Aerial photography small target rapid identification method in extra-high voltage environment evaluation

Abstract

The invention belongs to the field of image information processing, and mainly relates to deep learning target identification, in particular to an aerial photography small target rapid identificationmethod in extra-high voltage environment evaluation. The method is characterized by comprising the following steps: 1, establishing an aerial photography small target identification network model AerialNet in the extra-high voltage environment; 2, training a constructed AerialNet network model from the image training set; and 3, inputting a to-be-detected original image into the trained ArialNet network model to obtain a target category and position coordinates in the to-be-detected image. According to the method, the small target recognition network model AerialNet is constructed, and a residual error network feature enhancement module is designed; and therefore, loss of the feature information of the small target after convolution can be effectively prevented, and rich features of the small target can be retained and enhanced.

Description

technical field [0001] The invention belongs to the field of image information processing, and mainly relates to deep learning target recognition, specifically, a method for quickly identifying small targets in aerial photography in UHV environmental evaluation, and a network structure AerialNet for small target recognition in aerial images for constructing a deep learning model. Background technique [0002] In recent years, drones have played an increasing role in modern military and civilian fields. Aerial target detection has always been an important part of UAV research, and its importance is self-evident that it can complete specified monitoring and reconnaissance tasks in different environments. UHV power grid construction is facing more and more ecological constraints. Therefore, ensuring compliance with laws and regulations, quickly promoting the implementation of power grid projects, and avoiding subversive factors in the subsequent design and construction stages h...

AI Technical Summary

Problems solved by technology

[0005] At present, these two types of network models have their own advantages. The convolutional neural network model based on the candidate area has a high detection accuracy, but the steps are complicated, time-consuming, and the extracted feature map is too large, so it is difficult to effectively detect small targets. Feature extraction cannot accurately detect small targets in the image; the convolutional neural network based on regression is fast in target detection, but its network structure layer is deep, and the loss of small target information in the process of convolution and sampling is increasing. The more , the poorer the detection of small objects in the image

Images