Deep neural network-based plant protection strategy in rural garden landscape construction

Abstract

Rural landscaping enhances the quality of the environment in rural areas by planting trees and flowers and plants and at the same time can contribute to the protection of plants in rural areas. The occurrence and development of landscape plant diseases are very insidious, which may lead to poor tree growth and affect the annual growth of timber, or even lead to the death of the whole forest and the reverse succession of the forest. Plant leaf disease lesion location is random, and plant leaf image shooting is susceptible to light, background, angle and other objective factors can only rely on artificial identification, but the traditional artificial identification and physiological information extraction methods such as low accuracy, low efficiency and cumbersome. The rapid development of deep neural networks provides an effective solution for plant leaf recognition based on image analysis. In this paper, a DCA-ResNet model is constructed based on ResNet101 to recognize plant leaves by introducing the attention mechanism and optimizing the model structure. The validation of the DCA-ResNet model using public datasets on three species of plant leaves and plant leaf diseases achieved accuracy rates of 96.79%, 96.64%, and 96.44%, respectively.