Recent years, people pay great attention to the fusion of different kinds of sensors include camera, radar, LiDAR and so on. Since each sensor has its own advantage and disadvantage in automatic drive and obstacle detection, the fusion of sensors is more robust and reliable. In this paper, we present a novel target calibration of LiDAR-camera system. We use a 3D Flash LiDAR which has a resolution of 320×240, much cheaper and more reliable than Scanning LiDAR, and we use the camera which has a resolution of 1280×1024. We propose a novel target calibration method. The 3D target can provide both geometric features and visual features. This method is fast, easy to estimate all the six parameters of the extrinsic calibration. Our experiments validate our method and show that it achieves good accuracy.
Pose estimation by monocular is finding the pose of the object by a single image of feature points on the object, which must meet the requirements of detecting all the feature points and matching them in the image. But it will be difficult to obtain the correct pose if part of the feature points are occluded when the object moving a large scale. We proposed a method for finding the pose on the condition that the correspondences between the object points and the image points are unknown. The method combines two algorithms: one algorithm is SoftAssign, which constructs a weight matrix of feature points and image points, and determines the correspondences by iteration loop processing; the other algorithm is OI(orthogonal iteration), which derives an iterative algorithm which directly computes orthogonal and globally convergent rotation matrices.We nest the two algorithms into one iteration loop.An appropriate pose will be chosen from a set of reference poses as the initial pose of object at the beginning of the loop, then we process the weight matrix to confirm the correspondences and calculate the optimal solution of rotation matrices alternately until the object space collinearity error is less than the threshold, each estimation will be closer to the truth pose than the last one through every iteration loop. Experimentally, the method proved to be efficient and have a high precision pose estimation of 3D object with large-scale motion.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.