Strawberry anthracnose disease identification system based on hyperspectral analysis

Abstract

This invention discloses a strawberry anthracnose identification system based on hyperspectral analysis, comprising: a hyperspectral imaging device configured to photograph strawberry crops to obtain hyperspectral images; an image processing module comprising: a background removal unit configured to remove the background of the hyperspectral image using the vegetation index ExGRadj to generate a background-removed image; a selection unit connected to the background removal unit configured to select the region to be analyzed from the background-removed image; a feature analysis unit connected to the selection unit configured to perform hyperspectral image analysis on the region to be analyzed to extract hyperspectral feature values ​​of the crop; and a feature classification module connected to the image processing module and comprising a crop disease feature classification model configured to classify the crop according to the hyperspectral feature values ​​to determine whether the crop has a disease.

Claims

1. A strawberry anthracnose identification system, comprising: a hyperspectral imaging device configured to capture a strawberry crop to obtain a hyperspectral image of the strawberry crop; an image processing module comprising: a background removal unit connected to the hyperspectral imaging device and configured to remove the background of the hyperspectral image of the strawberry crop using a vegetation index ExGRadj and ExGR to generate a background-removed image of the strawberry crop, wherein ExGRadj = ExGB*1ExGR>-1*1y>180, and y is a luminance component; a selection unit connected to the background removal unit and configured to select an area to be analyzed from the background-removed image of the strawberry crop; a feature analysis unit connected to the selection unit and configured to perform hyperspectral image analysis on the area to be analyzed to extract hyperspectral feature values ​​of the strawberry crop; and a feature classification module connected to the image processing module and comprising a crop disease feature classification model, the crop disease feature classification model being configured to classify the strawberry crop according to the hyperspectral feature values, thereby determining whether the strawberry crop has a disease. The disease in question is strawberry anthracnose.

2. The strawberry anthracnose identification system as described in claim 1, wherein the hyperspectral feature values ​​are multiple reflectance values ​​measured at an effective spectral wavelength.

3. The strawberry anthracnose identification system as described in claim 1, wherein the hyperspectral feature values ​​are multiple reflectance values ​​measured at an effective spectral wavelength, the multiple reflectance values ​​reflecting the difference in health status of strawberry crops with and without disease, the degree of reaction of high and low bright spots of strawberry crops with disease, the degree of damage of strawberry crops with disease, and the texture characteristics of strawberry crops under healthy conditions.

4. The strawberry anthracnose identification system as described in claim 1, wherein the feature classification module includes a training unit configured to perform machine learning on a training dataset consisting of a plurality of the hyperspectral feature values ​​to establish the crop disease feature classification model.

5. The strawberry anthracnose identification system as described in claim 4, wherein the feature classification module further includes a numerical balancing unit, and each of the plurality of hyperspectral feature values ​​is classified as representing strawberry crop infected with the disease or representing strawberry crop not infected with the disease, thereby generating a first number representing strawberry crop infected with the disease and a second number representing strawberry crop not infected with the disease in the training dataset, wherein the numerical balancing unit is configured to adjust the difference between the first number and the second number within a certain range to generate an optimized training dataset for the training unit to perform machine learning.

6. The strawberry anthracnose identification system as described in claim 1, wherein at least a portion of the strawberry crop is green.

7. The strawberry anthracnose identification system as claimed in claim 1, wherein the hyperspectral imaging device is configured to photograph the strawberry crop at spectral wavelengths between 400 nm and 1000 nm.

8. The strawberry anthracnose identification system as described in claim 1, wherein the crop disease feature classification model is configured to classify the strawberry crop based on the hyperspectral feature values ​​and through one or more of a group consisting of random forest, support vector machine, and Xgboost, thereby determining whether the strawberry crop has a disease.

9. The strawberry anthracnose identification system as described in claim 1 further includes an early warning module connected to the feature classification module, configured to issue an alert message when the feature classification module determines that the strawberry crop has a disease.

Images