电子信息竞赛中的机器人目标检测技术研究

Research on robot target detection technology in electronic information contest

互联网智能技术促使以机器人为主体的电子信息竞赛开始出现,在电子信息竞赛中,传统的机器人目标检测技术的速度较慢、准确率较低。因此,研究提出了在深度学习基础上构建的YOLOv8算法,并对其轻量化版本YOLOv8-Tiny算法进行了改进,提出了SS-YOLO算法,同时利用实验对其性能进行了验证。实验结果表明,SS-YOLO的MAP最高可达85%,远高于YOLOv8-Tiny的68%;SS-YOLO算法呈现出57.1帧的FPS值和32.8 MB的模型大小。同时,SS-YOLO在7次召唤率范围内的准确率最高为100%,其平均准确率高达90%,远高于YOLOv8-Tiny的75%。综合来看,研究提出的SS-YOLO算法具备更高的检测速度和精度,同时在同等情况下模型更小,在实际的电子信息竞赛中具有较强的实用性和有效性。

Internet intelligence technology has promoted the emergence of robot-based electronic information competition. In the electronic information competition, the traditional robot target detection technology is slow and has low accuracy. Therefore, YOLOv8 algorithm based on deep learning is proposed, and its lightweight version YOLOv8-Tiny algorithm is improved, and SS-YOLO algorithm is proposed, and its performance is verified by experiments. The experimental results show that the maximum MAP of SS-YOLO can reach 85%, far higher than 68% of YOLOv8-Tiny;The SS-YOLO algorithm presents an FPS value of 57.1 frames and a model size of 32.8 MB. At the same time, SS-YOLO has the highest accuracy rate of 100% in the range of seven summoning rates, and its average accuracy rate is as high as 90%, far higher than 75% of YOLOv8-Tiny. In general, the SS-YOLO algorithm proposed in the study has higher detection speed and accuracy, and the model size is smaller under the same conditions, which has strong practicability and effectiveness in the actual electronic information competition.