Research on the Seed Respiration CO 2 Detection System Based on TDLAS Technology

,


Introduction
Respiration is an important physiological phenomenon of seeds. Tere is a big diference in nature and respiration strength between distinct seeds. Seeds produce CO 2 and consume O 2 in the process of aerobic respiration. Te intensity of seeds aerobic respiration refects the seed vitality [1][2][3] and the strength of seed metabolism. Terefore, the study of seed respiration is of great signifcance to understand the physiological and biochemical processes of seed vigor, seed metabolism, and seed germination.
Common methods for detecting seed respiration include the small basket method, Warburg microrespirometer, Q2 oxygen sensor technology, gas chromatography, and NDIR CO 2 sensor. Te small basket method [4] is also called the titration method with the advantage of simple operation.
However, the small basket method will cause damage to the seeds and has large human operating errors and low accuracy. Tus, it is not suitable for studying the process of seed respiration with the small basket method. Te Warburg microrespirometer [5] can improve the sensitivity and accuracy of seed respiration detection to a certain extent, but it is only applicable to the respiration detection of batches seeds. Due to its special breathing chamber structure, constant temperature is required during the experiment. Te Q2 oxygen sensor technology [6,7] can measure the oxygen consumption data of a single seed during the germination process and analyze the characteristics of the oxygen consumption curves comprehensively. But the sampling interval of seed respiration oxygen consumption data is about 30 minutes to 1 hour, so the time resolution is low and the measurement period generally takes more than 2 days. What is more, only the respiratory intensity of seed germination stage can be measured using Q2 oxygen sensor technology, for the confned space will have a coercion on seed respiration with time increasing. Gas chromatography is mostly used to measure high molecular weight gas, and it also has high measurement accuracy for CO 2 , but its detection process is more complicated. Te studied gas needs to be collected and flled into the gas chromatography, making it difcult to monitor seed respiration continuously. Besides, we can only get the gas volume fraction not the gas concentration directly. Terefore, there are still certain limitations in the application of seed respiration detection for using gas chromatography. Te above detection methods are all nonspectral analysis methods, and the measurement principle is relatively simple, but the biggest difculty is that the sampling process is complicated and not real time. Using the NDIR CO 2 sensor to measure seed respiration has fast response time and low detection cost, which shortens the seed respiration detection cycle to a certain extent. But unfortunately, the accuracy of NDIR CO 2 sensors is generally 30-50 ppm, and seed respiration detection devices need to be designed again and again according to diferent types of NDIR CO 2 sensors [8,9]. In summary, most traditional seed respiration detection methods require long-term collection of CO 2 emitted by the test samples. Te workload of detection and analysis is large, and most of them are destructive testing.
TDLAS has the characteristics of high sensitivity and high resolution and has been widely used in trace gas detection [10][11][12][13][14]. Many scholars at home and abroad have conducted research on TDLAS technology in the CO 2 concentration detection method and detection sensitivity [15][16][17][18][19]. TDLAS detection methods mainly include WMS and direct absorption spectroscopy. WMS can eliminate the low-frequency noise interference caused by the laser itself and the temperature and current during detection. Compared with direct absorption spectroscopy technology, WMS is easier to extract rich harmonic information and obtain a higher signal-to-noise ratio of the detection system, and the gas detection limit is easier to reach ppm or even ppb level [20,21]. Based on the quadrature demodulation method, the use of the frst harmonic to normalize the second harmonic (2f/1f ) for gas concentration inversion can eliminate the dependence of laser power [22].
Aiming at the problems of seed respiration CO 2 detection, combined with the development of TDLAS in CO 2 detection technology, we have developed a high-sensitivity seed respiration CO 2 detection system based on WMS-TDLAS. Ten, we compared the performance of our developed system with the NDIR CO 2 sensor and gas chromatography. A corn seeds respiration test was carried out in a laboratory environment. Te test results show that the seed respiration CO 2 detection system has reached the design expectations. Te main contributions or innovation of this article are as follows: (1) Based on Herriot cell, a 21 m efective optical path double-layer seed respiration detection device is designed for the frst time, and a seed respiration CO 2 detection system is built based on WMS-TDLAS technology. Te cost is relatively low, and real-time online detection can be realized, which solves the situation of no special seed respiration CO 2 detection sensor at present. (2) Te sampling frequency of the seed respiration CO 2 detection system was 7 Hz and the detection limit was 1.23 ppm, and the respiratory changes of a small number of seeds could be detected. Te linearity and stability of our developed system is good, and the CO 2 concentration of seed respiration online monitoring with high precision and high time resolution is achieved. Compared with gas chromatography or NDIR CO 2 sensor, its time resolution is tens to hundreds of times higher. Terefore, the seed respiration CO 2 detection system can efectively observe the changes of seed respiration and provide an accurate respiration detection sensor for the in-depth study of seed physiological changes. (3) A correction method for temperature and pressure is proposed with the second harmonic for concentration inversion. Te correction relationship between temperature or pressure changes and 2f peak value is calculated, which provides a theoretical method and correction equation for accurate inversion of seed respiration CO 2 concentration in diferent environments. Tus, the sensor can work in a wide range of temperature and pressure, which expands the application scenario of seed respiration detection.

