Search

With the rapid development of intelligent driving vehicles, multi-task visual perception based on deep learning emerges as a key technological pathway toward safe vehicle navigation in real traffic scenarios. However, due to the high-precision and high-efficiency requirements of intelligent driving vehicles in practical driving environments, multi-task visual perception remains a challenging task. Existing methods typically adopt effective multi-task learning networks to concurrently handle multiple tasks. Despite the fact that they obtain remarkable achievements, better performance can be achieved through tackling existing problems like underutilized high-resolution features and underexploited non-local contextual dependencies. In this work, we propose YOLOPv3, an efficient anchor-based multi-task visual perception network capable of handling traffic object detection, drivable area segmentation, and lane detection simultaneously. Compared to prior works, we make essential improvements. On the one hand, we propose architecture enhancements that can utilize multi-scale high-resolution features and non-local contextual dependencies for improving network performance. On the other hand, we propose optimization improvements aiming at enhancing network training, enabling our YOLOPv3 to achieve optimal performance via straightforward end-to-end training. The experimental results on the BDD100K dataset demonstrate that YOLOPv3 sets a new state of the art (SOTA): 96.9% recall and 84.3% mAP50 in traffic object detection, 93.2% mIoU in drivable area segmentation, and 88.3% accuracy and 28.0% IoU in lane detection. In addition, YOLOPv3 maintains competitive inference speed against the lightweight YOLOP. Thus, YOLOPv3 stands as a robust solution for handling multi-task visual perception problems. The code and trained models have been released on GitHub. [ABSTRACT FROM AUTHOR]

Car ownership in China reached 194 million vehicles at the end of 2016. The traffic congestion index (TCI) exceeds 2.0 during rush hour in some cities. Inefficient processing for minor traffic accidents is considered to be one of the leading causes for road traffic jams. Meanwhile, the process after an accident is quite troublesome. The main reason is that it is almost always impossible to get the complete chain of evidence when the accident happens. Accordingly, a police and insurance joint management system is developed which is based on high precision BeiDou Navigation Satellite System (BDS)/Global Positioning System (GPS) positioning to process traffic accidents. First of all, an intelligent vehicle rearview mirror terminal is developed. The terminal applies a commonly used consumer electronic device with single frequency navigation. Based on the high precision BDS/GPS positioning algorithm, its accuracy can reach sub-meter level in the urban areas. More specifically, a kernel driver is built to realize the high precision positioning algorithm in an Android HAL layer. Thus the third-party application developers can call the general location Application Programming Interface (API) of the original standard Global Navigation Satellite System (GNSS) to get high precision positioning results. Therefore, the terminal can provide lane level positioning service for car users. Next, a remote traffic accident processing platform is built to provide big data analysis and management. According to the big data analysis of information collected by BDS high precision intelligent sense service, vehicle behaviors can be obtained. The platform can also automatically match and screen the data that uploads after an accident to achieve accurate reproduction of the scene. Thus, it helps traffic police and insurance personnel to complete remote responsibility identification and survey for the accident. Thirdly, a rapid processing flow is established in this article to meet the requirements to quickly handle traffic accidents. The traffic police can remotely identify accident responsibility and the insurance personnel can remotely survey an accident. Moreover, the police and insurance joint management system has been carried out in Wuhan, Central China's Hubei Province, and Wuxi, Eastern China's Jiangsu Province. In a word, a system is developed to obtain and analyze multisource data including precise positioning and visual information, and a solution is proposed for efficient processing of traffic accidents. [ABSTRACT FROM AUTHOR]

Rényi entropy as a generalization of the Shannon entropy allows for different averaging of probabilities of a control parameter α. This paper gives a new perspective of the Kalman filter from the Rényi entropy. Firstly, the Rényi entropy is employed to measure the uncertainty of the multivariate Gaussian probability density function. Then, we calculate the temporal derivative of the Rényi entropy of the Kalman filter's mean square error matrix, which will be minimized to obtain the Kalman filter's gain. Moreover, the continuous Kalman filter approaches a steady state when the temporal derivative of the Rényi entropy is equal to zero, which means that the Rényi entropy will keep stable. As the temporal derivative of the Rényi entropy is independent of parameter α and is the same as the temporal derivative of the Shannon entropy, the result is the same as for Shannon entropy. Finally, an example of an experiment of falling body tracking by radar using an unscented Kalman filter (UKF) in noisy conditions and a loosely coupled navigation experiment are performed to demonstrate the effectiveness of the conclusion. [ABSTRACT FROM AUTHOR]

