A method of tea cicada recognition and counting based on convolutional neural networks

Abstract

The invention provides a method for identifying and counting the green leafhoppers based on a convolutional neural network. The method uses a VGG16 network structure as a backbone network to perform feature extraction of green leafhoppers containing yellow plates, and selects shallow features. The regression operation is performed, and the deep features of the rich semantic information of the backbone network are superimposed layer by layer on the shallow features through the combination of upsampling and convolution layers, thereby enriching the semantic information of the shallow features and avoiding the The image information is lost, and the automatic identification, position regression and automatic calculation of the tea leafhopper are completed; the identification of the tea leafhopper by this method can reach 98% in the case of 1% missed detection; at the same time, for high-definition pictures, If the processing speed of 1080P data reaches the ms level, it can quickly and accurately identify and calculate the small green leafhoppers, and can be widely used in the prevention and control of tea small green leafhoppers such as tea trees and fruit trees.

Description

Technical field [0001] The present invention involves the field of image processing technology, which specifically involves a method of identifying and counting tea -based tea cicada -based tea cicadas based on convolutional neural networks. Background technique [0002] Tea small green leaf cicada, adult body length of 3 to 4mm, body width of 1mm, light yellow green to green, compound eye gray -brown to dark brown, no single eye, tentacle rigid hair shape, black at the end; resemblance. Tea small green leaf cicada, nymphs are the main juice of piercing tea tree leaves, and inhabited on the back of the crop leaf, and the adults are ends after eating, and the eggs are mostly produced under the skin. ~ 17th generation, the generation overlap. Therefore, the tea small green leaf cicada is one of the most widely distributed and harmful parts of the tea districts in my country, and one of the key prevention and treatment of pests that affect the yield and quality of tea. About %, and ...

AI Technical Summary

Problems solved by technology

In the prior art, artificial methods are often used to carry out the monitoring of tea leafhoppers, that is, when the morning dew is not dry in sunny days or cloudy all-day investigations, the number of insect populations on 100 young leaves (30 bud tips) is randomly investigated; but this The method has high requirements for artificial professional knowledge, judgment experience (the green leafhopper is agile and lively, good at climbing and jumping) and the degree of vision (the adult body length of the green leafhopper is shorter, and the nymph body length is shorter), and it needs Manual screening, identification, and counting are time-consuming and labor-intensive. At the same time, the accuracy of this method is low, between 60% and 80%. The fluctuation range and error of each identification are large, and it cannot be completed accurately, quickly, and effectively. Identify and count tea leafhoppers, so as to take efficient and reasonable control measures

Images