Teoretical Principle.
A monochromatic light beam passing through a uniform gas medium will lose intensity due to optical absorption by gas molecules. Tis process can be described by Beer-Lambert law [23,24]: where I 0 is the intensity of laser without CO 2 absorption, P is the total gas pressure (atm), S(T) is the absorption line intensity as a function of temperature, C is the concentration of CO 2 produced by the respiration of tested seeds, L is the effectiveness of the absorption cell optical path (cm), φ(v) is the absorption linear function of gas molecules, α(v) is the absorption rate function, and σ(v) is the absorption cross section. TDLAS technology measurement methods include direct absorption spectroscopy and WMS. Among them, direct absorption spectroscopy [25,26] uses the change of laser intensity after gas absorption to analyze the concentration of the measured gas. Due to the low concentration of CO 2 produced by seed respiration, the gas absorbs less incident laser intensity. Terefore, the fuctuation of laser intensity, the density change of the gas to be measured, and the selection of the baseline will all cause large measurement errors in direct absorption spectroscopy. In order to solve the accurate online measurement of the CO 2 concentration of seed respiration in harsh environments, we use the WMS algorithm [27,28] to improve the detection accuracy and sensitivity of the TDLAS technology.

2
International Journal of Optics Inject a high-frequency sine wave into the laser for highfrequency modulation; the instantaneous frequency v(t) of the laser can be described as where v 0 is the laser center frequency, δv is the modulation amplitude, and f is the sine wave frequency. Ten, the cosine expansion of the laser intensity I t after gas absorption is carried out by Fourier series: Design two reference signals with a phase diference of 90°and use the phase sensitive detection (PSD) algorithm to obtain the n-th harmonic signal component An of the absorption spectrum signal as Ideally, I0 has nothing to do with v(t), so there is no amplitude modulation, and I0 is a constant. Tus, equation (4) can be rewritten as In general, when the absorption rate function α(v)L ≪ 1, Expanding the formula (6) according to Taylor series, we can get An as follows: As shown in formula (7), the n-th harmonic amplitude of CO 2 absorption signal of seed respiration is proportional to the gas concentration. Since the absorption spectrum of CO 2 is symmetrical about the center frequency, the odd harmonics of the absorbed energy are zero and the even harmonics reach the maximum at v0. For even harmonics of diferent orders, the value at v0 decreases as the order increases and the amplitude of the 2f signal is the largest. Terefore, considering the symmetry and amplitude of the harmonics and avoiding the infuence of the phase error, we use the amplitude of the second harmonic to perform concentration inversion and calculate the CO 2 concentration produced by seed respiration.

Selection of Spectral Absorption Lines.
Te seed respiration CO 2 detection system based on WMS-TDLAS technology is mainly used to monitor the concentration of seed respiration CO 2 in real time. By analyzing seed respiration intensity and respiration rate, seed vigor can be determined quickly, nondestructively, and efciently. Terefore, CO 2 is the target gas. In order to measure the concentration of CO 2 in seed respiration accurately, it is necessary to perform spectral simulation to determine whether the selected line has sufcient absorption line strength and can avoid interference from other gases. In the detection of seed respiration intensity, the main interference gas is H 2 O. Both CO 2 and H 2 O have absorption bands around 1.6 μm and 2.0 μm [29]. Te absorption line of CO 2 near 2.0 μm is stronger than that near 1.6 μm, and the absorption of H 2 O near 2.0 μm is weaker.
In an environment of 1 atm, room temperature, and a light path of 21 meters, based on the HITRAN 2016 database, the 2000 nm band is used to perform absorption simulation calculations for 400 ppm CO 2 and 1% H 2 O [30]. Te simulated absorbance of CO 2 is shown in Figure 1(a). It can be seen that there are several CO 2 absorption lines near the 2004 nm band and H 2 O has little interference with CO 2 in this band. Tere are three strong absorption lines near the frequency of 4989 cm − 1 -4993 cm − 1 , with the center frequencies of 4990 cm − 1 , 4991.25 cm − 1 , and 4992.5 cm − 1 , respectively. Considering the cost-efectiveness of the laser, we choose the 2004 nm DFB laser (nanoplus, maximum power 5 mW, Germany) for seed respiration intensity detection.
When the efective optical length, temperature, and pressure are determined, the driving current will afect the center frequency of the absorption spectrum. Figure 1(b) shows the absorption spectrum of CO 2 under diferent driving currents International Journal of Optics (50 mA-90 mA). In order to select the CO 2 absorption line with a center frequency of 4991.25 cm − 1 accurately, we set the DFB laser driving current to 70 mA.

