基于深度学习的道路车辆目标检测系统设计

针对现有道路车辆目标检测系统由于计算量过大，且在复杂背景下容易出现误检的问题，设计了一种基于深度学习的道路车辆目标检测系统；系统硬件主要由信号输入模块、控制模块、中央处理模块和输出通道模块四部分组成；引入XCV50E芯片构建控制模块，通过高速随机存取存储器(RAM)快速分配信号；利用TMS320C6202芯片设置中央处理模块，将PCI9054作为信号输出模块的核心设备；软件设计中，完成用户登录、数据采集及处理、模型训练等设计；引入深度学习策略，先采用直方图均衡法处理环境光线干扰的问题，然后建立改进YOLOv4模型，处理噪声等干扰信息，最后基于注意力机制完成图像特征提取优化，进而更精准地完成道路车辆目标检测；实验结果表明，所提系统具有很好的鲁棒性，能够很好地减少计算量，提高检测准确率。

Design of Road Vehicle Target Detection System Based on Deep Learning

Aiming at the problem that the existing road vehicle target detection system is prone to misdetection due to excessive computation and complex background, a road vehicle target detection system based on deep learning is designed. The system hardware is mainly composed of four parts: signal input module, control module, central processing module and output channel module. Introduce the XCV50E chip to build a control module, and quickly distribute signals through high-speed RAM. Use TMS320C6202 chip to set up the central processing module, and take PCI9054 as the core device of the signal output module. In the software design, the design of user login, data acquisition and processing, model training, etc. is completed. The deep learning strategy is introduced. First, the histogram equalization method is used to deal with the problem of environmental light interference, then the improved yolov4 model is established to deal with noise and other interference information. Finally, the image feature extraction optimization is completed based on the attention mechanism, and then the road vehicle target detection is completed more accurately. The experimental results show that the proposed system has good robustness, can reduce the amount of calculation and improve the detection accuracy.