Influenced by the ecological environment and other factors, the occurrence of diseases and insect pests in lotus root will seriously affect the quality and yield of lotus root in recent years. With the improvement of living standards and the development of lotus industry chain, people"s demand for green, high-yield and high-quality products is increasing. At present, many planters are unable to accurately identify the diseases and pests of lotus due to lack of professional knowledge of disease and pest control. There is short of efficient and low-cost automatic identification technology for the prevention and control of lotus diseases and pests. The early diagnosis and identification of diseases and insect pests is of great significance for the control of diseases and insect pests in lotus root fields. In order to efficiently and accurately realize the early identification and diagnosis of lotus leaf diseases and insect pests, this paper proposes a lotus leaf disease and insect pests identification model based on Improved DenseNet. Firstly, this paper compares different dynamic data enhancement methods. The result shows that the accuracy of resizing and randomly resizing the image is higher than directly resizing to the same size. The loss of detail information in part of the image is caused by resizing the image, which affects the model recognition effect. The accuracy of the model was improved by 3.54% compared with 81.47% by using the data enhancement method of Resize, Random Resized Crop, Random Horizontal Flip and Random Adjust Sharpness. Adamax optimizer was used to replace Stochastic Gradient Optimization optimizer, the accuracy of DenseNet model on lotus leaf disease and pest data set has been improved by 3.04%. DenseNet model is adopted as backone The Stem module uses multi-layer small convolution for fast dimensionality reduction, and uses a branch structure to combine convolution and maximum pooling, which improves the ability of the model to extract shallow features at a lower operating cost. The Squeeze and Exitattention mechanism module and sharpened cosine similarity convolution are introduced in the Denselayer layer of the Dense Block and the Transition Layer. This method improves the recognition ability of the model to lotus leaf disease, and verifies the effectiveness of sharpened cosine convolution to improve the performance of the model. Based on the Plantvillage dataset, migration learning achieved 91.34% recognition accuracy, which was 9.87% higher than before improvement and optimization. We conducted migration learning on the Plantvillage dataset. The accuracy of the improved model was 91.34%, which was 9.87% higher than that before improvement and optimization. The improved model is applied to the identification of lotus field diseases and insect pests in UAV images. In this paper, the calibration area of lotus leaf is cut and predicted by reasoning, then different masks are generated according to the model prediction results and added to the UAV image to generate the distribution map of lotus field diseases and insect pests, realizing the recognition of lotus field diseases and insect pests in the UAV image. This paper realizes automatic classification and recognition of l lotus leaf rot, virus disease, spodoptera litura, leaf rot and leaf spot disease. It provides a new method for efficient and accurate identification and dynamic monitoring of lotus root diseases. It provides information support for variable pesticide application and flight path planning in plant disease prevention and control based on UAV.