Experimental Systems.
Our developed double-layer seed respiration device includes seed respiration chamber and Herriott cell. A rectangular cover is arranged above the seed respiration chamber, and a silicone ring is used to seal between the rectangular cover and the seed respiration chamber. When the silicone ring is squeezed and deformed by the pressure from the cover, the double-layer seed respiration device is sealed to avoid the infuence of external air in the seed respiration detection. A flter screen is set between the seed respiration chamber and the Herriott cell with a steel mesh structure; thus, the gas between the seed respiration chamber and the Herriott cell is fully difused, while impurities and dust in the seeds are prevented from falling into the Herriott cell. Te respiration space of seeds is relatively large, so there is no need to consider the inhibition of seed respiration by changes in concentration of CO 2 and O 2 .
Te Herriott cell is composed of two spherical mirrors A and B placed on opposite sides, with a schematic diagram of the basic structure shown in Figure 2.
Te laser emits from the entrance hole and then exits from the exit hole after multiple refections, forming a complete refection light path. Te light path transmission matrix is shown as where N is the number of refected light spots on one of the spherical mirrors, K is the number of times the light ray rotates around the optical axis, and K < N, N is not divisible by K. We defne where d is the distance between the two spherical mirrors, R 1 and R 2 are the radii of curvature of spherical mirrors A and B. Generally, R 1 � R 2 � R, and d can be calculated as A relatively large efective optical path can be obtained by increasing the number of refections. When a light spot is close to the exit hole with a large radius, part of the laser may leak from the Herriott cell and reach the detector during the refection process. At this time, there will be a large error between the measured optical path and the target optical path. Terefore, the size of the entrance hole and the exit hole should be as small as possible. In addition, the distribution of light spots should be close to the edge of the spherical mirrors to accommodate more light spots. Due to the infuence of refectivity of spherical mirrors, too many refections means more light intensity attenuation, which will reduce the signal-to-noise ratio and afect the stability of the optical path. Based on the above research and development concept, we use the TracePro software (V8.1, LAMBDA research corporation, 28 light spots are formed on the spherical mirror B, and the total optical path is about 20.41 m. With the increase in the number of refections, the size of the light spots does not increase, indicating that the Herriott cell has a better focusing ability for the light path. Te two spherical mirrors are connected by a cylinder in the middle, and the lenses are coated with a midinfrared high-refection flm. Te entrance and exit laser adopts an adjustable structure, equipped with FC/APC standard connectors (F028APC-2000, THORLABS). Based on the above simulation results, we developed a double-layer seed respiration device with a volume of about 1.5 L, which has the advantages of long optical path, relatively small volume, wide temperature adaptation range, strong focusing ability, corrosion resistance, and easy installation and operation. Te double-layer seed respiration device is shown in Figure 4.
Te seed respiration CO 2 detection system is shown in Figure 5, which mainly includes a double-layer seed respiration device for placing seeds (one to hundreds of seeds can be placed), a circuit and control module, and a data processing module. Te DFB laser is driven by a temperature current controller, and the scanning frequency of the driver is 50 Hz. Te laser is focused by the collimator into the Herriott cell of the double-layer seed respiration device to detect the CO 2 concentration. Te laser is received by the detector after multiple refections in the Herriott cell, and then after twostage amplifcation, the signal is collected and recorded by the DAQ data acquisition card. Ten, the upper computer software performs spectral signal fltering, 2f signal demodulation, and inversion of gas concentration.
When the temperature changes, the temperature drift of the circuit components will aggravate the varieties in the working conditions of the DFB laser. Terefore, it is necessary to control the temperature of the DFB laser accurately and perform the temperature compensation. In addition, in order to obtain the stable wavelength of the DFB laser, we designed a wavelength closed-loop locking method. After the spectrum signal is collected by the upper computer software, the absorption peak position is calculated. Te direction of wavelength drift is determined according to the error between the current absorption peak position and the target absorption peak position. And then, a wavelength adjust signal is sent to the MCU (STM32H743VIT6) to adjust the output wavelength of the DFB laser.

