Efficient graph attentional network for 3D object detection from Frustum-based LiDAR point clouds

LiDAR-based 3D object detection is important for autonomous driving scene perception, but point clouds produced by LiDAR are irregular and unstructured in nature, and cannot be adopted by the conventional Convolutional Neural Networks (CNN). Recently, Graph Convolutional Networks (GCN) has been proved as an ideal way to handle non-Euclidean structure data, as well as for point cloud processing. However, GCN involves massive computation for searching adjacent nodes, and the heavy computational cost limits its applications in processing large-scale LiDAR point cloud in autonomous driving. In this work, we adopt a frustum-based point cloud-image fusion scheme to reduce the amount of LiDAR point clouds, thus making the GCN-based large-scale LiDAR point clouds feature learning feasible. On this basis, we propose an efficient graph attentional network to accomplish the goal of 3D object detection in autonomous driving, which can learn features from raw LiDAR point cloud directly without any conversions. We evaluate the model on the public KITTI benchmark dataset, the 3D detection mAP is 63.72% on KITTI Cars, Pedestrian and Cyclists, and the inference speed achieves 7.9 fps on a single GPU, which is faster than other methods of the same type.     

An effective deep learning model for grading abnormalities in retinal fundus images using variational auto-encoders

Diabetic retinopathy (DR) and Diabetic Macular Edema (DME) are severe diseases that affect the eyes due to damage in blood vessels. Computer-aided automated grading will help clinicians conduct disease diagnoses at ease. Experiments of automated image processing with deep learning techniques using CNN produce promising results, especially in the medical imaging domain. However, the disease grading tasks in retinal images using CNN struggle to retain high-quality information at the output. A novel deep learning model based on variational auto-encoder to grade DR and DME abnormalities in retinal images is proposed. The objective of the proposed model is to extract the most relevant retinal image features efficiently. It focuses on addressing less relevant candidate region generation and translational invariance present in images. The experiments are conducted in IDRID dataset and evaluated using accuracy, U-kappa, sensitivity, specificity and precision metrics. The results outperform compared with other state-of-art techniques.     

A job grouping approach for planning container transfers at automated seaport container terminals

This paper proposes a practical job grouping approach, which aims to enhance the time related performance metrics of container transfers in the Patrick AutoStrad container terminal, located in Brisbane, Australia. It first formulates a mathematical model of the automated container transfers in a relatively complex environment. Apart from the consideration on collision avoidance of a fleet of large vehicles in a confined area, it also deals with many other difficult practical challenges such as the presence of multiple levels of container stacking and sequencing, variable container orientations, and vehicular dynamics that require finite acceleration and deceleration times. The proposed job grouping approach aims to improve the makespan of the schedule for yard jobs, while reducing straddle carrier waiting time by grouping jobs using a guiding function. The performance of the current sequential job allocation method and the proposed job grouping approach are evaluated and compared statistically using a pooled t-test for 30 randomly generated yard configurations. The experimental results show that the job grouping approach can effectively improve the schedule makespan and reduce the total straddle carrier waiting time.     

State feedback sliding mode control without chattering by constructing Hurwitz matrix for AUV movement

This paper presents a new method to eliminate the chattering of state feedback sliding mode control (SMC) law for the mobile control of an autonomous underwater vehicle (AUV) which is nonlinear and suffers from unknown disturbances system. SMC is a well-known nonlinear system control algorithm for its anti-disturbances capability, while the chattering on switch surface is one stiff question. To dissipate the well-known chattering of SMC, the switching manifold is proposed by presetting a Hurwitz matrix which is deducted from the state feedback matrix. Meanwhile, the best switching surface is achieved by use of eigenvalues of the Hurwitz matrix. The state feedback control parameters are not only applied to control the states of AUV but also connected with coefficients of switching surface. The convergence of the proposed control law is verified by Lyapunov function and the robust character is validated by the Matlab platform of one AUV model.     

