Target detection in UAV aerial images has become one of the hot topics in recent years. Traditional detection methods have difficulty meeting both light weighting and accuracy requirements. This paper proposes a new target detection algorithm to address the issue of insufficient light weighting: using EficientNet as the main backbone network, replacing the backbone network of YOLOX, continuing to use YOLOX’s detection head and neck, and using this algorithm on the drone aerial dataset VisDrone. It was found that the average accuracy and performance are basically the same as other models, greatly improving the computational speed and efficiency of the model. Compared with other backbone networks added to YOLOX, the mAP value is improved by at least 3.4%. Compared with other advanced target detection algorithms, mAP value increased by at least 1.2%. When processing UAV target identification tasks, the YOLOX model based on EfficientNet-B0as the main trunk network has a strong recognition ability for the image targets whose color is similar to the background color, relatively fuzzy and small scale.
Target recognition algorithm for UAV aerial images based on improved YOLO-X
2023-10-11
2653788 byte
Conference paper
Electronic Resource
English
Small Target Detection Algorithm in Remote Sensing Image Based on Improved Yolo
Springer Verlag | 2021
|Ship detection based on improved YOLO algorithm
IEEE | 2023
|LI-YOLO: An Object Detection Algorithm for UAV Aerial Images in Low-Illumination Scenes
DOAJ | 2024
|