Signal Processing.
In order to obtain the measurement data of the project quickly and accurately, we prefer to flter and smooth the obtained seed respiratory CO 2 spectral signal by using Savitzky Golay based on empirical mode decomposition (EMD-SG) fltering algorithm [31,32] after cumulative average and normalization. EMD-SG fltering algorithm can fully retain the useful information in the spectral signal and improve the signal-to-noise ratio. Ten, the PSD algorithm is used to extract the 2f signal from the denoised seed respiratory CO 2 spectral signal. Finally, the ratio of the peak value of the 2f signal to the peak value of the calibrated gas 2f signal is used to calculate the CO 2 concentration of seed respiration. Te detail step of the specifc data acquisition process is presented as follows and is shown in Figure 6: (1) Obtain the efective seed respiration CO 2 absorbance spectrum signal and fltering We use the EMD-SG algorithm for denoising and fltering shown in Figure 7. Te CO 2 absorbance spectrum is decomposed by EMD to get IMF and residual component.  International Journal of Optics 5 optimization (PSO) algorithm with an appropriate window size and polynomial order. (2) Extract 2f signal with PSD algorithm and its parameters setting Te most important thing in the demodulation of 2f signal is to set the parameters reasonably to enhance the signal-to-noise ratio and obtain a high-quality second harmonic lineshape. In order to avoid the infuence of the phase error, the PSD algorithm is used to extract the 2f signal from the denoised effective seed respiration CO 2 absorption spectrum. Te phase diference between two reference signals of the PSD algorithm is 90°, which ensures that the quadrature output result remains constant with the initial phase changing. Te denoised efective seed respiration CO 2 absorption spectrum X (t) is expressed as Fourier series: And the reference signal is set as We multiply efective seed respiration CO 2 absorption spectrum X (t) and the reference signal R (t) to get V out (t): V out (t) � V r V s1 sin ω r t + θ sin ω 0 t + φ 1 + V r V s2 sin ω r t + θ sin 2ω 0 t + φ 2 + V r V s3 sin ω r t + θ sin 3ω 0 t + φ 3 + · · · + V r sin ω r t + θ n(t).  V out (t) � 0.5V r V s1 cos ω 0 − ω r t + φ 1 − θ − 0.5V r V s1 cos ω 0 + ω r t + φ 1 + θ + 0.5V r V s2 cos 2ω 0 − ω r t + φ 2 − θ − 0.5V r V s2 cos 2ω 0 + ω r t + φ 2 + θ + 0.5V r V s3 cos 3ω 0 − ω r t + φ 3 − θ − 0.5V r V s3 cos 3ω 0 + ω r t + φ 3 + θ + · · · + V r n(t) sin ω r t + θ .

International Journal of Optics
In order to extract the 2f signal, we set the parameter as ω r � 2ω 0 ; then, V out (t) is In addition to the DC signal, the output signal V out (t) also contains 1st order harmonic frequency, 2nd order harmonic frequency, and high order harmonic frequency components. Low-pass fltering is used to retain the DC signal 0.5V r V s2 cos (φ 2 − θ) from V out (t). It is necessary to set the low-pass flter cutof frequency ω s < ω 0 , and the pass-band cut-of frequency ω p should be greater than the scanning signal frequency. In this project, the signal scanning frequency is set as 50 Hz and the modulation frequency is set as 200 KHz. We use the Butterworth flter to flter the signal with its parameters of ω p � 2 × 10 − 5 rad/s, ω s � 0.064rad/s, the maximum attenuation of the pass band Rp � 1 dB, and the minimum attenuation of the stop band Rs � 55 dB. (3) Correction of 2f signal in diferent temperature and pressure In the process of concentration inversion, if temperature and pressure vary widely, they need to be corrected. Te seed respiration CO 2 detection system developed in this study is operated under standard atmospheric pressure or a gas environment of the same pressure during the test process, so the pressure change can be ignored in the measurement process. However, the temperature varies greatly when the user measures at diferent times in diferent weather, so it is necessary to correct the measurement results of seed respiration at diferent temperatures. Under the same modulation factor, the second harmonic peak value will decrease as the temperature increases. Figure 8 shows the relationship between the peak value of the second harmonic and the temperature under the environment of 21 m efective optical path, 0°C-45°C, 0.8°atm-1.5 atm.
At 1 atm, the third-order polynomial is used to analyze CO 2 with a volume fraction of 1%. Te peak change of the 2f signal of gas is ftted, and the ftting equation is obtained as    International Journal of Optics T 0 (T 0 � 20°C). Te corrected second harmonic peak F T is shown in formula (15), and similar methods can be used for correction under other pressure conditions.

