Application Case | Application Of ATH3500 Hyperspectral Imager In Rapid Wheat Year Identification

Preface

Wheat is one of the most widely distributed food crops in the world. In my country, the planting area and output of wheat are second only to rice. The germination rate of wheat seeds is crucial to increasing wheat production and income, and the germination rate of wheat seeds is closely related to the storage year of wheat seeds. Generally speaking, as the storage year increases, the germination rate of wheat seeds decreases, so it is of great significance to identify the storage years of wheat seeds. The traditional method of identifying the storage year of wheat seeds relies on experienced agronomists to smell the seeds and check the quality of the seeds. This method is time-consuming and labor-intensive, and has large errors. Imaging hyperspectral technology combines the advantages of graphics technology and spectral technology, and can simultaneously obtain image information and spectral information that reflect the external characteristics, internal physical structure and chemical composition of the sample to be tested. It has been widely used in the field of non-destructive testing of agricultural products, such as crop information diagnosis, pesticide residue detection, internal and external quality prediction, etc. This time, wheat seeds are taken as an example, and imaging hyperspectral technology is used to identify the year of wheat seeds, providing a new technical reference for the time identification of agricultural products.

Advantages of hyperspectral

Hyperspectral reflects the characteristics of high-resolution optical information, which uses many very narrow electromagnetic wave bands (usually <10nm) to obtain relevant data from objects of interest. Hyperspectral images are acquired by an imaging spectrometer, which provides hundreds of narrow-band spectral information for each pixel, producing a complete and continuous spectral curve. Make surface information of wheat seeds easier to discover.

Hyperspectral has the following characteristics:

(1) There are many spectral features. The imaging spectrometer has 480 bands in the visible and near-infrared spectral regions.

(2) High spectral resolution. The sampling interval of the imaging spectrometer is small, and the resolution is less than 2nm. Fine spectral resolution reflects the subtle characteristics of the spectrum of ground objects.

(3) The amount of data is rich. As the number of bands increases, the amount of data increases exponentially.

(4) It can provide spatial domain information and spectral domain information, that is, "unification of images and spectra", and the spectral curve obtained by the imaging spectrometer can be compared with the measured spectral curve of similar ground objects.

1.Comparison with visible light results

The visible light image only consists of light information in three bands of red, green and blue (RGB). When identifying the year of wheat seeds, it is monitored through artificial naked eye interpretation, which is inefficient and has limited accuracy. The hyperspectral image results have 480 bands of light information. For wheat seed years, unique band characteristics can be extracted to accurately identify them. At the same time, combined with machine learning algorithms, automated monitoring can be achieved.

2.Comparison with multispectral results

The main difference between multispectral and hyperspectral is the number and narrowness of the bands. Multispectral images usually consist of 10-30 bands of light information, which are easily confused with the reflection spectra of other ground objects. Having a higher level of spectral detail in hyperspectral images can provide better vegetation discrimination capabilities. For example, when detecting poppies, hyperspectral images can effectively distinguish poppies from other similar plants and improve accuracy.

1. Multispectral images usually refer to 10 to 30 bands expressed in pixels, and each band can be obtained by using a remote sensing radiometer.

2. Hyperspectral images contain very narrow bands, and hyperspectral images have hundreds of bands (for example, ATH8010 has 480 bands).

Imaging hyperspectral equipment

This test uses wheat seeds as the research object, and uses a hyperspectral imager (spectral range 400-1000nm) to collect hyperspectral data of the test objects to identify wheat seeds of different years. The system is mainly composed of hyperspectral cameras, light sources, obscura, stage and other accessories. The original platform is equipped with a uniform sliding conveyor belt device to ensure that the movement rate of the sample to be tested matches the acquisition frame rate of the camera. Process analysis in real time. The actual scene is shown in Figure 1.

Figure 1 ATH3500 teaching imaging spectrometer.

Data collection and analysis

4.1 Image preprocessing

The original image data of wheat seeds collected by the imaging hyperspectral imager are preprocessed. The preprocessing process mainly includes two parts: the first part is wavelength calibration, the second part is reflectance calibration.

The first is wavelength calibration. When collecting data, each hyperspectral camera has its corresponding calibration file. You need to import the calibration file to the wavelength value in ENVI to convert the wavelength of data collection into the actual wavelength of the camera.

The second step is reflectance calibration. The formula for reflectance calibration is as follows:

Reftarget=DNtarget/DNpanel×Refpanel

In the formula: Reftarget is the reflectance of the target object; Refpanel is the reflectance of the standard reference plate; DNtarget is the value of the target object in the original image; DNpanel is the value of the standard reference plate in the original image.

4.2 Data analysis and results

Table 1 shows the RGB composite hyperspectral image of wheat seeds in different years.

Since hyperspectral imaging data contain both spatial and spectral information, each pixel on the wheat seed corresponds to a spectrum. Figure 2 shows the reflectance spectra of wheat seeds in different years after data preprocessing. According to the reflection spectrum characteristics of wheat seeds, the spectral trends obtained in different years are basically consistent. In the entire wavelength range, the spectral curves of wheat seeds in each year are relatively smooth. There are three obvious differences near 500 and 700nm. These two bands are used to construct the NDVI spectral index. The calculation formula of NDVI is as follows:

Figure 2 Schematic diagram of spectral curves of wheat seeds in different years.

Figure 3 Changes in spectral characteristics of wheat NDVI in different years.

(2) By analyzing the curve changes of the NDVI spectral index and its corresponding year, it can be seen that, except for the wheat seeds in the abnormal year 2020, as the year of wheat seed collection increases, from the overall change trend, its NDVI value decreases. A comprehensive analysis of the change curve between NDVI of wheat seeds and years shows that the changes in wheat seeds in 2020 tend to be different from the overall changes. Therefore, when analyzing the relationship between NDVI and years, the interference of abnormal year data on the data analysis results is deleted .

(3) By establishing a scatter plot and trend line of the NDVI spectral index and its corresponding year, it can be seen that the coefficient of determination of NDVI and its corresponding year is 0.9854, and the modeling model is y=-100.62x+2068.69 (where y is the year, x is the NDVI value). Judging from the coefficient of determination of the prediction model, the model constructed with NDVI and its corresponding year has higher accuracy.

Conclusion

The wheat seed year identification analysis based on hyperspectral images uses the obtained NDVI modeling model y=-100.62x+2068.69 to predict the wheat seed years in 2019, 2021, and 2022. The predicted years and actual years are shown in Figure 5 As shown in the figure, it can be seen from the figure that the predicted year is rounded, and the predicted year of the model constructed by NDVI is completely consistent with the actual year. Therefore, the corresponding year of wheat seeds can be quickly identified through hyperspectral data.

Figure 4 Wheat seed forecast.