A method for processing target material distribution images in a coherent imaging system

Abstract

The present invention discloses a method for processing a target material distribution image in a coherent imaging system, and belongs to the technical field of coherent imaging image processing. The method includes configuring a reference laser interference signal subdivision multiple so that the exposure trigger frequency is more than 2 times higher than the highest frequency of the light interference signal in the observation band; designing a special narrow-band bandpass FIR filter according to the characteristic absorption wavelength of the target material component to be measured, the reference laser wavelength, and the reference laser interference signal subdivision multiple; based on the area array coherent imaging technology, acquiring the interference signal data cube image frame by frame according to the exposure trigger frequency; sending the newly added frame image data of the data cube to the FIR filter for parallel real-time processing to obtain a two-dimensional normalized image of the material distribution; and obtaining a two-dimensional concentration distribution image with quantitative concentration based on the background removal processing of the normalized image. The present invention realizes synchronous calculation in the process of acquiring the coherent imaging data cube frame image, saves storage resources, and improves the material distribution imaging rate.

Description

Technical Field

[0001] The invention belongs to the technical field of coherent imaging image processing, and in particular relates to a method for processing a target material distribution image in a coherent imaging system. Background Art

[0002] In order to realize the identification and imaging of the material components of the sample to be tested, the most commonly used technical means is to use spectral imaging technology. By combining the functions of imaging and spectral analysis, it is possible to provide spectral information of the sample to be tested in two-dimensional space, identify the components and concentration of the sample to be tested through spectral analysis, and then draw a material component distribution image.

[0003] Spectral imaging technology is mainly divided into three technical routes: filter type, dispersion type, and area array coherence type. The filter type realizes imaging in different bands by inserting filters in the imaging light path, but the spectral resolution is low and the number of filters is limited; the dispersion type often uses slit + grating spectrometry to realize spectral measurement, but the slit limits the light flux and causes the signal-to-noise ratio to be limited. Compared with the first two technologies, the area array coherent imaging spectroscopy technology has the advantages of high light flux (high signal-to-noise ratio), wide spectral range and high spectral resolution, and can better circumvent the principle limitations of the first two technologies. Its technical principles are as follows: Figure 1 shown.

[0004] The area array coherent imaging spectroscopy system uses an interferometer to coherently modulate the optical signal of the sample to be tested containing the target material component. The fixed mirror of the interferometer remains stationary, and the moving mirror completes a push scan. Each pixel of the area array detector can generate a series of analog interference signals containing the intensity information of each wavelength. The signal is output through the analog channel of the area array detector and collected by the ADC to obtain an interference signal data cube. The cube is a huge three-dimensional array, whose ij axis is the pixel coordinate axis of the area array detector, and the t axis is the time axis. The discrete sequence of each pixel along the time axis corresponds to a series of interference sequences in different fields of view. Traditional image processing methods will wait for the interference signal data cube to be completely collected and cached, and then perform FFT (fast Fourier transform) on the interference sequence of each pixel independently to obtain a spectral sequence data cube. The ij axis remains unchanged, and the λ axis is the wavelength axis. The discrete sequence of each pixel along the wavelength axis corresponds to a series of spectral sequences in different fields of view.

[0005] like Figure 2 As shown in the figure, the traditional method is based on the single-pixel spectral sequence in the spectral sequence data cube. , Represents the element number in the sequence; nonlinear least squares fitting is performed independently on each pixel spectral sequence to obtain the composition + concentration information, and the results of each pixel are summarized to generate a material composition and concentration distribution map. Each pixel in the image corresponds to the composition and concentration information of the sample to be tested in different fields of view.

[0006] However, this traditional method has high requirements for data caching and processing computing power. For example, for a 640×512 pixel array detector, each of the 320,000 pixels in total will output a complete interference signal sequence. Each interference signal sequence requires Points, Generally, it takes an integer between 12 and 15, corresponding to 4096, 8192, 16384, 32768, etc.