Seed Respiration CO 2 Detection System Performance.
In order to use the seed respiration CO 2 detection system to achieve real-time, high-precision, and nondestructive testing of seed respiration, we have prepared a series of test experiments to detect system performance. When the ambient temperature changes, the laser intensity will change. In order to ensure that the detector can receive stable laser intensity during continuous seed respiration detection, we placed the seed respiration CO 2 detection system in a temperature regulator (MXA600, China) to test the detected laser intensity at diferent temperatures. With 5°C as the step size, set the temperature variable in the range of 5°C-45°C, and a temperature current controller is used to control the output of the 2004 nm DFB laser. Under the condition of constant laser source output, test the laser intensity change for 1 hour continuously. A photodetector is used to convert the optical signal into an electrical signal, and the voltage amplitude is used to express the laser intensity. Te test voltage output is shown in Figure 9 with a fuctuation of about 1.7%, which means that the output laser intensity of the double-layer seed respiration device fuctuates very little without interference. It can be seen that the seed respiration CO 2 detection system has good laser output stability and can meet the requirements of seed respiration measurement.
In order to analyze the linearity of the seed respiration CO 2 detection system, we prepared diferent concentrations of standard CO 2 for testing under the conditions of 20°C and 1 atm. Before the testing, pour high-purity nitrogen into the Herriott cell for 5-6 minutes to drain the gas in the container and keep the modulation and demodulation parameters unchanged during the measurement. Te second harmonic absorption signal of CO 2 is calculated according to the EMD-SG algorithm and the PSD algorithm as shown in Figure 10(a). Te peak value of the 2f signal and the CO 2 concentration present a good linear relationship as shown in Figure 10(b) (adj. R 2 ≈ 0.999), indicating that the EMD-SG, PSD algorithm, and modulation and demodulation parameters used in this article are suitable for CO 2 detection in this system.
In order to further analyze the stability of the system, Allan variance is introduced. Allan variance is not only an important indicator of the stability of the detection system but also can predict the detection limit. We fush in 3008 ppm CO 2 standard gas into the Herriott cell and calculated the Allan variance during the 1000 s. It can be seen from Figure 11 that the Allan variance shows a trend of decreasing frst and then increases with the integration time. When the integration time is 1/7 s, the Allan variance is 143.34 ppm. As the integration time increases, the stability of the system is better, and the Allan variance decreases. When the integration time is equal to 51 s, the detection limit of the system is 1.23 ppm. Tis is because increasing the integration time can reduce Gaussian noise efectively, thereby reducing the detection limit and improving system stability. Ten, mainly due to the infuence of system drift and noise, Allan variance begins to increase again.
Te repeatability of the seed respiration CO 2 detection system can be calculated based on the relative standard deviation (RSD). In the same measurement environment as possible, 1008 ppm of standard CO 2 is fushed into the double-layer seed respiration device and sealed, and repeated measurements are performed 50 times from time to time. Te RSD of our developed system is calculated to be 0.00046. Table 1 compares the seed respiration CO 2 detection system with commonly used CO 2 detection methods. Compared with the widely used NDIR CO 2 sensor and gas chromatography, the seed respiration CO 2 detection system has better repeatability, and the detection limit is basically the same as gas chromatography, which is much higher than that of the NDIR CO 2 sensor. When the measuring frequency is about 0.2 Hz, the overall performance of the seed respiration CO 2 detection system is better. In addition, compared with the gas chromatography, the seed respiration CO 2 detection system has the advantages of simple operation, no need for assistance such as carrier gas, no sampling limitation, and direct concentration inversion.
In order to compare the stability of the seed respiration CO 2 detection system and the NDIR CO 2 sensor during continuous measurement, we used the above two detection methods to measure 702 ppm standard CO 2 continuously for about 30 minutes. (Because the gas chromatography needs to collect gas and inject it into the device before detecting, it is not suitable for high-resolution continuous monitoring of seed respiration intensity, so the gas chromatography is not used as a comparison object.) Figure 12(a) shows two methods within 30 minutes for the detection results of 702 ppm standard CO 2 . Te measurement frequency of the seed respiration CO 2 detection system is International Journal of Optics signifcantly higher than that of the NDIR CO 2 sensor, with a good ability to capture the details of signal changes. During the test, both systems fuctuate within a certain range. Figures 12(b) and 12(c) calculate the residuals of the 702 ppm standard gas measured by the two methods, respectively. Te fuctuation range of the seed respiration CO 2 detection system is about 5 ppm-7 ppm and the fuctuation range of the NDIR CO 2 sensor is about 20 ppm-28 ppm. Obviously, the measurement results of our developed system are more stable, and the fuctuation range is much lower than that of the NDIR CO 2 sensor. Terefore, the seed respiration CO 2 detection system is much more suitable for   detecting the respiratory intensity of a small amount of seeds.
In order to verify that our developed system can detect the CO 2 concentration produced by seed respiration in real time and with high accuracy, we selected a small amount of sweet corn seeds and waxy corn seeds (1 g, about 40 seeds) as samples to measure the respiration intensity in a laboratory environment and calculate breathing rate. For ensuring the accuracy of the seed respiration rate calculation and taking the amount of calculation into account, we calculated the seed respiration rate every 0.5 hours. Te average value of 20 sampling points before and after each time interval was taken as the seed respiration intensity at that moment shown in Table 2. Figure 13 shows the changes in the intensity and respiration rate of less corn seeds (1 g) visually and carefully. We can fnd that with time increasing, the respiration intensity of sweet corn seeds and waxy corn seeds gradually increases. Te respiration intensity and respiration rate changes of sweet corn seeds and waxy corn seeds in the same weight are diferent. Te respiration intensity of sweet corn  Figure 13: Variation trend of respiration intensity and respiration rate of 1 g sweet corn seeds and waxy corn seeds.  Figure 12: Stability test of the seed respiration CO 2 detection system and the NDIR CO 2 sensor. (a) Test results of 702 ppm CO 2 detection with the seed respiration CO 2 detection system and NDIR CO 2 sensor; (b) residual of the seed respiration CO 2 detection system at 702 ppm; (c) residual of NDIR CO 2 sensor at 702 ppm. seeds changes faster, and the respiration rate changes of waxy corn seeds are more uniform. Te experiments also show that the seed respiration CO 2 detection system can achieve high-resolution continuous observation of a small amount of seed respiration and show detailed information such as the change trend of seed respiration intensity and the details of respiration rate changes. Using this system to measure the respiration intensity and respiration rate of diferent varieties of seeds to observe their respiration process, we can summarize and analyze the respiration changes of diferent varieties seeds and further study physiological and biochemical issues such as seed vigor.