The existing deep learning methods for point cloud classification are trained using abundant labeled samples and used to test only a few samples. However, classification tasks are diverse, and not all tasks have enough labeled samples for training. In this paper, a novel point cloud classification method for indoor components using few labeled samples is proposed to solve the problem of the requirement for abundant labeled samples for training with deep learning classification methods. This method is composed of four parts: mixing samples, feature extraction, dimensionality reduction, and semantic classification. First, the few labeled point clouds are mixed with unlabeled point clouds. Next, the mixed high-dimensional features are extracted using a deep learning framework. Subsequently, a nonlinear manifold learning method is used to embed the mixed features into a low-dimensional space. Finally, the few labeled point clouds in each cluster are identified, and semantic labels are provided for unlabeled point clouds in the same cluster by a neighborhood search strategy. The validity and versatility of the proposed method were validated by different experiments and compared with three state-of-the-art deep learning methods. Our method uses fewer than 30 labeled point clouds to achieve an accuracy that is 1.89–19.67% greater than existing methods. More importantly, the experimental results suggest that this method is not only suitable for single-attribute indoor scenarios but also for comprehensive complex indoor scenarios. [ABSTRACT FROM AUTHOR]

The loose combination (LC) and the tight combination (TC) are two different models in the combined processing of four global navigation satellite systems (GNSSs). The former is easy to implement but may be unusable with few satellites, while the latter should cope with the inter-system bias (ISB) and is applicable for few tracked satellites. Furthermore, in both models, the inter-frequency bias (IFB) in the GLObal NAvigation Satellite System (GLONASS) system should also be removed. In this study, we aimed to investigate the performance difference of ambiguity resolution and position estimation between these two models simultaneously using the single-frequency data of all four systems (GPS + GLONASS + Galileo + BeiDou Navigation Satellite System (BDS)) in three different environments, i.e., in an open area, with surrounding high buildings, and under a block of high buildings. For this purpose, we first provide the definition of ISB and IFB from the perspective of the hardware delays, and then propose practical algorithms to estimate the IFB rate and ISB. Thereafter, a comprehensive performance comparison was made between the TC and LC models. Experiments were conducted to simulate the above three observation environments: the typical situation and situations suffering from signal obstruction with high elevation angles and limited azimuths, respectively. The results show that in a typical situation, the TC and LC models achieve a similar performance. However, when the satellite signals are severely obstructed and few satellites are tracked, the float solution and ambiguity fixing rates in the LC model are dramatically decreased, while in the TC model, there are only minor declines and the difference in the ambiguity fixing rates can be as large as 30%. The correctly fixed ambiguity rates in the TC model also had an improvement of around 10%. Once the ambiguity was fixed, both models achieved a similar positioning accuracy. [ABSTRACT FROM AUTHOR]

A non-linear filtering algorithm based on the alpha-divergence is proposed, which uses the exponential family distribution to approximate the actual state distribution and the alpha-divergence to measure the approximation degree between the two distributions; thus, it provides more choices for similarity measurement by adjusting the value of α during the updating process of the equation of state and the measurement equation in the non-linear dynamic systems. Firstly, an α -mixed probability density function that satisfies the normalization condition is defined, and the properties of the mean and variance are analyzed when the probability density functions p (x) and q (x) are one-dimensional normal distributions. Secondly, the sufficient condition of the alpha-divergence taking the minimum value is proven, that is when α ≥ 1 , the natural statistical vector's expectations of the exponential family distribution are equal to the natural statistical vector's expectations of the α -mixed probability state density function. Finally, the conclusion is applied to non-linear filtering, and the non-linear filtering algorithm based on alpha-divergence minimization is proposed, providing more non-linear processing strategies for non-linear filtering. Furthermore, the algorithm's validity is verified by the experimental results, and a better filtering effect is achieved for non-linear filtering by adjusting the value of α. [ABSTRACT FROM AUTHOR]