| CPC A01B 79/005 (2013.01) [G06T 7/0012 (2013.01); G06V 10/82 (2022.01); G06V 20/188 (2022.01); G06N 3/0464 (2023.01); G06Q 10/04 (2013.01); G06Q 50/02 (2013.01); G06T 9/002 (2013.01); G06T 2207/20081 (2013.01); G06T 2207/20084 (2013.01); G06T 2207/30188 (2013.01); G06V 10/40 (2022.01); G06V 10/764 (2022.01); G06V 10/774 (2022.01); G06V 10/80 (2022.01); G06V 20/13 (2022.01); G06V 40/172 (2022.01)] | 6 Claims |
|
1. A crop yield estimation method based on grade identification and weight decision-making, comprising:
acquiring a visible light image of a target crop field at maturity by using a remote sensing device;
inputting the visible light image of the target crop field at maturity into a yield grade classification model, to obtain yield grade output values corresponding to all yield grades of the visible light image of the target crop field at maturity, wherein the yield grades are determined according to a yield grade rule, the yield grade rule is that a difference between adjacent yield grades is not greater than a product of a predetermined percentage and an average value of all known sample crop yields, and the number of yield grades is not less than a set number;
performing normalization on each of the yield grade output values to obtain a confidence score corresponding to each of the yield grade output values, and sorting the confidence scores in descending order;
selecting, from the confidence scores sorted in descending order, a first m confidence scores, and performing normalization on the first m confidence scores to obtain m yield grade weights wherein m is a predetermined positive integer not less than 3; and
respectively multiplying the m yield grade weights and corresponding yield grades, and adding up all products, to obtain an estimated yield of the target crop field at maturity, wherein:
the yield grade classification model is obtained by training a deep convolutional neural network based on sample data, the sample data comprises sample input data and corresponding label data, the sample input data is an image of a sample crop field at maturity, and the label data is a yield grade of the image of the sample crop field at maturity;
wherein the inputting the visible light image of the target crop field at maturity into the yield grade classification model, to obtain yield grade output values corresponding to all yield grades of the visible light image of the target crop field at maturity comprises:
stitching and orthorectifying the visible light image of the target crop field at maturity based on image features, to obtain preprocessed visible light images, wherein the image features at least comprise photo heading overlap features and side overlap features;
clipping, based on a boundary of the target crop field at maturity, the preprocessed visible light image, to obtain a plurality of images of an initial target crop field at maturity;
performing random sampling for equal size and no repetition on the plurality of images of the initial target crop field at maturity, to obtain images of the target crop field at maturity after sampling;
performing data format conversion on each of the images of the target crop field at maturity after sampling, to obtain a plurality of images of the target crop field at maturity after data format conversion; and
inputting each of the images of the target crop field at maturity after data format conversion into the yield grade classification model, to obtain yield grade output values corresponding to all yield grades of each image of the target crop field at maturity after data format conversion.
|