vins-fusion单目IMU跑港大MARS数据集注意事项

vins-fusion跑港大MARS数据集只能跑单目IMU的。

港大MARS数据集链接如下：

HKU MaRS Labhttps://mars.hku.hk/dataset.html

港大MARS数据集里的单目图片消息类型是sensor_msgs/CompressedImage，而vins订阅的图像消息类型是sensor_msgs/Image，所以需要把sensor_msgs/CompressedImage消息类型转换成sensor_msgs/Image消息类型。

可以用一个python脚本实现（代码参考自：sensor_msgs/CompressedImage消息类型转换成sensor_msgs/Image消息类型_compressedimage转为image-CSDN博客）

import rospy
from sensor_msgs.msg import CompressedImage, Image
from cv_bridge import CvBridge, CvBridgeError

bridge = CvBridge()

def compressed_image_callback(msg):
    try:
        cv_image = bridge.compressed_imgmsg_to_cv2(msg)
    except CvBridgeError as e:
        rospy.logerr("CvBridgeError: %s", e)
        return

    try:
        image_msg = bridge.cv2_to_imgmsg(cv_image, encoding="bgr8")
    except CvBridgeError as e:
        rospy.logerr("CvBridgeError: %s", e)
        return

    image_msg.header = msg.header

    pub.publish(image_msg)

if __name__ == '__main__':
    rospy.init_node("compressed_to_image")
    sub = rospy.Subscriber("compressed_image", CompressedImage, compressed_image_callback, queue_size=1)
    pub = rospy.Publisher("image", Image, queue_size=1)
    rospy.spin()

FAST-LIVO自己跑bag包时，也是借助image_transport把压缩图像话题转为原始图像话题跑的
https://gitee.com/huanglilong/FAST-LIVO/blob/main/launch/mapping_avia.launch

 

vins订阅的MARS数据集里的imu话题注意订阅大疆飞控的imu话题/dji_osdk_ros/imu，不要订阅livox激光雷达的imu话题/livox/imu，因为/livox/imu里的加速度值是普通加速度值的十分之一，会导致vins一直没法初始化过去，终端一直报misalign visual structure with IMU。