[0007] Since the more points there are, the higher the spectral resolution after Fourier transformation, the more conducive it is to nonlinear least squares fitting spectrum analysis of material composition and concentration. According to the spectral resolution requirements, 8192, 16384, 32768 data points are generally used, and each data point is the result of 12-16 bits ADC quantization. Therefore, an interference signal data cube and a spectrum sequence data cube after Fourier transformation are three-dimensional arrays with a very large total data volume, and the complete data volume is about 32Gbits~171Gbits. In addition, the traditional method has to wait for the interference signal data cube to be completely collected and cached, and then the interference sequence of each pixel is independently FFT, so the demand for massive data storage is inevitable. Independent nonlinear least squares fitting of each pixel sequence in the spectrum sequence data cube requires a huge amount of floating-point operations, and the total amount of calculation is very large.

[0008] It can be seen that the storage requirements and computational overhead of existing traditional methods are very large, which is not conducive to the miniaturization design of instruments and equipment, and also affects the real-time performance of the output, and is not conducive to application in situations requiring rapid imaging.

[0009] Therefore, how to quickly and efficiently process the massive data of area array coherent imaging and quickly generate the target material distribution image from the massive three-dimensional array data of the initial interference signal data cube has become a technical problem that needs to be urgently solved in area array coherent imaging spectroscopy technology. Summary of the invention

[0010] In order to solve the above technical problems, the present invention provides a method for processing target material distribution images in a coherent imaging system, which can quickly and efficiently process the massive data of area array coherent imaging, greatly reduce the size of data cache space and the amount of calculation, and quickly generate target material composition and concentration distribution images with a certain accuracy.

[0011] To achieve the above purpose, the technical solution adopted by the present invention is as follows:

[0012] A method for processing a target material distribution image in a coherent imaging system, the method comprising the following steps:

[0013] Step 1: According to the Nyquist sampling theorem, the reference laser interference signal subdivision multiple R is configured according to different observation bands, so that the exposure trigger frequency is more than twice the highest frequency of the optical interference signal in the observation band;

[0014] Step 2: According to the characteristic absorption wavelength of the target material component to be measured , Reference laser wavelength , refer to the subdivision multiple R of the laser interference signal, and design a special narrow-band passband FIR filter;

[0015] Step 3: Based on the area array coherent imaging technology, the interference signal data cube image is acquired frame by frame according to the exposure trigger frequency set in step 1;

[0016] Step 4: Send the newly added frame image data of the interference signal data cube to a narrowband bandpass FIR filter for parallel real-time processing, obtain a two-dimensional original image of material distribution based on the convolution result, and perform normalization processing on the two-dimensional original image of material distribution to obtain a two-dimensional normalized image of material distribution;

[0017] Step 5: Based on the background normalized image that does not contain the target substance components to be measured and the two-dimensional normalized image of the sample to be measured that includes the target substance components to be measured, a two-dimensional concentration distribution image for quantitative concentration is calculated.

[0018] The beneficial effects of the present invention are:

[0019] 1) The present invention can process the newly added data of the interference signal data cube in real time, and convolve the newly added data with the narrowband bandpass FIR filter. There is no need to wait for the interference signal data cube to be collected and cached before performing FFT, which saves cache space, improves the real-time performance of calculation, and also saves the design of high-speed data interface for uploading to the host computer. The processing can be completed in the internal processor of the device (such as FPGA, ASIC, etc.);

[0020] 2) The present invention does not require Fourier transform and nonlinear least squares fitting, but convolves the interference signal data cube with a specially designed narrowband bandpass FIR filter, processes in real time and calculates the sum of squares, and finally converts the sum of squares into RMS values, and uses the RMS values ​​to form a two-dimensional image of material distribution. The floating-point operations are very few, and it is very suitable for fixed-point operations of processors such as FPGA or ASIC;

[0021] 3) The present invention can also perform quantitative analysis with a certain accuracy. The target RMS result is divided by the background RMS value, and the division result is converted into absorbance, which is then converted into concentration information with a certain accuracy according to the Lambert-Beer law. This can realize two-dimensional imaging of material distribution + concentration, and is suitable for applications where the target component distribution needs to be quickly analyzed and the accuracy requirement is not very high;

[0022] 4) The present invention can reduce the single pixel interference signal sequence according to the imaging speed requirement The length of the microscope can be shortened, and the acquisition time can be reduced. Rapid imaging of material distribution can be completed in seconds or even sub-seconds, thus improving the imaging speed.

