Garden nursery stock intelligent detection and counting method based on UAV and convolutional neural network

Abstract

The invention belongs to the technical field of garden plant intelligent monitoring, and discloses a garden nursery stock intelligent detection counting method based on UAV and a convolutional neuralnetwork, and the method comprises the steps: collecting garden seedling image data through an unmanned plane under the conditions of sufficient illumination, no wind or breeze; screening the acquiredimage data according to the picture definition, and deleting an unclear image or an image with too dark light; cutting the image data, and unifying image color representation at the same time; preprocessing the image data; utlizing a Mask RCNN deep learning framework based on image segmentation to construct a convolutional neural network oriented to garden nursery stock intelligent detection and counting; and carrying out garden nursery stock intelligent detection and counting by utilizing the constructed convolutional neural network for garden nursery stock intelligent detection and counting.According to the invention, unmanned aerial vehicle remote sensing and the convolutional neural network are combined to carry out intelligent detection and counting of garden nursery stocks, and theblank of domestic intelligent detection and counting of garden nursery stocks can be filled.

Description

technical field [0001] The invention belongs to the technical field of intelligent monitoring of garden plants, in particular to an intelligent detection and counting method for garden seedlings based on a UAV and a convolutional neural network. Background technique [0002] Landscaping plants are rich in tree species and are the main components of urban greening systems. With the development of urbanization, the area of ​​garden seedling bases in my country has been expanding year by year. However, there is no accurate estimate of the distribution and number of garden seedlings. Due to the lack of sufficient attention, the intelligent detection of garden seedlings has been neglected for a long time, and there are few reports on the intelligent detection and counting of garden seedlings at home and abroad. In recent years, the maturity of drone technology and deep learning algorithms has provided conditions for the detection and counting of garden seedlings. Therefore, it i...

AI Technical Summary

Problems solved by technology

There are many studies on deep learning algorithms and UAV applications, but there are few reports on large-scale intelligent detection and counting of garden seedlings combined with UAVs and convolutional neural networks

[0004] The existing garden seedling detection and counting methods are mainly traditional manual field surveys, which adopt the method of counting while marking to avoid repetition and omission of manual counting. Although this method has high accuracy of counting results, it is time-consuming, labor-intensive and inefficient. Low

[0005] Through the above analysis, the problems and defects of the prior art are: (1) the traditional survey of garden seedlings adopts the method of manual on-the-spot investigation, which is slow and not suitable for large-scale rapid detection; (2) in order to obtain samples of garden seedlings in different periods The data needs to be detected multiple times, and the feasibility is low; (3) the detection cost is high, and it is not suitable for multiple detection and counting of large-scale garden seedlings

[0006] The difficulty in solving the above problems and defects is: Due to the limitations of the growth characteristics of garden seedlings, the lighting conditions of garden seedling bases, the background factors of UAV aerial photography, and the characteristics of selected deep learning algorithms, the high-definition photos of garden seedlings taken by UAV aerial photography cannot be directly for training

Images