Server of crop growth stage determination system, growth stage determination method, and storage medium storing program
A server of a crop growth stage determination system includes a processor. The processor inputs first images obtained by image capturing crops in a manner such that crop shapes are extractable. The Processor inputs growth stages each indicating a level of physiological growth of the crops for each of the first images. The processor constructs a learned model by performing deep learning to associate images of the crops and growth stages of the crops based on the input first images and the input growth stage. The processor inputs a second image obtained by image capturing crops a growth stage of which is unknown, in a manner such that crop-shapes are extractable. The processor determines the growth stage for the input second image based on the learned model. The processor outputs the determined growth stage.
1. A server of a crop growth stage determination system, the server comprising a processor configured to:
input first images obtained by image capturing crops in a manner such that crop shapes are extractable based on satisfying preset image-capturing conditions;
input growth stages each indicating a level of physiological growth of the crops for each of the first images;
construct a learned model by performing deep learning to associate images of the crops and growth stages of the crops based on the input first images and the input growth stage;
input a second image obtained by image capturing crops a growth stage of which is unknown, in a manner such that crop-shapes are extractable based on satisfying the preset image-capturing conditions;
determine the growth stage for the input second image based on the learned model; and output the determined growth stage,
wherein the crops are paddy rice, and
the processor performs deep learning to associate images, from which shapes of the paddy rice are extractable, with growth stages including panicle initiation, the panicle initiation being a time point at which differentiation of a young panicle wrapped in a leaf sheath starts.
2. The server of the crop growth stage determination system according to claim 1 , wherein
the processor excludes at least one of the input first image or the input second image that does not satisfy the preset image-capturing conditions.
3. The server of the crop growth stage determination system according to claim 1 , wherein
the processor inputs the first images for which each of an image capturing time period is restricted.
4. The server of the crop growth stage determination system according to claim 1 , wherein
the processor inputs at least one of the input first image or the input second image that is captured in accordance with an application program for displaying a guide for a composition of an image of crops to be captured.
5. The server of the crop growth stage determination system according to claim 1 , wherein
the processor divides the second image into a plurality of regions from which the crop shapes are individually extractable, determines the growth stage for each of the regions based on the learned model, and determines the growth stage of the second image from combined determination results of the regions.
6. The server of the crop growth stage determination system according to claim 5 , wherein
the processor sets a satisfaction criterion for the determination result of the divided regions, and determines the growth stage of the second image without including a determination result that does not meet the satisfaction criterion in the determination result of the regions.
7. The server of the crop growth stage determination system according to claim 1 , wherein
the first images and the second image are gray-scale images.
8. The server of the crop growth stage determination system according to claim 1 , the server further comprising:
a storage configured to store as a history record the growth stage determined by the processor,
wherein the processor outputs, based on the determined growth stage and the history record stored in the storage, at least one of the number of days that have passed in the determined growth stage, or the number of days that remain until a next growth stage.
9. A crop growth stage determination method, comprising:
inputting first images obtained by image capturing crops in a manner such that crop shapes are extractable are input based on satisfying preset image-capturing conditions;
inputting growth stages each indicating a level of physiological growth of the crops is input for each of the first images;
constructing a learned model by performing deep learning to associate images of the crops and growth stages of the crops based on the input first images and the input growth stage;
inputting a second image obtained by image capturing crops a growth stage of which is unknown, in a manner such that crop-shapes are extractable based on satisfying the preset image-capturing conditions:
determining the growth stage for the input second image based on the learned model; and
outputting the determined growth stag;
wherein the crops are paddy rice, and
the method includes performing deep learning to associate images, from which shapes of the paddy rice are extractable, with growth stages including panicle initiation, the panicle initiation being a time point at which differentiation of a young panicle wrapped in a leaf sheath starts.
10. A non-transitory storage medium storing a program implemented by a processor, the program causing the processor to:
input first images obtained by image capturing crops in a manner such that crop shapes are extractable are input based on satisfying preset image-capturing conditions;
input growth stages each indicating a level of physiological growth of the crops is input for each of the first images;
construct a learned model by performing deep learning to associate images of the crops and growth stages of the crops based on the input first images and the input growth stage;
input a second image obtained by image capturing crops a growth stage of which is unknown, in a manner such that crop-shapes are extractable based on satisfying the preset image-capturing conditions:
determine the growth stage for the input second image based on the learned model; and
output the determined growth stage;
wherein the crops are paddy rice, and
the processor performs deep learning to associate images, from which shapes of the paddy rice are extractable, with growth stages including panicle initiation, the panicle initiation being a time point at which differentiation of a young panicle wrapped in a leaf sheath starts.