共查询到20条相似文献,搜索用时 328 毫秒
1.
2.
3.
4.
《交通信息与安全》2015,(6)
车载行人识别系统由于存在检测距离精确度不高及受遮挡影响较大等问题,在弯道及交叉口情况下适应性差。为提高行人防碰撞系统的预警效果,提出在车路协同环境下的行人目标信息融合算法研究。采用路侧和车载摄像头检测行人轨迹信息,通过Kalman滤波进行信息预处理,其次分别通过时间对准、空间对准、轨迹关联和信息融合完成对行人目标的位置估计。最后,搭建实车实验平台,对提出的信息融合算法进行验证。实验结果显示,对于X方向的行人轨迹误差,通过轨迹融合后,行人轨迹最大绝对误差、绝对平均误差相比于融合前均有大幅度减小,分别为50.00%,55.56%;对于Y方向的行人轨迹误差,通过轨迹融合后,行人轨迹最大绝对误差、绝对平均误差相比于融合前均有大幅度减小,分别为40.00%,62.07%。实验结果表明,该融合算法提高了行人轨迹检测精度,增强了系统的预警精确度。 相似文献
5.
针对智能汽车道路目标检测任务中单一传感器感知能力有限、多传感器后融合处理复杂等问题,提出了一种基于Transformer交叉注意力机制的多模态感知融合方法。首先,利用交叉注意力机制能较好地融合多模态信息的优势,搭建了基于深度学习方式的端到端融合感知网络,用以接收视觉与点云检测网络的输出,并进行后融合处理。其次,对点云检测网络的三维目标信息进行高召回处理,与视觉图像检测器输出的道路目标信息一同作为网络的输入。最后,通过网络实现二维目标信息向三维信息的融合,输出对三维目标检测信息的修正,从而得到准确度更高的后融合检测信息。在KITTI公开数据集上的验证指标表明,通过所提融合方法引入二维检测信息后,相比较PointPillars、PointRCNN、PV-RCNN及CenterPoint四种基准方法,对车辆、骑行人、行人3种类别的综合平均提升分别为7.07%、2.82%、2.46%、1.60%。通过与基于规则的后融合方法对比,所提融合网络在行人和骑行人中等、困难样本检测上,分别有平均1.88%与4.90%的提升。进一步表明所提方法具有更强的适应性与泛化能力。最后,进行了实车试验平台的搭建及算... 相似文献
6.
基于改进YOLOv2模型的驾驶辅助系统实时行人检测 总被引:1,自引:0,他引:1
《汽车工程》2019,(12)
为解决驾驶辅助系统(ADAS)对复杂背景行人和小尺寸行人检测精度较低的问题,基于深度神经网络模型YOLOv2建立了ADAS实时行人检测模型YOLOv2-P。首先在特征提取网络中采用参数化修正线性单元激活函数,以从训练数据中自适应地学习参数,并在行人检测网络中采用多层特征图融合方法,将低层特征图信息与高层特征图信息进行融合;然后使用交叉熵损失函数替代YOLOv2模型中的sigmoid激活函数,并对宽度、高度损失函数进行归一化处理;最后采用迭代自组织数据分析算法对行人数据集中行人边界框尺寸进行聚类。试验结果表明:相比于YOLOv2,YOLOv2-P对复杂背景行人及小尺寸行人的检测精度有明显提升,能够满足ADAS行人检测准确性和实时性需要。 相似文献
7.
8.
9.
10.