[0023] 5) When generating a narrow-band pass FIR filter according to the characteristic absorption band of the target material component to be measured, the passband can be set to only one characteristic absorption band or a combination of multiple absorption bands, which can be flexibly adjusted according to imaging needs;

[0024] 6) You can also use other digital filters, such as IIR filters, to achieve similar effects;

[0025] 7) The computational complexity of the present invention is greatly reduced, and it can be implemented using FPGA, ASIC, ARM, GPU, DSP, CPU, etc., and has a wide range of applications. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 This is the principle diagram of the area array coherent imaging spectroscopy technology;

[0027] Figure 2 The traditional method is the process from the spectral sequence data cube to the material component concentration distribution map;

[0028] Figure 3 This is a schematic diagram of generating a subdivision trigger signal with reference to the laser interference signal of the solution of the present invention (R=4);

[0029] Figure 4 Schematic diagram of the interference signal data cube composition;

[0030] Figure 5 This is a schematic diagram of the process of generating coefficients of a special narrow-band bandpass FIR filter in the solution of the present invention;

[0031] Figure 6 This is a schematic diagram of the real-time image processing process of the interference signal data cube of the solution of the present invention. DETAILED DESCRIPTION

[0032] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0033] A method for processing a target material distribution image in a coherent imaging system comprises the following steps:

[0034] Step 1: According to the Nyquist sampling theorem, the subdivision multiple R of the reference laser interference signal is configured according to different observation bands, so that the exposure trigger frequency is more than twice the highest frequency of the optical interference signal in the observation band. Figure 3As shown in the case, the subdivision multiple R=4, the vertical axis is the signal amplitude, the horizontal axis is the time, the solid line sine wave is the reference laser interference signal, and the dotted line sine wave is an example wavelength light interference signal in the observation band of the sample to be measured. Only one cycle waveform is drawn. The solid line sine wave interference signal is subdivided to generate trigger pulses Trig1, Trig2, Trig3, Trig4, Next-Trig1, etc. At the same time, selecting different subdivision multiples according to the observation band can reduce the amount of data in the interference signal data cube required when measuring light signals with longer wavelengths.

[0035] In the coherent imaging system, a reference laser interference signal is used to trigger the high-frame-rate image acquisition of equal optical path difference of the optical interference signal of the sample to be measured. This is mainly due to the single and stable frequency characteristic of the reference laser wavelength, which can be used as a scale for measuring the optical interference signal of the sample to be measured.

[0036] By using light sources + optical lenses + detectors of different bands to build an area array coherent interferometer, and subdividing the interference signal of the reference laser according to the observation band, the area array coherent interference signal data cube collection of visible light, near-infrared, mid-infrared and far-infrared band light signals can be realized respectively.

[0037] The subdivision multiple R is determined according to the following conditions: represents the shortest wavelength of the observed band, Indicates the reference laser wavelength:

[0038] ,

[0039] Take HeNe laser as an example, its wavelength is 632.8nm:

[0040] If the wavelength of the sample light to be measured is a visible light signal between 400nm and 632.8nm, the reference laser interference signal should be divided into 4 equal parts (R=4). In this way, at least 2.53 points can be collected per cycle of the interference signal of the light signal with a wavelength of 400nm and longer, meeting the requirements of the Nyquist sampling theorem.

[0041] If the wavelength of the sample light to be tested is a visible light + near-infrared light signal between 632.8nm and 1265.6nm, the reference laser interference signal is divided into two equal parts (R=2). In this way, at least 2 points can be collected per cycle of the interference signal of the light signal with a wavelength of 632.8nm and longer, meeting the requirements of the Nyquist sampling theorem.

[0042] If the wavelength of the sample light to be measured is 1265.6nm or longer near-infrared and above light signals, the reference laser interference signal does not need to be equally divided (R=1). In this way, at least 2 points can be collected per cycle of the interference signal of the light signal with a wavelength of 1265.6nm or longer, meeting the requirements of the Nyquist sampling theorem.

[0043] If the storage and computing power of the data processing system is sufficient, R can also be uniformly set to 4 or a larger subdivision multiple value, but this will result in an increase in the amount of data in the interference signal data cube.

