Ancher-free remote sensing image target detection method and system based on scene enhancement

Abstract

The invention discloses an anchor-free remote sensing image target detection method and system based on scene enhancement, and the method comprises the following steps: 1, carrying out the linear enhancement of an obtained remote sensing image data set in a balance coefficient hybrid enhancement mode, and obtaining an enhanced training set; 2, constructing a target detection model based on scene enhancement, training the target detection model through the training set obtained in the step 1 until a preset stop condition is met, and obtaining a trained target detection model, wherein the trained target detection model is used for remote sensing image target detection. According to the method, a more convenient and robust balance coefficient hybrid enhanced data augmentation mode is provided, the feature extraction capability and the category prediction capability of the network are enhanced by using scene information, and the detection precision is improved.

Description

technical field [0001] The invention belongs to the technical field of remote sensing image processing and target detection, in particular to an anchor-free remote sensing image target detection method and system based on scene enhancement. Background technique [0002] Target detection in remote sensing images is to detect the existing target categories in high-resolution aerial images and to give the position information of each target. In recent years, object detection tasks in the field of remote sensing images have gained more and more application scenarios; for example, in urban planning, UAV detection, intelligent monitoring, etc. Although traditional detection methods such as DPM are still used in this field, CNN-based deep learning methods have gradually gained a dominant position; He Kaiming et al. proposed FasterR-CNN, FPN, MaskR-CNN, Joseph proposed YOLOv1, YOLOv2, Algorithms such as YOLOv3 have achieved great success in the field of natural environment image ta...

AI Technical Summary

Problems solved by technology

Some scholars are also exploring its reasonable application in target detection tasks, such as the context model feature pyramid (FPN), which has obtained some results and conclusions, but its methods are limited to specific sub-tasks, which are difficult to apply to anchor- In the free target detection method

[0004] Data enhancement plays an important role in image processing and deep learning models. In the field of target detection, commonly used data enhancement methods include horizontal-vertical flipping, random cropping, scale transformation, elastic distortion, etc., but they are all limited to specific data sets. , it is difficult to widely adapt to remote sensing image datasets

In addition, most of the existing data enhancement methods will increase the number of samples in the training set, which will lead to problems such as long model training time and complex label conversion.

Images