融合毫米波雷达与深度视觉的多目标检测与跟踪 总被引:1,自引:0,他引:1
《汽车工程》2021,(7)
针对现有融合毫米波雷达与传统机器视觉的车辆检测算法准确率较低与实时性较差的问题,本文中对多目标检测与跟踪进行研究。首先,利用阈值筛选和前后帧数据关联方法对毫米波雷达数据进行预处理,进而提出一种用于毫米波雷达数据跟踪的自适应扩展卡尔曼滤波算法。然后,为提高目标检测精度与速度,基于采集到的实车数据集训练卷积神经网络,完成深度视觉的多车辆检测。最后,采用决策级融合策略融合毫米波雷达与深度视觉信息,设计了一种用于复杂交通环境下前方车辆多目标检测与跟踪的框架。为验证所设计的框架,进行了不同交通环境下的实车实验。结果表明:该方法可实时检测跟踪前方车辆,具有比融合毫米波雷达与传统机器视觉的车辆检测方法更好的可靠性与鲁棒性。 相似文献
11.
交通信息感知作为交通信息基础设施最为关键的功能之一,可为交通态势预判、信号控制等交通应用场景提供重要的数据与决策支撑,是践行“交通强国”战略的基石。目前,常用交通信息感知手段主要包括地磁线圈、雷达、视频、红外等,但这些单一的交通信息感知设备普遍存在信息感知不全面、精度不高的问题。本文研究基于雷达与视频融合的交通信息感知技术,将毫米波雷达和视觉传感器两者的检测数据有机融合,从而实现大范围、精准而全面的交通信息感知,以期为交通的智能化、信息化发展奠定基础。 相似文献
12.
智能车辆安全辅助驾驶技术研究近况 总被引:3,自引:2,他引:3
论述了安全辅助驾驶技术的研究现状、研究的必要性以及研究进展。安全辅助驾驶技术包括车道偏离预警与保持、前方车辆探测及安全车距保持、行人检测、驾驶员行为监测、车辆运动控制与通讯等。分析了各种传感器的优缺点及其在实际应用过程中存在的问题,基于单一传感器不能很好地解决安全辅助驾驶技术可靠性和环境适应能力的要求,应结合激光雷达技术解决图像模糊问题,利用红外传感器增强机器视觉识别的可靠性,未来的安全辅助驾驶技术应该采取多种传感器融合的技术,结合毫米波雷达和激光雷达系统具有深度测量精确的特点,将极大的推动汽车安全辅助驾驶系统的应用和推广。 相似文献
13.
轨迹数据驱动的行人行为分析建模在公共场合异常事件监测、人车冲突风险评估等方面具有重要意义,广布的交通视频监控是行人群轨迹数据的重要来源。行人轨迹具有趋势性和规律性,提取的原始轨迹信息冗余较大,且密集行人群频繁遮挡,不同行人轨迹易发生误匹配,导致数据失真。针对以上问题,根据行人轨迹的局部结构特征和数值特性,设计一种改进的两阶段自适应滑窗轨迹压缩算法ATSSW (Adaptive Two Stage Sliding Window)和基于轨迹局部转向角的误匹配识别和分割方法ABTDS (Angle-based Trajectory Detection and Segmentation),清洗和压缩行人轨迹数据。首先,ATSSW算法考虑轨迹各坐标分量的数值分布特征,将提取到的所有原始轨迹分为漂移和非漂移2类,采取不同的策略分别压缩2类轨迹;然后,ABTDS算法分析压缩后的轨迹局部转角特征,辨识误匹配轨迹样本;最后,ABTDS算法分割误匹配样本,并用分割后的轨迹更新原始轨迹数据集。研究结果表明:ATSSW算法压缩了653条原始行人轨迹,总压缩信息损失1 002.04,总平均轨迹压缩率为6.07%,总平均轨迹压缩保留率为95.35%;原始轨迹集中存在126条误匹配轨迹,ABTDS算法辨识并成功分割了其中的107条,检出率为84.92%;所提算法抑制了原始行人轨迹中漂移点和误匹配现象所致的干扰,减少了原始轨迹数据噪声,可提高轨迹数据驱动的行人行为建模精确度;适当压缩原始轨迹,可减轻轨迹数据存储处理的负担。 相似文献
14.
