Multi-object Recognition Method Based on Improved YOLOv2 Model

A method of vehicle multi-object identification and classification based on the YOLOv2 algorithm is proposed to solve problems low detection rate, poor robustness, unsatisfactory effect for classical type real road environment. Based algorithm, network structure YOLOv2-voc improved according actual conditions. The training model was obtained ImageNet data fine-tuning technology, analysis results object characteristics. This paper structure, namely called YOLOv2-voc_mul. In order verify validity method, experiments are performed using samples from simple backgrounds complex compared with existing YOLOv2, YOLOv2-voc, YOLOv3 models after 70000 iterations, respectively. show that YOLOv2-voc_mul has an accuracy 98.6% under background, mAP (mean Average Precision) different reaches 87.81%. Under average 92.09% 89.64% single four models. summary, our better accuracy, a false good robustness.