Nighttime Vehicle Object Detection Based on Improved YOLOv7

Detecting vehicles using deep learning is a crucial area of research in computer vision. Detecting vehicles at night can be challenging due to factors such as low illumination, complex lighting, and other environmental conditions, which can negatively impact the accuracy of detection. As a result, achieving high detection accuracy for night vehicle images can be difficult. Therefore, this paper combines image enhancement algorithm and object detection algorithm to study vehicle object detection in night scene. Firstly, Laplacian sharpening algorithm is applied to enhance the images. Then, the improved YOLOv7 algorithm is applied to detect the enhanced images. Compared with the original YOLOv7 on the self-made night vehicle dataset, improved YOLOv7 only has 1.6% higher parameters, but has 2.5% less computation, and brings 1.9% higher mAP0.5 and 1.7% higher mAP0.5:0.95.