| 融合深度学习的无人驾驶多目标检测算法 |
| |
| 引用本文： | 朱玉刚.融合深度学习的无人驾驶多目标检测算法[J].教育技术导刊,2019,18(9):42-46. |
| |
| 作者姓名： | 朱玉刚 |
| |
| 作者单位： | 杭州电子科技大学 电子信息学院，浙江 杭州 310018 |
| |
| 摘 要： | 目标检测是目前计算机视觉领域非常热门的研究方向，是无人驾驶技术重要环节。为提高行车过程中目标检测精度并改善基于单发多目标检测器在训练时出现的正负样本失衡问题，基于车载视频，运用深度学习算法中具有强大性能的SSD模型，通过引入Focal Loss函数设计新的损失函数，解决样本失衡问题；同时在不降低检测速率的情况下，提高检测精度。基于自行采集的车载视频数据集进行实验，结果表明，改进后SSD模型的mAP相较于原始SSD模型提高了3%，达到74%。
|
| 关 键 词： | 深度学习 无人驾驶 SSD 目标检测 Focal Loss |
| 收稿时间： | 2019-02-11 |
Multi-Object Detection Algorithm for Unmanned Driving Based on Deep Learning |
| |
| ZHU Yu-gang.Multi-Object Detection Algorithm for Unmanned Driving Based on Deep Learning[J].Introduction of Educational Technology,2019,18(9):42-46. |
| |
| Authors: | ZHU Yu-gang |
| |
| Institution: | School of Electronics and Communication，Hangzhou Dianzi University， Hangzhou 310018， China |
| |
| Abstract: | Object detection is a very popular direction in the field of computer vision， especially in the field of unmanned driving technology. In order to improve the detection accuracy of objects encountered in the driving process and improve the imbalance of positive and negative samples based on single Shot MultiBox Detector （SSD） in training， based on vehicle video， this paper uses SSD model which has strong performance in deep learning algorithm， and designs a new loss function by introducing Focal Loss function to solve the problem of sample imbalance. In this way， the detection accuracy is improved without reducing the detection rate. This method is more effective and convenient. Based on the self-collected vehicle video data set， the experimental results show that the mAP （mean Average Precision） of the improved SSD model is 3% higher than that of the original SSD model， up to 74%. |
| |
| Keywords: | deep learning unmanned vehicle SSD object detection Focal Loss |
|
| 点击此处可从《教育技术导刊》浏览原始摘要信息 |
| 点击此处可从《教育技术导刊》下载免费的PDF全文 |