Robust anti-sliding control of autonomous vehicles in presence of lateral disturbances

Path following control problem of autonomous vehicles is investigated, concerning both unmeasurable sliding effects and lateral disturbances which lead to some difficulties in designing autonomous control under complex environment. To deal with the sliding effects, sideslip angles are modeled and reconstructed by estimating the tire cornering stiffness, which plays important role in analyzing the sliding effects. To this end, a Luenberger-type observer is designed, which is able to identify the tire cornering stiffness adaptively even in presence of time-varying lateral disturbances. Furthermore, to guarantee high-precision guidance, a sliding mode controller is designed based on chained system theory, and this controller is shown to be robust to both the lateral disturbances and the inaccuracy of the sliding reconstruction. Simulations illustrate that the proposed methods can reconstruct the sliding angles and provide high-accuracy anti-sliding control even in presence of the time-varying lateral disturbances.     

感兴趣区域的非线性变换多描述编码

介绍一种基于感兴趣区域的非线性变换的多描述图像编码,能够与目前的标准图像压缩算法兼容,对图像的感兴趣区域进行更好的保护。由感兴趣区域的非线性几何变换产生冗余,对图像进行采样形成多描述编码。通过仿真实验,证实当图像在差错信道下传输而丢失描述时,该算法能够使图像感兴趣区域得到更好的恢复。     

Adaptive multi-scale segmentation of surface data using unsupervised learning of seed positions

This paper presents a method for multi-scale segmentation of surface data using scale-adaptive region growing. The proposed segmentation algorithm is initiated by an unsupervised learning of optimal seed positions through the surface attribute clustering with a two-criterion score function. The seeds are selected as consecutive local maxima of the clustering map, which is computed by an aggregation of the local isotropic contrast and local variance maps. The proposed method avoids typical segmentation errors caused by an inappropriate choice of seed points and thresholds used in the region-growing algorithm. The scale-adaptive threshold estimate is based on the image local statistics in the neighborhoods of seed points. The performance of this method was evaluated on LiDAR surface images.     

一种NURBS曲面拟合精度的量化评测算法

曲面重构是逆向工程中的核心技术之一,由于NURBS曲面在光顺性和局部可编辑等方面所具有的优点,使其成为点云数据自由曲面重构的常见形式。目前对NURBS曲面重构技术的研究上取得了一些成果,但各方法在拟合精度和效果上各有参差,因此有必要对NURBS曲面拟合精度评价算法进行研究。在采用NURBS实现曲面拟合的基础上,对拟合精度的量化指标进行研究,设计了一种基于区域划分的搜索迭代算法,可快速地计算得到原始点云与NURBS曲面的偏差。     

Uniqueness of the Perfect Fusion Grid on ℤ d

Region merging methods consist of improving an initial segmentation by merging some pairs of neighboring regions. In a graph, merging two regions, separated by a set of vertices, is not straightforward. The perfect fusion graphs defined in J. Cousty et al. (J. Math. Imaging Vis. 30:(1):87–104, 2008) verify all the basic properties required by region merging algorithms as used in image segmentation. Unfortunately, the graphs which are the most frequently used in image analysis (namely, those induced by the direct and the indirect adjacency relations) are not perfect fusion graphs. The perfect fusion grid, introduced in the above mentioned reference, is an adjacency relation on ℤ ^d which can be used in image analysis, which indeed induces perfect fusion graphs and which is “between” the graphs induced by the direct and the indirect adjacencies. One of the main results of this paper is that the perfect fusion grid is the only such graph whatever the dimension d.

基于感兴趣区梯度方向直方图的行人检测

针对以梯度方向直方图作为人体特征的行人检测存在向量维数较大、检测时间较长的问题,提出基于感兴趣区梯度方向直方图的行人检测方法,分别在头部及四肢等重点区域计算梯度方向直方图,有效地减少了向量维数。实验结果表明,该方法在检测率基本不变的情况下提高了检测速度。