[0044] Step 2: According to the characteristic absorption wavelength of the target material component to be measured , Reference laser wavelength , refer to the laser interference signal subdivision multiple R, design the narrowband bandpass FIR filter coefficients, the process is as follows Figure 5 As shown, the process is as follows:

[0045] Step 2.1: Set the characteristic absorption wavelength of the target substance to be measured (unit: μm), reference laser wavelength (unit μm) converted to wave number (cm -1 ) units, the conversion formula is:

[0046] ,

[0047] Refers to the μm unit wavelength value that needs to be converted. Refers to the converted wave value, the conversion result is (Unit: cm -1 )and (Unit: cm -1 ).

[0048] Step 2.2: Calculate the sampling frequency of the key parameter FIR filter of the present invention :

[0049] ,

[0050] The sampling frequency It is not related to the actual ADC sampling rate of the optical interference signal of the sample to be measured, but only to the reference laser wave number value after conversion to wave number units. and subdivision multiples This is a key point that is different from the sampling frequency setting of the FIR filter in traditional digital signal processing. This greatly simplifies the filter design in the present invention and avoids the influence of the differences in structure and interference signal data cube sampling rate of different planar array coherent interferometers on the filter design.

[0051] Step 2.3: If only one characteristic absorption band is targeted, the FIR filter type is set to bandpass. Set the upper and lower limits of the characteristic absorption band of the target material component in wavenumber (cm -1 ) are the units of and , then the lower passband frequency of the FIR filter is The calculation formula (unit Hz) is:

[0052] ,

[0053] Passband upper frequency The calculation formula (unit Hz) is:

[0054] ,

[0055] Stopband upper and lower frequency , The width of the transition band between the upper and lower limit frequencies of the passband It can be set to 10-100Hz as needed. The amplitude attenuation value between the passband and the stopband Set to around 60-100dB as needed.

[0056] The closer the upper and lower limit frequencies of the passband are to the characteristic absorption wavelength and the transition band width, the The narrower the value, the lower the amplitude attenuation. The higher the value, the more obvious the brightness difference between the target material component and the non-target material component in the final material distribution imaging result. However, the higher the FIR filter order, the greater the amount of calculation. It is controlled at around 200-2000 orders and saved after fixed-point quantization.

[0057] Based on the above FIR design parameters, the calculation tool is called to design and generate the final FIR filter coefficient sequence , Represents the element number in the sequence, and the filter order is recorded as .

[0058] Step 2.4: If only multiple characteristic absorption bands are targeted, refer to step 2.3 to design FIR filters for each characteristic absorption band respectively, align each group of coefficients by serial number, and add up the accumulated average as the final FIR filter coefficients.

[0059] In particular, for plastics, silicon wafers, drugs, polluted gases, etc., solid, liquid and gas can be imaged after pre-processing. Solids and liquids can be imaged by sample preparation + broadband fill light transmission light, and polluted gases can be imaged by measuring the thermal radiation emitted by themselves (the air mass to be measured is high-temperature gas) or the radiation after the background thermal radiation passes through the air mass to be measured (the air mass to be measured is low-temperature gas).

[0060] Based on a standard spectral library containing various material components, a corresponding FIR filter coefficient sequence can be generated according to the characteristic absorption band of each component, and a new spectral library can be formed to adapt to the processing method of the present invention.

[0061] In the unknown component identification scenario, the FIR filter coefficient sequences of different components can be selected from the spectrum library to perform real-time acquisition and processing on the interference signal data cube according to the method of the present invention. According to the light and dark contrast in the final generated image, a reasonable judgment threshold is set to qualitatively judge whether the selected target component is contained.

[0062] When measuring the spectral transmittance characteristics of a material, you can manually set the target band and And generate the corresponding FIR filter coefficient sequence, collect and process the interference signal data cube in real time according to the method of the present invention, and quantitatively obtain the transmittance of the material in this band according to the light and dark contrast in the final generated image.

[0063] Step 3: Based on the area array coherent imaging technology, according to the exposure trigger frequency set in step 1, the interference signal data cube image is acquired frame by frame. After multiple equal optical path difference exposure imaging, the multiple exposure images are arranged according to the time axis t to obtain the interference signal data cube, such as Figure 4 As shown in the figure, the i and j axes are the horizontal and vertical axes of the pixel distribution on the photosensitive surface of the array detector (such as visible light CMOS or infrared focal plane detector), and t is the time axis. Each pixel data changes with the time axis t to form a series of discrete single pixel interference signal sequences. , subscript , Used to distinguish different pixel coordinates (corresponding to the i and j axes of the array detector). Represents the element number in the discrete interference sequence, and the total length is , Generally, it is an integer between 12 and 15.

