Matti Kutila

Driver Distraction Detection with a Camera Vision System

Driver assistance systems and electronics (e.g. navigators, cell phones, etc.) steal increasing amounts of driver attention. Therefore, the vehicle industry is striving to build a driving environment where input-output devices are smartly scheduled, allowing sufficient time for the driver to focus attention on the surrounding traffic. To enable a smart human-machine interface (HMI), the drivers momentary state needs to be measured. This paper describes a facility for monitoring the distraction of a driver and presents some early evaluation results. The module is able to detect the drivers visual and cognitive workload by fusing stereo vision and lane tracking data, running both rule-based and support-vector machine (SVM) classification methods. The module has been tested with data from a truck and a passenger car. The results show over 80% success in detecting visual distraction and a 68-86 % success in detecting cognitive distraction, which are satisfactory results.

Road condition monitoring system based on a stereo camera

This paper presents a method and evaluation results to monitor and detect road conditions (ice, water, snow and dry asphalt). The developed device bases on light polarization changes when reflected from road surface. The recognition capability has been improved with texture analysis which estimates contrast content of an image. Test drives were performed with a vehicle equipped also with a commercial solution from Vaisala which was used as the reference sensor. The results show that the proposed solution does not currently adapt to different conditions perfectly well. Therefore, further development targets have been identified to include not only more adjustable classifier but externally lighting to stabilizing ambient illumination.

Driver cognitive distraction detection : feature estimation and implementation

Abstract This article focuses on monitoring a drivers cognitive impairment due to talking to passengers or on a mobile phone, daydreaming, or just thinking about other than driving-related matters. This paper describes an investigation of cognitive distraction, firstly, giving an overall idea of its effects on the driver and, secondly, discussing the practical implementation of an algorithm for detection of cognitive distraction using a support vector machine (SVM) classifier. The evaluation data have been gathered by recruiting 12 professional drivers to drive for approximately 45 min in various environments and inducing cognitive tasks, i.e. arithmetic calculations. According to the prior knowledge and the experimental analysis, gaze, head and lane-keeping variances over a 15 s time window were selected indicative features. The SVM classifiers performance was optimized through exhaustive parameter tuning. The executed tests show that the cognitive workload can be detected with approximately 65-80 per cent confidence despite the fact that the test material represented medium-difficulty cognitive tasks (i.e. the induced workload was not very high). Thus, it could be assumed that a more challenging cognitive task would yield better detection results.

Utilization of Optical Road Surface Condition Detection around Intersections

This paper presents experimental results regarding road condition monitoring by machine vision techniques. The system will be further developed by VTT in the European INTERSAFE-2 project, focused on applications at road intersections. Knowing the presence of adverse conditions such as icy or wet roads, important effects can be obtained for intersection safety by means of more effective driver assistance functions. A very good reliability up to 93% has been found for the detection of icy roads, while lower values around 60% have been measured in the case of wet surfaces, due to a higher sensitivity to environmental conditions, especially outdoor light. Future steps will include an advanced classification algorithm and the implementation of active lighting. Integration of the vision equipment into a cooperative system for intersection safety is being investigated.

Slippery Road Detection by Using Different Methods of Polarised Light

Road friction measurement is an important issue for active safety systems on vehicles; hence knowledge of this key parameter can significantly improve the interventions on vehicle dynamics. This study compares two different on-board sensors for the classification of road conditions with polarised infrared light. Several tests are performed on a dedicated track, with focus on detection of dry or wet surfaces, and the presence of ice or snow. The work shows the capability of both sensors to provide a correct classification. In particular, results indicate how the monitored area, the presence of active illumination and the mounting position influence measurements and response times. It is concluded that both systems classify different road conditions in all cases. Performance of the Road eye system varied from 80 to 90% whereas the camera based IcOR achieved 70-80% accuracy level. Since these are being prototype sensors more development is needed before implemented into advanced safety applications.

Multi-camera-based smoke detection and traffic pollution analysis system

This article studies the smoke and exhaust detection system that has been developed for monitoring exhaust gases to enforce environmental laws and regulations. In many highly populated countries the HSU (Hart ridge Smoke Unit) grade is used to impose penalties. In many cases, HSU values over 40 ... 50 are leading to legal actions. This paper proposes a method that adopts two cameras, a far infrared camera and a high-resolution visible wavelength camera, as a detection system for smoky vehicle detection. The far infrared camera is used for detecting the location of the vehicle exhaust fumes. This thermal information is fused with visible spectrum information from the high-resolution camera. An algorithm evaluates if the identified vehicles are causing visible exhaust smoke. If smoke is detected, the system stores evidence for further actions. The first prototype version of the system needed an automatic adaptation procedure in order to calibrate far infrared and high-resolution images together. Mechanically, the system can be set up quickly in the chosen roadside location. A developed prototype system is one step towards future tools for authorities to automatically detect and classify vehicles emitting smoke. If a permanent set-up is desired, the system can be installed on a lamp post, beneath an overhead bridge or on other similar structures.

