基于激光雷达和摄像机融合的智能车障碍物识别方法研究

针对智能车环境感知中单一传感器所存在的局限性问题,本文提出一种通过激光雷达融合摄像机来感应识别智能车前方障碍物的方法。首先,通过激光雷达与摄像机之间的校准,来实现目标的三维数据的图像投影,并进行视觉图像与目标的三维雷达数据的融合,以提取障碍物候选区域。其次,提出了一种基于卷积神经网络和SVM的障碍物识别模型,用于训练KITTI数据库中的数据,检测视觉图像中的行人和车辆目标,以此来得到所需要的单帧下各传感器的目标检测数据。实验结果表明,所提出的模型在KITTI中选择的小数据集上获得的模型在实际测试中具有良好的性能,具有可靠的识别能力和良好的分类结果。

Research on Intelligent Vehicle Obstacle Recognition Based on Lidar and Camera Fusion

Aiming at the limitation of the single sensor in the smart car environment perception, it proposed a method to identify the obstacle in front of the smart car through the laser radar fusion camera. Firstly, the image projection of the target three-dimensional data was realized by the calibration between the laser radar and the camera, and the fusion of the visual image and the target three-dimensional radar data was performed to extract the obstacle candidate region. Secondly, an obstacle recognition model based on convolutional neural network and SVM was proposed to train the data in the KITTI database to detect pedestrians and vehicle targets in the visual image, so as to obtain the required single-frame sensors. Target detection data. The experimental results show that the proposed model has good performance in the actual test, and has reliable recognition ability and good classification results.