[0064] Step 4: During the area array coherent imaging process, the processor (such as FPGA) detects the reference laser interference signal trigger pulse, and controls the area array detector (such as visible light CMOS or infrared focal plane detector) to expose and image once, and obtain a new frame of interference data. The interference sequence corresponding to each pixel on the area array detector adds 1 bit of new data. The new frame and the previously processed old frame constitute a new interference signal data cube.

[0065] For the new frame image data, the entire image processing process is as follows Figure 6 As shown, this is the key calculation step for realizing real-time online image processing of interference signal data cube in the present invention.

[0066] The area array detector generally does not output the voltage of all pixels of an image (representing the light intensity detected by the detector pixels) at one time, but uses M pins (usually 4-16) to transmit the voltage of M pixels in each clock cycle, and completes the voltage output of all pixels of an image after multiple clock cycles. All M output voltages are sampled and quantized by M ADCs at the same time, that is, digital quantities that can be stored and calculated are obtained, such as Figure 6 The 4 dark data at the top of the "New Frame Map".

[0067] In order to achieve the purpose of online real-time processing of the interference sequence of each pixel, M-way FIR filter convolution calculation units are implemented in parallel in a processor (such as FPGA), and a scheduling switching module is used to send the M pixel values ​​sampled and quantized in each clock cycle to the M-way "FIR filter convolution calculation unit".

[0068] The FIR filter convolution calculation unit converts the interference signal sampling sequence of each pixel into The FIR filter coefficient sequence generated in step 2 Convolution obtains the sequence of each pixel after filtering , the calculation formula is as follows:

[0069] ,

[0070] In the formula, Represents the convolution operation, which is a process of calculating the single-pixel interference signal sequence from Capture the latest "Numbers of values ​​​​compose the "sliding window of the interference pattern to be convolved" and the FIR filter coefficient sequence Phase convolution, Indicates the index of the value in the "sliding window of the interference image to be convolved". As new frames are continuously input, the convolution results form a filtered interference signal sequence , i and j are still pixel coordinates, n represents the element number in the sequence, and the total number is , For single pixel Sequence length.

[0071] Each time a new frame is input, the sliding window convolution calculation is completed, and the real-time result of the convolution calculation for each pixel is calculated. It is sent to the corresponding RMS calculation module, which is divided into two calculation stages:

[0072] Step 4.1.1: Before the moving mirror of the area array coherence interferometer completes the scanning of the entire measuring optical path, that is, In the stage of continuously generating new data, we will first filter the real-time results of each pixel. The square sum of the values ​​of each pixel is calculated. Use register cache, each register is usually around 48~96 bits, to ensure that the square sum result does not overflow;

[0073] Step 4.1.2: When the moving mirror of the planar array coherent interferometer completes the scanning of the entire measuring optical path, No new data, please wait" "After a clock cycle, the FIR convolution is completed. There is no new data for the value, so the square and calculation results Convert to RMS result , the formula is as follows:

[0074] ,

[0075] The root mean square calculation result of each pixel By pixel coordinates , Stored as a set of two-dimensional arrays, each pixel RMS result The storage bit width is unified (the unified bit width is generally around 16-40 bits), which is the two-dimensional original image of the material distribution .

[0076] In this stage, the contribution of light signals in non-characteristic absorption bands is eliminated by performing real-time convolution processing on the interference signal sequence of each pixel with a specially designed narrow-bandpass FIR filter to obtain a filtered interference signal sequence.

[0077] Calculate the RMS value of the interference signal sequence after filtering as the value representing the light signal intensity of the characteristic absorption band. If the sample area to be tested corresponding to the pixel field of view has strong absorption in the current characteristic absorption band, the light in this band will be absorbed, and the RMS value will be low. It will appear darker in the normalized image, indicating that the concentration of the target substance component in this area is high.

