In order to solve the problem of occluded the vehicle ahead during driving, pedestrians and other targets are small, and it is prone to problems such as missed detection, false detection. This paper proposes an improved network model based on YOLOv4_Tiny. Before training the model, data augmentation is performed by means of image segmentation. Increase the feature scale of the model, and determine the size of the final increased feature scale through experimental comparison. The attention mechanism SE module is introduced in the enhanced (Feature Pyramid Networks) FPN of YOLOv4_Tiny. The test results show that compared with YOLOv4_Tiny, the (average precision) AP of YOLONew_Tiny in large target trucks, cars, and small target pedestrians is improved by 0.73%, 3.85% and 12.95% respectively. The (mean Average Precision) mAP is improved by 5.48%. This shows that the improved YOLOv4_Tiny network model can effectively improve the detection ability of forward targets during driving.
Research on Unmanned Vehicle Detection Method Based on Improved YOLOv4_Tiny
25.09.2022
4761584 byte
Aufsatz (Konferenz)
Elektronische Ressource
Englisch
A Lightweight Anti-Unmanned Aerial Vehicle Detection Method Based on Improved YOLOv11
DOAJ | 2024
|UNMANNED AERIAL VEHICLE DETECTION SYSTEM AND UNMANNED AERIAL VEHICLE DETECTION METHOD
Europäisches Patentamt | 2019
|UNMANNED AERIAL VEHICLE DETECTION SYSTEM AND UNMANNED AERIAL VEHICLE DETECTION METHOD
Europäisches Patentamt | 2018
|UNMANNED AERIAL VEHICLE DETECTION SYSTEM AND UNMANNED AERIAL VEHICLE DETECTION METHOD
Europäisches Patentamt | 2019
|Europäisches Patentamt | 2021
|