Abstract: Provided is a method and system for determining plant leaf surface roughness. The method includes: acquiring a plurality of continuously captured zoomed-in leaf images by using a zoom microscope image capture system; determining a feature match set according to the zoomed-in leaf images; removing de-noised images of which the number of feature matches in feature match set is less than a second set threshold, to obtain n screened images; combining the n screened images to obtain a combined grayscale image; and determining plant leaf surface roughness according to the combined grayscale image. In the present disclosure, first, a plurality of zoomed-in leaf images are directly acquired by the zoom microscope image capture system quickly and accurately; the zoomed-in leaf images are then screened and combined to form a combined grayscale image; finally, three-dimensional roughness of a plant leaf surface is determined quickly and accurately according to the combined grayscale image.

Abstract: A method and a system for identifying a field boundary are provided. The method includes pre-processing acquired original images to obtain an image data set; training an improved U-Net network to obtain a segmentation model of field regions by taking images in the image data set as inputs, and marked images corresponding to the input images as outputs; wherein the marked images are binary images for distinguishing the field regions from non-field regions, an inverted residual block is set between maximum pooling layers of an encoder of the improved U-Net network, and an inverted residual block is set between deconvolutional layers of a decoder of the improved U-Net network; inputting a current frame image into the segmentation model, outputting a current binary image, and determining a field-side boundary line and a field-end boundary line based on a frame association algorithm and an RANSAC line fitting method.

Abstract: Provided is a method and system for determining plant leaf surface roughness. The method includes: acquiring a plurality of continuously captured zoomed-in leaf images by using a zoom microscope image capture system; determining a feature match set according to the zoomed-in leaf images; removing de-noised images of which the number of feature matches in feature match set is less than a second set threshold, to obtain n screened images; combining the n screened images to obtain a combined grayscale image; and determining plant leaf surface roughness according to the combined grayscale image. In the present disclosure, first, a plurality of zoomed-in leaf images are directly acquired by the zoom microscope image capture system quickly and accurately; the zoomed-in leaf images are then screened and combined to form a combined grayscale image; finally, three-dimensional roughness of a plant leaf surface is determined quickly and accurately according to the combined grayscale image.