In this paper, we propose to fuse the LIDAR and monocular image in framework of conditional random field detect road robustly challenging scenarios. points are aligned with pixels by cross calibration. Then boosted decision tree based classifiers trained for point cloud respectively. The scores two kinds treated as unary potentials corresponding pixel nodes field. fused can be solved efficiently graph cut. Extensive experiments tested on KITTI-Road benchmark show that our method reaches...

Multi-object tracking is a crucial problem for autonomous vehicle. Most state-of-the-art approaches adopt the tracking-by-detection strategy, which two-step procedure consisting of detection module and module. In this paper, we improve both steps. We by incorporating temporal information, beneficial detecting small objects. For module, propose novel compressed deep Convolutional Neural Network (CNN) feature based Correlation Filter tracker. By carefully integrating these two modules,...

Road detection is a key task for autonomous land vehicles. Monocular vision-based road-detection algorithms are mostly based on machine learning approaches and usually cast as classification problems. However, the pixel-wise classifiers faced with ambiguity caused by changes in road appearance, illumination weather. An effective way to reduce model contextual information structured prediction. Currently, widely used prediction Markov random field or conditional field. field-based methods...

Negative obstacles for field autonomous land vehicles (ALVs) refer to ditches, pits, or terrain with a negative slope, which will bring risks in travel. This paper presents feature fusion based algorithm (FFA) obstacle detection LiDAR sensors. The main contributions of this are fourfold: (1) A novel three-dimensional (3-D) setup is presented. With setup, the blind area around vehicle greatly reduced, and density data improved, critical ALVs. (2) On basis proposed mathematical model point...

A unified Mie and fractal model for light scattering by biological cells is presented. This shown to provide an excellent global agreement with the angular dependent elastic spectroscopy of over whole visible range (400 700 nm) at all angles (1.1 165 deg) investigated. from bare cell nucleus found dominate in forward directions, whereas random fluctuation background refractive index within cell, behaving as a continuous medium, other angles. Angularly aided demonstrated be effective...

Many road detection algorithms require pre-learned information, which may be unreliable as the scene is usually unexpectable. Single image based (i.e., without any information) techniques can adopted to overcome this problem, while their robustness needs improving. To achieve robust from a single image, paper proposes general shape prior enforce detected region road-shaped by encoding into graph-cut segmentation framework, where training data automatically generated predicted of current...

Federated learning (FL) as an emerging distributed machine (ML) paradigm enables participants to train their on-device data locally and share model parameters with others by the parameter server. Differing from centralized ML, FL splits high requirements of training computing power server clients, which is well adapted unmanned aerial vehicle (UAV) swarms scattered nodes, heterogeneous data, limited power. However, pre-trained models are unsatisfactory in unfamiliar scenes most existing...

Global localization is a challenging problem in which autonomous vehicle has to estimate the self-position with respect priori map using perception results. In this paper we present vision-based method for Vehicles urban environment. The process consists of two stages: coarse topological and fine metric map. represented by holistic image feature provides location, whereas from relatively slow, but more accurate. It possible integrate stages make precise reliable localization. location system...

In this paper, we propose a multi-cue fusion approach to detect the road boundary using stereo vision, which fits with few edge points. Firstly areas are determined in accordance normal vector information. Based on cues of vector, height and color area, three Bayes models established respectively. Then Naive framework could provide confidence level each point fuses kinds cues. highest would be output. Finally, support regression (SVR) method curb curves according correct Extensive...

In this paper, we propose to fuse the geometric information of a 3-D LiDAR and monocular camera detect urban road region ahead an autonomous vehicle. Our method takes advantage both high definition data continuity in image representation. First, obtain efficient representation organized 2-D inverse depth map, by projecting points onto camera's plane. Through new representation, can acquire intermediate representations scenes extracting vertical horizontal histograms normalized depth. The...

Range images are commonly used representations for 3D LiDAR point cloud in the field of autonomous driving. The approach generating a range image is generally regarded as standard approach. However, there do exist two different types approaches to image: In one approach, row defined laser ID, and other elevation angle. We named first Projection By Laser ID (PBID), second Elevation Angle (PBEA). Few previous works have paid attention difference these approaches. this work, we quantitatively...

