Efficient L-shape fitting for vehicle detection using laser scanners

Zhang, Xiao; Xu, Wenda; Dong, Chiyu; Dolan, John M.

doi:10.1109/ivs.2017.7995698

Cited by 102 publications

(63 citation statements)

References 14 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The proposed BoxNet method was thoroughly compared with the state-of-the-art 2D bounding box fitting algorithm Search-based L-shape Fitting (SLF) [16], in which area, closeness and variance were used as the objectives to optimize the box that can tightly embrace all the points -a non-learning generic strategy focusing on the observed points without considering the correlation prior between points and object geometry. In order to have a fair comparison, we implemented and ran the SLF method only on the testing samples.…”

Section: Statistical Resultsmentioning

confidence: 99%

See 1 more Smart Citation

BoxNet: A Deep Learning Method for 2D Bounding Box Estimation from Bird's-Eye View Point Cloud

Nezhadarya

Liu²,

Liu³

2019

2019 IEEE Intelligent Vehicles Symposium (IV)

View full text Add to dashboard Cite

We present a learning-based method to estimate the object bounding box from its 2D bird's-eye view (BEV) LiDAR points. Our method, entitled BoxNet, exploits a simple deep neural network that can efficiently handle unordered points. The method takes as input the 2D coordinates of all the points and the output is a vector consisting of both the box pose (position and orientation in LiDAR coordinate system) and its size (width and length). In order to deal with the angle discontinuity problem, we propose to estimate the double-angle sinusoidal values rather than the angle itself. We also predict the center relative to the point cloud mean to boost the performance of estimating the location of the box. The proposed method does not rely on the ordering of points as in many existing approaches, and can accurately predict the actual size of the bounding box based on the prior information that is obtained from the training data. BoxNet is validated using the KITTI 3D object dataset, with significant improvement compared with the state-of-the-art non-learning based methods.

show abstract

Section: Statistical Resultsmentioning

confidence: 99%

“…No estimation of width or length was mentioned in the method. Recently, three evaluation criteria were proposed in [16] within an optimization framework to generate a bounding box that can embrace all the points. In [17], the points were iteratively clustered to two orthogonal segments using their ordered information.…”

Section: Related Workmentioning

confidence: 99%

BoxNet: A Deep Learning Method for 2D Bounding Box Estimation from Bird's-Eye View Point Cloud

Nezhadarya

Liu²,

Liu³

2019

2019 IEEE Intelligent Vehicles Symposium (IV)

View full text Add to dashboard Cite

show abstract

“…In our localization approach based on square landmarks, several presented methods are applied. In order to represent a square landmark with four corners, we refer to the L-Shape Fitting method proposed in [10]. The following position estimate stage employs Particle Filter (also known as Monte Carlo Localization) [11].…”

Section: Related Workmentioning

confidence: 99%

“…Although it can only perceive a single layer of 3-D information, it is fully applicable for the detection of square-like landmarks, the needed feature of which lies in the corners of its cross section. To complete this task, an L-Shape Fitting approach [10] is utilized, which will deduce the coordinates of the four corner points for each detected square item. Furthermore, these produced corner coordinates will be filtered and used in the measurement update of Particle Filter, associating with an off-line map containing the ground truth coordinates of the landmarks.…”

Section: Overviewmentioning

confidence: 99%

“…After acquiring the scanning data of the environmental objects, segmentation is firstly employed, separating data points into different clusters. Afterwards, we follow the L-shape fitting approach proposed in [10], the consequence of which represents all eligible point cloud shape with a rectangle. To apply this result for localization, we calculate the corners of all deduced rectangles.…”

Section: Square-like Landmarks Extractionmentioning

confidence: 99%

See 1 more Smart Citation

Self-Localization of Parking Robots Using Square-Like Landmarks

Chen

et al. 2018

2018 IEEE International Conference on Robotics and Biomimetics (ROBIO)

View full text Add to dashboard Cite

In this paper, we present a framework for selflocalization of parking robots in a parking lot innovatively using square-like landmarks, aiming to provide a positioning solution with low cost but high accuracy. It utilizes square structures common in parking lots such as pillars, corners or charging piles as robust landmarks and deduces the global pose of the robot in conjunction with an off-line map. The localization is performed in real-time via Particle Filter using a single line scanning LiDAR as main sensor, an odometry as secondary information sources. The system has been tested in a simulation environment built in V-REP, the result of which demonstrates its positioning accuracy below 0.20 m and a corresponding heading error below 1 • .

show abstract

Particle Filter for Trajectories of Movers from Laser Scanned Dataset

Sharif

2019

Pattern Recognition and Artificial Intelligence

View full text Add to dashboard Cite

Efficient L-shape fitting for vehicle detection using laser scanners

Cited by 102 publications

References 14 publications

BoxNet: A Deep Learning Method for 2D Bounding Box Estimation from Bird's-Eye View Point Cloud

BoxNet: A Deep Learning Method for 2D Bounding Box Estimation from Bird's-Eye View Point Cloud

Self-Localization of Parking Robots Using Square-Like Landmarks

Particle Filter for Trajectories of Movers from Laser Scanned Dataset

Contact Info

Product

Resources

About