Automotive LIDAR sensor development scenarios for harsh weather conditions

This article focuses on development baseline for a novel LIDAR for future autonomous cars, which require perception not only in clear weather, but also under harsh weather conditions such as fog and rain. Development of automotive laser scanners is bound to the following requirements: maximize sensor performance, assess the performance level and keep the scanner component costs reasonable (<;1000 €) even if more expensive optical and electronic components are needed. The objective of this article is to review the existing automotive laser scanners and their capabilities to pave the way for developing new scanner prototypes, which are more capable in harsh weather conditions. Testing of scanner capabilities has been conducted in the northern part of the Finland, at Sodankylä Airport, where fog creates a special problem. The scanner has been installed in the airport area for data gathering and analyzes if fog, snow or rain are visible in the scanner data. The results indicate that these conditions degrade sensor performance by 25%, and therefore, future work in software module development should take this into account with in-vehicle system performance estimations concerning the visual range of the scanner. This allows the vehicle to adapt speed, braking distance and stability control systems accordingly.

The DESERVE project: Towards future ADAS functions

This article introduces the objectives and structure of the European research project DESERVE that is co-funded by the ARTEMIS-JU and national funding bodies. The project started in September 2012 with a duration of 3 years. The project aims to establish a new embedded SW and HW design by using a more efficient development process (including the enabling general platform concept and tool chain) in order to overcome challenges in reducing component costs and development time of future ADAS functions for modern vehicles. Both the process and the platform concept will be demonstrated with innovative ADAS functions in 3 passenger cars, 1 truck and 1 motorcycle. Embedded hardware and software units have been developed for improving electronic horizon band of vehicles by detecting objects in front. Moreover, driver/motorcycle rider awareness is analysed by monitoring his/her actions online. The systems need to be robust and reliable in different environment conditions (night time, rain, etc.). The DESERVE platform distinguishes three layers of intelligence: perception, application and intervention&warning control. The demonstrators will be based on software development tools from Elektrobit (ADTF) and Intempora (RTmaps). These tools are used to create 10 innovative ADAS applications as part of an integral ADAS development platform, following a new design and development process. Since the project is highly application oriented, the requirements have been adapted mainly from the ISO 26262 standard and the AUTOSAR framework which ensures compatibility with the existing automotive software environment.

In-vehicle sensor data fusion for road friction monitoring

This study presents the in-vehicle road friction estimation module to provide information for the driver to adapt his/her driving style. Our main aim is to enable support for saving fuel, minimising ecological impact and improving traffic safety. The module is based on the IcOR camera system developed by VTT which has been combined with inertia unit. Moreover, GPS receiver has been implemented to provide geographical position of the vehicle which is important for the infrastructure side of cooperative traffic services. The test results indicate clear correlation with the lateral accelerations and IcOR road state measures. The most important outcome is the chance to reduce number of false alarms of the IcOR system with acceleration based measurement. Moreover, the benefit of the acceleration data is the possibility to maintain the detection reliability of IcOR system in level of 70 % when driving in dark environment such as tunnels or night time.

Measurement of Driver’s Visual Attention Capabilities Using Real-Time UFOV Method

This paper proposes a new real-time method to measure the driver’s useful field of view (UFOV) while driving a car in ordinary traffic situations in an urban environment. This is called the real-time useful field of view (rUFOV) method to discriminate it from conventional UFOV measurement, which is typically performed offline and with laboratory equipment developed by Visual Awareness Inc. The proposed real-time method first tracks traffic objects that appear in the driver’s peripheral vision using a road video camera, checks the degree of the driver’s attention to these objects using a driver monitoring camera, and finally calculates the percentage reduction in the driver’s UFOV using a database acquired over an extended period of time. Preliminary results showed better performance than originally expected. The rUFOV method was then incorporated into a driving simulation environment to enable more precise measurement of the driver’s gaze angle. This enabled the performance of safer tests for identifying conditions under which mental load reduced the driver’s visual capabilities, thus increasing the possibility of hasty driving, as well as the incorporation of more accurate control parameters into simulation software for risky driving scenarios. Consequently, this paper proposes a new methodology for measuring the driver’s UFOV as a potential real-time driver support system with automatic intrusive HMI adaptation and immediate alarm functions. The evaluation was conducted in two phases. First, the system was tested in real traffic using typical vehicle equipment and technically worked with a performance level of 81%.In the second phase, more test runs were performed in the simulator environment, which enabled near accident scenarios to be created without risking traffic safety and it was also measured its reaction time..