of the relative deviation value of the spectral data point to a preset amplification factor is used as the initial weight of the spectral data point; if the spectral data point belongs to the peak signal segment, the normalized result of the target noise value of the spectral data point is used as the initial weight of the spectral data point.

2. The method for acquiring and detecting hazardous chemicals based on Raman spectroscopy as described in claim 1, characterized in that, The method for obtaining the target similarity sequence of the peak points includes: For any peak point, the signal segment between the peak point and its adjacent peak points is denoted as the signal segment to be analyzed. The signal segment to be analyzed includes a left signal segment and a right signal segment. The amplitude difference between each spectral data point in the signal segment to be analyzed and its adjacent spectral data points is denoted as the local similarity of the corresponding spectral data point. The new sequence obtained by smoothing the sequence formed by the local similarity of the spectral data points in the left signal segment to be analyzed and the new sequence obtained by smoothing the sequence formed by the local similarity of the spectral data points in the right signal segment to be analyzed are respectively denoted as the left target similarity sequence and the right target similarity sequence of the peak point. Both the left target similarity sequence and the right target similarity sequence belong to the target similarity sequence of the peak point.

3. The method for acquiring and detecting hazardous chemicals based on Raman spectroscopy as described in claim 2, characterized in that, The method for obtaining the baseline signal segment and the peak signal segment includes: Based on the distance between the spectral data points corresponding to all minimum values ​​in the target similarity sequence of the peak points and the peak points, the peak signal segments on the original Raman spectral signal are obtained; All signal segments on the original Raman spectrum signal other than the peak signal segment are recorded as baseline signal segments.

4. The method for acquiring and detecting hazardous chemicals based on Raman spectroscopy as described in claim 3, characterized in that, A method for obtaining peak signal segments on the original Raman spectral signal based on the distances between the spectral data points corresponding to all minimum values ​​in the target similarity sequence of the peak points and the peak points includes: In each target similarity sequence of the peak point, among all the spectral data points corresponding to the minimum values, the spectral data point closest to the peak point is selected as the feature intersection point corresponding to the peak point. The signal segments between the feature intersection points corresponding to each peak point are all recorded as the peak signal segments on the original Raman spectral signal.

5. The method for acquiring and detecting hazardous chemicals based on Raman spectroscopy as described in claim 1, characterized in that, The method for obtaining the true peak probability feature value of the peak signal segment includes: The difference in the horizontal coordinate between the right-end data point and the left-end data point of the peak signal segment is recorded as the horizontal coordinate difference value, and the ratio of the amplitude of the peak point on the peak signal segment to the horizontal coordinate difference value is recorded as the energy concentration characterization value. The amplitude of the peak point on the peak signal segment, the energy concentration characterization value, and the coefficient of variation of the set consisting of all target similarity sequences of the peak point on the peak signal segment are multiplied together, and the result of the multiplication is recorded as the true peak probability feature value of the peak signal segment.

6. The method for data acquisition and detection of hazardous chemicals based on Raman spectroscopy as described in claim 1, characterized in that, The method for obtaining the local density feature values ​​of the peak signal segment includes: Obtain the weighted difference set corresponding to the peak signal segment. The a-th weighted difference in the weighted difference set is the result of multiplying the difference in the horizontal coordinate between the peak point on the peak signal segment and the peak point on the a-th neighboring signal segment in the set of neighboring signal segments of the peak signal segment with the distance factor of the a-th neighboring signal segment. The negative correlation mapping result of the accumulated result of the weighted difference set is recorded as the local density feature value of the peak signal segment.

7. The method for acquiring and detecting hazardous chemicals based on Raman spectroscopy as described in claim 1, characterized in that, A method for obtaining the target noise level value of each spectral data point in the peak signal segment based on the ratio of the amplitude of each spectral data point in the peak signal segment to the amplitude of the peak point in the corresponding peak signal segment and the noise characterization value includes: The result of negatively correlating the ratio of the amplitude of each spectral data point in the peak signal segment to the amplitude of the peak point in the peak signal segment, and then multiplying it by the noise characterization value of the peak signal segment, is normalized and denoted as the target noise level value of the corresponding spectral data point.

Images