Unmanned aerial vehicle tree type identification method based on improved SSD learning model

Abstract

The invention discloses an unmanned aerial vehicle tree type identification method based on an improved SSD learning model. The method comprises the following steps: firstly, performing clustering by adopting a K-means + + method, respectively obtaining scale and aspect ratio information of a default candidate box, and providing parameters for a convolutional neural network model in the next step; secondly, constructing a convolutional neural network model, improving a reference SSD learning model, adding a filtering layer capable of adaptively generating different filters for different types of picture noise, adding a global feature fusion network at the last of the model, improving a loss function of the reference SSD learning model, and training the model by using a pre-acquired picture; and finally, detecting tree types based on the improved SSD learning model. According to the method, training is carried out on the tree data set to extract the feature representations of the generalized trees of different types, the distinguishability between the features is highlighted, the small target detection and picture noise resistance capabilities are improved, and the tree type identification accuracy is improved.

Description

technical field [0001] The invention belongs to the field of artificial intelligence target recognition, and relates to the determination of tree types in a UAV aerial photography scene, in particular to a UAV tree type identification method based on an improved SSD learning model. Background technique [0002] Tree species classification is of great value in the field of smart gardens and plays an important role in tree resource management and monitoring and carbon storage estimation. UAV remote sensing is a low-altitude remote sensing technology. Compared with various high-resolution satellites, UAVs are less disturbed by atmospheric factors in the process of acquiring images, and have good prospects in small-area remote sensing applications. The use of drones to obtain images with ultra-high spatial resolution provides a new method for extracting spatial information of tree species, and the technology of accurately and quickly extracting tree species classification inform...

AI Technical Summary

Problems solved by technology

However, the disadvantage of uniform dense sampling is that training is difficult, mainly because of the imbalance between positive samples and negative samples (background), resulting in low model accuracy.

Images