[0078] Due to the differences in single-frame exposure time, background light intensity, and transmittance of the sample to be measured when the area array coherent imaging system collects the interference signal data cube image, the root mean square results under different measurement scenarios The actual effective bit width during storage varies, so the original two-dimensional image of the material distribution Normalize the image to a fixed bit width (usually 8-16 bits) to facilitate the host computer software to save it in a common image format. The steps are as follows:

[0079] Step 4.2.1. Find the RMS value of all pixels Maximum RMS value , determine the position of the highest non-zero bit in its binary format;

[0080] Step 4.2.2, starting from the highest non-zero bit, intercept S bits of binary data including the highest bit to the lower bits (the S value is selected according to the quantization accuracy required for the host computer to store and display the image, generally 8-16 bits) as the normalization result of the pixel;

[0081] Step 4.2.3, refer to the maximum RMS value above The intercepted bit segment intercepts the S-bit binary data of the same segment for the binary format of other pixel values, and the intercepted result is calculated according to the pixel coordinates. , Save to a new set of two-dimensional arrays as a two-dimensional normalized image of material distribution .

[0082] Through this two-dimensional normalized image , the distribution state of the target substance components in the sample to be tested can be qualitatively distinguished. The darker the area in the image, the more target substance components there are, and the brighter the area, the less target substance there is.

[0083] Based on this image, you can perform processing such as inversion, pseudo color, contrast enhancement, and contour extraction to help users view the distribution of materials.

[0084] Step 5: Normalize the material distribution in two dimensions Based on this, the following steps can be followed to obtain a quantitative two-dimensional concentration distribution image: :

[0085] Step 5.1: Follow steps 1 to 4 above to scan and obtain a background normalized image that does not contain the target substance to be measured. ;

[0086] Step 5.2: Generate a two-dimensional normalized image of the sample containing the target substance to be tested. Normalize image with background Divide, each Divide the pixel value by The value of the pixel with the same coordinates in the corresponding pixel is saved as a floating point number. The value of each pixel after processing is recorded as the transmittance , a decimal between 0 and 1.

[0087] Step 5.3: Transmittance of each pixel Convert to absorbance according to the following formula :

[0088] ,

[0089] According to Beer-Lambert law:

[0090] ,

[0091] in is the absorption coefficient of the target substance, is the sample thickness, Calculate the target material concentration for each pixel in the two-dimensional image. :

[0092] ,

[0093] In order to make the quantitative analysis results of concentration more accurate, a calibration experiment was carried out using a variety of samples with known concentrations covering the measurement range to determine the correction function and correction coefficient. After calibration, the corrected concentration result is obtained. , to achieve concentration quantification.

[0094] Step 5.4: The corrected target substance concentration of each pixel The floating point value is then re-fixed to obtain a two-dimensional concentration distribution image with quantitative concentration. Similarly, the image can be processed by inverting, pseudo-coloring, enhancing contrast, extracting contours, etc. to further improve the image readability.

[0095] The specific embodiments described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for processing a target material distribution image in a coherent imaging system, characterized in that: The method comprises the following steps: Step 1: According to the Nyquist sampling theorem, the reference laser interference signal subdivision multiple R is configured according to different observation bands, so that the exposure trigger frequency is more than twice the highest frequency of the optical interference signal in the observation band; Step 2: According to the characteristic absorption wavelength of the target material component to be measured , Reference laser wavelength , refer to the subdivision multiple R of the laser interference signal, and design a special narrow-band passband FIR filter; Step 3: Based on the area array coherent imaging technology, the interference signal data cube image is acquired frame by frame according to the exposure trigger frequency set in step 1; Step 4: Send the newly added frame image data of the interference signal data cube to a narrowband bandpass FIR filter for parallel real-time processing, obtain a two-dimensional original image of material distribution based on the convolution result, and perform normalization processing on the two-dimensional original image of material distribution to obtain a two-dimensional normalized image of material distribution; Step 5: based on the background normalized image not including the target substance component to be measured and the two-dimensional normalized image of the sample to be measured including the target substance component to be measured, a two-dimensional concentration distribution image for quantitative concentration is calculated; Wherein, the step 2 comprises: Step 2.1: Set the characteristic absorption wavelength of the target substance to be measured , Reference laser wavelength Convert to their corresponding wave number units , ; Step 2.2: Calculate the sampling frequency of the key parameter FIR filter ; Step 2.3: For a characteristic absorption band, the FIR filter type is set to bandpass, and the upper and lower limits of the characteristic absorption band of the target material component are set in units of wavenumbers. and , then the lower passband frequency of the FIR filter is The calculation formula is: ， Passband upper limit frequency The calculation formula is: ， Combined sampling frequency , passband lower limit frequency , Passband upper limit frequency , passband stopband transition band width , Passband and Stopband Amplitude Attenuation Value , generate the FIR filter coefficient sequence , Represents the element number in the sequence, and the filter order is recorded as ; Step 2.4: For multiple characteristic absorption bands, design a FIR filter for each characteristic absorption band according to step 2.3, align each group of coefficients by serial number, and add up the accumulated average as the final FIR filter coefficient.