Pedestrian detection plays an essential role in the navigation system of autonomous vehicles. Multisensor fusion-based approaches are usually used to improve performance. In this study, we aimed develop a score pedestrian algorithm by integrating data two light and ranging systems (LiDARs). We first evaluated two-stage object-detection pipeline for each LiDAR, including object proposal fine classification. The scores from these different classifiers were then fused generate result using...

A novel method based upon multi-scale windows for efficient detection of the vehicles is described in this paper. This presumes that local image vehicle most symmetrical on center while least edge vehicle. However, value would not be only maximum actual images, which causes troubles locating vehicles. We present a new both symmetry axis and edges can detected simultaneously. also concocted strategy sequential processing order to improve detection. Experimental results shows our algorithm...

Automated lane detection is a vital part of driver assistance systems in intelligent vehicles. In this study, multi‐lane method based on omnidirectional images presented to conquer the difficulties stemming from limited view field rectilinear cameras. The contributions study are twofold. First, extract features markings under various illumination and road‐surface scenarios, feature extractor anisotropic steerable filter proposed. Second, parabola model used fit straight as well curved lanes....

Autonomous driving with artificial intelligence technology has been viewed as promising for autonomous vehicles hitting the road in near future. In recent years, considerable progress made Deep Reinforcement Learnings (DRLs) realizing end-to-end driving. Still, safely and comfortably real dynamic scenarios DRL is nontrivial due to reward functions being typically pre-defined expertise. This paper proposes a human-in-the-loop algorithm learning personalized behavior progressive way....

Federated learning (FL) has been proposed as a promising distributed paradigm to realize edge artificial intelligence (AI) without revealing the raw data.Nevertheless, it would incur inevitable costs in terms of training latency and energy consumption, due periodical communication between user equipments (UEs) geographically remote central parameter server.Thus motivated, we study joint aggregation association problem minimize total cost, where model over multiple cells just happens at...

This paper proposes a fast and robust algorithm for traffic sign detection recognition. The includes two stages: In the first stage, Adaboost based red pixels model of speed limit in Lab color space is built. Then used to extract area latent signs. After that, improved Hough Transform locate signs precisely. second template matching recognize sign. At same time, new method rejecting non-signs presented, which recognition rate complex outdoor scenes.

The significant economic contributions of the tourism industry in recent years impose an unprecedented force for data mining and machine learning methods to analyze data. intrinsic problems raw are largely related complexity, noise nonlinearity that may introduce many challenges existing techniques such as rough sets neural networks. In this paper, a novel method using SVM-based classification with two nonlinear feature projection is proposed analysis. first based on ISOMAP (Isometric...

Many classical visual odometry and simultaneous localization mapping methods are able to achieve excellent performance, but mainly restricted on the static scenes suffer degeneration when there many dynamic objects. In this paper, an efficient coarse-to-fine algorithm is proposed for moving object detection in autonomous driving. A motion-based conditional random field task modeled. Particularly, initial dynamic-static segmentation, a superpixel-based binary segmentation processed, further...

Pedestrian detection is a key technology in autonomous driving perception system. Although the current vision-based pedestrian has obtained very good performance, camera sensitive to light and shadow. In addition, it unable provide precise location information, which difficult address problem. To tackle these issues, LIDAR subsystem applied here order extract object structure features train an SVM classifier. Additionally, association of detections can be solved 3D world coordinates by...

Free space detection is crucial to autonomous vehicles while existing works are not entirely satisfactory. As cameras have many advantages on environment perception, a stereo vision-based robust free method proposed which mainly depends geometry information and Gaussian process regression. In this work, in order improve the performance by exploiting multiple source information, we apply Bayesian framework conditional random field inference fuse multimodal including 2-D image 3-D point...

Negative obstacle detection has been a challenging topic. In the previous researches, distance that negative obstacles can be detected is so near vehicles have to travel at very low speed. this paper, algorithm from image sequences proposed. When are far vehicle, color appearance models used as cues of detecting obstacles, while get closer, geometrical extracted stereo vision. Furthermore, different combined in Bayesian framework detect sequences. Massive experiments show proposed quite...