Conclusions
In this study, a sensor system of a double-layer seed respiration device based on Herriott cell is designed independently and creatively, which solves the problem of unable to detect seed respiration with high precision in real time. Te sensor has a volume of about 1.5 L and an efective optical path of about 21 meters. Using a 2004 nm laser as the light source, a set of the seed respiration CO 2 detection system based on WMS-TDLAS technology was built. Te theoretical calculation of this system was done under diferent temperatures and pressures, and the concentration inversion correction equation was obtained when the temperature and pressure changed. Te laser output has good stability and voltage fuctuation range is about 1.7%. Te system linearity can reach about 0.999, and the detection limit is 1.23 ppm at 51 s. Compared with the commonly used NDIR CO 2 sensor and gas chromatography, our system has higher measurement frequency, lower detection limit, and better stability. Finally, we used the developed system to test the sweet corn seeds and waxy corn seeds respiration intensity of a small amount (1 g) and calculated their respiration rate. Te test results showed that our system can show detailed information clearly, such as the change trend of seed respiration intensity and respiration rate. Te seed respiration CO 2 detection system realizes high-precision continuous online monitoring of seed respiration and has broad application prospects in the felds of seed respiration detection and rapid nondestructive detection of seed vigor.

Data Availability
Te data used to support the fndings of this study are available from the corresponding authors upon request.

Conflicts of Interest
Te authors declare that there are no conficts of interest.