2. A method for processing target material distribution images in a coherent imaging system according to claim 1, characterized in that: In step 1, the subdivision multiple R of the reference laser interference signal is taken according to the following conditions: represents the shortest wavelength of the observed band, Indicates the reference laser wavelength: 。 3. The method for processing target material distribution images in a coherent imaging system according to claim 1, characterized in that: The step 3 includes: acquiring interference signal data cube images frame by frame according to the exposure trigger frequency set in step 1, arranging multiple exposure images according to the time axis t, and obtaining an interference signal data cube. Each pixel data changes with the time axis t to form a series of independent interference sequences, which are recorded as , subscript , represents pixel coordinates, Represents the element index in a sequence.

4. The method for processing target material distribution images in a coherent imaging system according to claim 1, characterized in that: In step 4, the newly added frame image data of the interference signal data cube is sent to the narrow-band bandpass FIR filter for parallel real-time processing, and the two-dimensional original image of the material distribution is obtained based on the convolution result, including: The convolution calculation unit of the FIR filter is to sample the interference signal of each pixel in a sequence With the FIR filter coefficient sequence Convolution obtains the sequence of each pixel after filtering , the convolution calculation formula is as follows: ， In the formula, Represents the convolution operation, which is a process of calculating the single-pixel interference signal sequence from Capture the latest "Numbers of values ​​make up the "sliding window of the interference pattern to be convolved" and the FIR filter coefficient sequence Phase convolution, Represents the index of the value in the "sliding window of the interference pattern to be convolved"; The sequence after filtering each pixel Send it to the corresponding RMS calculation module to calculate the RMS value of each pixel, and calculate the RMS value of each pixel according to the pixel coordinates. , It is stored as a set of two-dimensional arrays, which is the two-dimensional original image of material distribution.

5. The method for processing target material distribution images in a coherent imaging system according to claim 1, characterized in that: In step 4, the step of normalizing the two-dimensional original image of material distribution to obtain a two-dimensional normalized image of material distribution includes: Step 4.2.

1. Find the RMS value of all pixels Maximum RMS value , determine the maximum RMS value The position of the highest non-zero bit in binary format; Step 4.2.2, starting from the highest non-zero bit, including the highest non-zero bit, intercept S bits of binary data from the lower bits inward as the maximum RMS value The pixel normalization result of the corresponding pixel; Step 4.2.3, refer to the maximum RMS value The bit segment to be cut off is the one with the largest RMS value removed. The binary format of other pixel values ​​of the corresponding pixel intercepts the S-bit binary data of the same segment, and the interception result is calculated according to the pixel coordinates , Save to a new set of two-dimensional arrays as a two-dimensional normalized image of material distribution .

6. The method for processing target material distribution images in a coherent imaging system according to claim 1, characterized in that: The step 5 comprises: Step 5.1, follow steps 1 to 4 to scan in advance to obtain a background normalized image that does not contain the target substance component to be measured; Step 5.2: Divide the two-dimensional normalized image of the sample to be tested containing the target substance component by the background normalized image to obtain the transmittance of each pixel. ; Step 5.3: Transmittance of each pixel Convert to absorbance according to the following formula : ， According to Beer-Lambert law: ， in is the absorption coefficient of the target substance, is the sample thickness, The concentration of the target substance component is calculated for each pixel in the two-dimensional image. : ， Concentration of substances Perform calibration and obtain the corrected concentration result ; Step 5.4: The corrected concentration result of each pixel The corresponding floating-point values ​​are then re-fixed to obtain a two-dimensional concentration distribution image with quantitative concentration.

Images