S. H. Jeong J. E. Lee S. U. Choi J. N. Oh K. H. Lee 《International Journal of Automotive Technology》2012,13(7):1133-1140
Recently, the advanced driver assistance system (ADAS), which helps mitigate car accidents, has been developed using environmental detection sensors, such as long and short range radar, lidar, wide dynamic range cameras, ultrasonic sensors and laser scanners. Among these detection sensors, radars can quickly provide drivers with reliable information about the velocity, distance and direction of a target obstacle, as well as information about the vehicle in changing weather conditions. In the adaptive cruise control system (ACCS), three radar sensors are usually needed because two short range radars are used to detect objects in the adjacent lane and one long range radar is used to detect objects in-path. In this paper, low-cost radar based on a single sensor, which can detect objects in both the adjacent lane and in-path, is proposed for use in the ACCS. Before designing the proposed radar, we analyzed the world-wide radar technology and market trends for ACCS. Based on this analysis, we designed a novel radar sensor for the ACCS using radar components, such as an antenna, transceiver module, transceiver control module and signal processing algorithm. Finally, target detection experiments were conducted. In the experimental results, the proposed single radar can successfully complete the detection required for the ACCS. In the conclusion, the perspective and issues in the future development of the ACCS radar are described. 相似文献
15.
16.
介绍了上海通用汽车某车型在前期开发工作中长测距雷达布置设计的一系列要求以及相关设计方法和设计经验。主要包括雷达空间布置要求,行人保护的分析,雷达支架的动态分析,以及与之相关的造型外观设计优化需求,为长测距雷达在其他轿车上的布置提供参考。 相似文献
17.
18.
S. H. Jeong C. G. Choi J. N. Oh P. J. Yoon B. S. Kim M. Kim K. H. Lee 《International Journal of Automotive Technology》2010,11(3):409-416
This paper presents a low cost design and implementation of a parallel parking assist system (PPAS) based on ultrasonic sensors.
Generally, a PPAS requires several types of sensors, such as an ultrasonic sensor, camera sensor, radar sensor and laser sensor
for parking space detection. However, our proposed PPAS only requires two ultrasonic sensors on the front and lateral sides
for parking space detection. Moreover, a steering angle sensor and wheel speed sensor installed in the vehicle are used to
obtain vehicle position information for localization in ultrasonic range data. The hardware architecture of the PPAS based
on an electronic control unit (ECU) module, sensor modules and a human machine interface (HMI) module was proposed. Moreover,
the software architecture of the PPAS is based on system initialization, scheduling, recognition and a control algorithm.
In particular, a novel sensor algorithm was proposed to minimize the vehicle corner error of the ultrasonic sensor. A prototype
of the PPAS based on the proposed architecture was constructed. The experimental results demonstrate that the implemented
prototype is robust and successfully performs parking space detection and automatic steering control. Finally, the low cost
design and implementation of the PPAS was possible due to the cheap ultrasonic sensors, simple hardware design and low computational
complexity of the proposed algorithm. 相似文献
19.
J. Han O. Heo M. Park S. Kee M. Sunwoo 《International Journal of Automotive Technology》2016,17(3):483-491
For robust vision-based forward collision warning (FCW) and autonomous emergency braking (AEB) systems, not only reliable detection performance including high detection rate and low false positives but also accurate measurement output of a target vehicle is required. Especially, in order to reduce false alarm or activation of FCW/AEB systems, the systems require the precise measurement output of a target object, such as position, velocity, acceleration, and time-to-collision (TTC). In this study, we developed a measurement estimation algorithm of a target vehicle using a monocular camera. This method estimates two cases of vehicle widths for a target vehicle by using the detected lane information and a pin-hole camera model. After that, the position, velocity, acceleration, and TTC of a target vehicle are estimated by using a Kalman filter for the each estimated vehicle width. To improve robustness, the both estimation results using the detected lane information and the pinhole camera model are fused. This estimation algorithm was evaluated and compared with the state-of-the-art technology. As a result, the proposed measurement output estimation method can improve the performance of the FCW/AEB systems. 相似文献