FullStop: A Camera-Assisted System for Characterizing Unsafe Bus Stopping

Detecting Vehicles’ Relative Position on Two-Lane Highways Through a Smartphone-Based Video Overtaking Aid Application

Abstract: In this paper we present a smartphone-based real-time video overtaking architecture for vehicular networks. The developed application aims to prevent head-on collisions that might occur due to attempts to overtake when the view of the driver is obstructed by the presence of a larger vehicle ahead. Under such conditions, the driver does not have a clear view of the road ahead and of any vehicles that might be approaching from the opposite direction, resulting in a high probability of accident occurrence. Our application relies on the use of a dashboard-mounted smartphone with the back camera facing the windshield, and having the screen towards the driver. A video is streamed from the vehicle ahead to the vehicle behind automatically, where it is displayed so that the driver can decide if it is safe to overtake. One of the major challenges is the way to pick the right video source and destination among vehicles in close proximity, depending on their relative position on the road. For this purpose, we have focused on two different methods: one relying solely on GPS data, and the other involving the use of the camera and vehicle heading information. Our experiments show that the faster method, using just the location information, is prone to errors due to GPS inaccuracies. A second method that depends on data fusion from the optical sensor and GPS, although accurate over short distances, becomes more computationally intensive, and its performance significantly depends on the quality of the camera.

Ethical AI for Automated Bus Lane Enforcement

Abstract: There is an explosion of camera surveillance in our cities today. As a result, the risks of privacy infringement and erosion are growing, as is the need for ethical solutions to minimise the risks. This research aims to frame the challenges and ethics of using data surveillance technologies in a qualitative social context. A use case is presented which examines the ethical data required to automatically enforce bus lanes using camera surveillance and proposes ways of minimising the risks of privacy infringement and erosion in that scenario. What we seek to illustrate is that there is a challenge in using technologies in positive, socially responsible ways. To do that, we have to better understand the use case and not just the present, but also the downstream risks, and the downstream ethical questions. There is a gap in the literature in this aspect as well as a gap in the actual thinking of researchers in terms of understanding and responding to it. A literature review and detailed risk analysis of automated bus lane enforcement is conducted. Based on this, an ethical design framework is proposed and applied to the use case. Several potential solutions are created and described. The final chosen solution may also be broadly applicable to other use cases. We show how it is possible to provide an ethical AI solution for detecting infringements that incorporates privacy-by-design principles, while being fair to potential transgressors. By introducing positive, pragmatic and adaptable methods to support and uphold privacy, we support access to innovation that can help us mitigate current emerging risks.

Distinctive Image Features from Scale-Invariant Keypoints

Abstract: This paper presents a method for extracting distinctive invariant features from images that can be used to perform reliable matching between different views of an object or scene. The features are invariant to image scale and rotation, and are shown to provide robust matching across a substantial range of affine distortion, change in 3D viewpoint, addition of noise, and change in illumination. The features are highly distinctive, in the sense that a single feature can be correctly matched with high probability against a large database of features from many images. This paper also describes an approach to using these features for object recognition. The recognition proceeds by matching individual features to a database of features from known objects using a fast nearest-neighbor algorithm, followed by a Hough transform to identify clusters belonging to a single object, and finally performing verification through least-squares solution for consistent pose parameters. This approach to recognition can robustly identify objects among clutter and occlusion while achieving near real-time performance.

SIFT Flow: Dense Correspondence across Scenes and Its Applications

Abstract: While image alignment has been studied in different areas of computer vision for decades, aligning images depicting different scenes remains a challenging problem. Analogous to optical flow, where an image is aligned to its temporally adjacent frame, we propose SIFT flow, a method to align an image to its nearest neighbors in a large image corpus containing a variety of scenes. The SIFT flow algorithm consists of matching densely sampled, pixelwise SIFT features between two images while preserving spatial discontinuities. The SIFT features allow robust matching across different scene/object appearances, whereas the discontinuity-preserving spatial model allows matching of objects located at different parts of the scene. Experiments show that the proposed approach robustly aligns complex scene pairs containing significant spatial differences. Based on SIFT flow, we propose an alignment-based large database framework for image analysis and synthesis, where image information is transferred from the nearest neighbors to a query image according to the dense scene correspondence. This framework is demonstrated through concrete applications such as motion field prediction from a single image, motion synthesis via object transfer, satellite image registration, and face recognition.

Activity recognition from accelerometer data

Abstract: Activity recognition fits within the bigger framework of context awareness. In this paper, we report on our efforts to recognize user activity from accelerometer data. Activity recognition is formulated as a classification problem. Performance of base-level classifiers and meta-level classifiers is compared. Plurality Voting is found to perform consistently well across different settings.

A tutorial survey on vehicular ad hoc networks

Abstract: There has been significant interest and progress in the field of vehicular ad hoc networks over the last several years. VANETs comprise vehicle-to-vehicle and vehicle-to-infrastructure communications based on wireless local area network technologies. The distinctive set of candidate applications (e.g., collision warning and local traffic information for drivers), resources (licensed spectrum, rechargeable power source), and the environment (e.g., vehicular traffic flow patterns, privacy concerns) make the VANET a unique area of wireless communication. This article gives an overview of the field, providing motivations, challenges, and a snapshot of proposed solutions.

Nericell: rich monitoring of road and traffic conditions using mobile smartphones

Abstract: We consider the problem of monitoring road and traffic conditions in a city. Prior work in this area has required the deployment of dedicated sensors on vehicles and/or on the roadside, or the tracking of mobile phones by service providers. Furthermore, prior work has largely focused on the developed world, with its relatively simple traffic flow patterns. In fact, traffic flow in cities of the developing regions, which comprise much of the world, tends to be much more complex owing to varied road conditions (e.g., potholed roads), chaotic traffic (e.g., a lot of braking and honking), and a heterogeneous mix of vehicles (2-wheelers, 3-wheelers, cars, buses, etc.).To monitor road and traffic conditions in such a setting, we present Nericell, a system that performs rich sensing by piggybacking on smartphones that users carry with them in normal course. In this paper, we focus specifically on the sensing component, which uses the accelerometer, microphone, GSM radio, and/or GPS sensors in these phones to detect potholes, bumps, braking, and honking. Nericell addresses several challenges including virtually reorienting the accelerometer on a phone that is at an arbitrary orientation, and performing honk detection and localization in an energy efficient manner. We also touch upon the idea of triggered sensing, where dissimilar sensors are used in tandem to conserve energy. We evaluate the effectiveness of the sensing functions in Nericell based on experiments conducted on the roads of Bangalore, with promising results.