A method for recognizing crop images taken by an unmanned aerial vehicle

Abstract

The invention provides a crop image recognition method photographed by an unmanned aerial vehicle. The invention relates to a crop image recognition method photographed by an unmanned aerial vehicle (UAV), which is characterized in that the method comprises the following steps: (1) constructing attribute information of the crop image photographed by the UAV and performing preprocessing to obtain acrop image data set; S2. the convolution neural network model is pre-trained with the idea of transfer learning; S3, fine-tuning the convolution neural network pre-trained in the step S2 by using thecrop image data set obtained in the step S1, extracting features of different layers of the convolution neural network model, and combining the features to obtain image feature representations; S4, classifying the image features obtained in the step S3 by using the SVM classifier, completing the crop image classification, obtaining the classification result, and finally inputting the crop image captured by the unmanned aerial vehicle into the convolution neural network model in the step S3 for recognition. The invention can more effectively identify the target image data by using the labeledsample of the target image under the condition that the image data set is limited.

Description

technical field [0001] The present invention relates to the technical field of image processing and recognition, and more specifically, to a crop image recognition method captured by a drone. Background technique [0002] In recent years, image recognition technology has developed rapidly, especially deep learning has significantly improved the performance of image recognition. Using deep learning to identify crop images taken by drones can effectively promote the development of traditional agriculture to modern agriculture. [0003] However, deep learning requires huge samples to realize model training, and the image data taken by drones is limited, making it difficult to achieve effective training; related research shows that the learned features are closely related to the task of recognition, while traditional The feature recognition algorithm of the convolutional neural network is difficult to meet the needs of the real scene in terms of accuracy, especially the high-le...

AI Technical Summary

Problems solved by technology

[0003] However, deep learning requires huge samples to realize model training, and the image data taken by drones is limited, making it difficult to achieve effective training; related research shows that the learned features are closely related to the task of recognition, while traditional The feature recognition algorithm of the convolutional neural network is difficult to meet the needs of the real scene in terms of accuracy, especially the high-level features are relatively abstract semantic features, and it is easy to lose the detailed information of the image

Images