Small target detection algorithm based on improved PointPillars

Journal Information

J. Hefei Univ. Tech. (Nat. Sci.), Jul. 2026, Vol. 49 No. 7: 865-871

DOI: 10.3969/j.issn.1003-5060.2026.07.001

Authors

ZHANG Bingli $ ^{1,2} $, YANG Chenglei $ ^{1,2} $, PAN Zehao $ ^{1,2} $,

(1. School of Automobile and Traffic Engineering, Hefei University of Technology, Hefei 230009, China; 2. Anhui Provincial Research Center for Intelligent Automotive Engineering, Hefei 230009, China)

Abstract and Keywords

Abstract: In view of the missing and false detection of small and medium targets in the traditional point cloud detection algorithm, this paper proposes a small target detection algorithm based on improved PointPillars. Firstly, feature pyramid module is added into the backbone network to optimize the detection of small targets by combining high-level semantic information and low-level positioning information. Secondly, by adding the residual module and the convolutional block attention module (CBAM), the channel information and spatial information are combined to complete the extraction of deep features. Finally, three detection heads output feature maps of different sizes to achieve accurate recognition of different types of targets. The feasibility analysis of the proposed algorithm is carried out on KITTI data set. Compared with traditional PointPillars algorithm, the proposed algorithm improves the mean average precision (mAP) for small target recognition by 1.74%. The results of real vehicle test show that the algorithm is accurate for road target detection, and the detection frequency of 65 Hz meets the real-time requirement.

Keywords: feature pyramid; residual module; attention mechanism; road target; intelligent driving

作者信息

张炳力 $ ^{1,2} $，杨程磊 $ ^{1,2} $，潘泽昊 $ ^{1,2} $，王怪昕 $ ^{1,2} $，王欣雨 $ ^{1,2} $，王焱辉 $ ^{1,2} $

(1. 合肥工业大学汽车与交通工程学院，安徽合肥 230009；2. 安徽省智能汽车工程研究中心，安徽合肥 230009)

摘要和关键词

摘要: 针对传统点云检测算法中小目标漏检、误检等问题, 文章提出一种改进的 PointPillars 小目标检测算法。首先, 在主干网络中引入特征金字塔模块, 融合高层次语义信息与低层次定位信息来优化对小目标的检测; 其次, 加入残差模块和卷积块注意力模块 (convolutional block attention module, CBAM), 结合通道信息与空间信息来完成对深层特征的提取; 最后, 通过三检测头输出不同尺寸的特征图, 以实现对不同类别目标的精确识别。在公开数据集 KITTI 上进行的可行性分析结果表明, 相较于传统 PointPillars 算法, 该文算法对小目标识别的平均精度均值提升了 1.74%。实车试验结果表明, 该文算法对道路目标检测结果准确, 检测频率达 65 Hz, 满足实时性需求。

关键词: 特征金字塔；残差模块；注意力机制；道路目标；智能驾驶

Funding

安徽省科技重大专项资助项目（202203a05020008）；长三角科技创新共同体联合攻关专项资助项目（2022CSJGG1501）和济宁市产业创新重大技术“全球揭榜”资助项目（2022JBZP002）

合肥工业大学学报(自然科学版)

Navigation