Presentation on theme: "Drowsy Driver Warning System Project Description Special Thanks The team would like to thank Dr. Roy Czernikowski for his assistance and guidance throughout."— Presentation transcript:
Drowsy Driver Warning System Project Description Special Thanks The team would like to thank Dr. Roy Czernikowski for his assistance and guidance throughout the project. The team would also like to thank Richard Tolleson for his continued technical support and miscellaneous project supplies. The team recognizes Michael Snook for his assistance in the area of circuit design and analysis of the team’s electrical components. Finally the team would like to show its appreciation towards their classmates for their positive feedback and suggestions, which ultimately contributed to the success of the project. Resources 1: http://www.sleepfoundation.org/site/apps/nlnet/content3.aspx?c=huIXKjM0IxF&b=2464479&ct=3445645 2: MPT: http://mplab.ucsd.edu/grants/project1/free-software/mptwebsite/introduction.html 3: AForge: http://code.google.com/p/aforge Acknowledgements/Resources Mark Cataldi, Jesse Harvey, Dieter Laskowski Senior Design Projects II – Fall 2008 Department of Computer Engineering, Rochester Institute of Technology, Rochester, NY ComponentQuantity Total Retail Cost Total Cost to Students Laptop with Windows XP and USB drivers 1$1000.00Already Owned Logitech QuickCam® Pro 9000 2$200.00$175.00 IR LEDs2$6.00 USB to Serial Adapter 1$20.00Free Drowsiness slows reaction time, decreases awareness, and impairs judgment just like drugs or alcohol. Independent studies 1 of the Pennsylvania Turnpike and the New York State Thruway estimate that 50 percent of fatal crashes on those roads are caused by drowsy drivers. Many people would never consider drinking and driving, but many fail to recognize that driving drowsy can be just as fatal as driving drunk. The Drowsy Driver Warning System is designed to prevent a driver from falling asleep at the wheel. Two cameras are mounted inside the automobile cabin. One of the cameras faces the driver in order to monitor the driver’s eyes. The other camera faces the front windshield of the automobile in order to monitor lane position. When the system determines that the driver is becoming drowsy or the automobile is drifting out of its current lane, the warning system will enable an audible buzzer and massage pad in order to awaken the driver. The Team Team members from left to right: Mark Cataldi, Jesse Harvey and Dieter Laskowski Project Costs Image processing is handled by two software libraries: Machine Perception Toolbox (MPT) and a custom lane departure detection application that utilizes the AForge.Net framework. The Machine Perception Toolbox 2 (MPT) supplies cross-platform libraries for real- time perception primitives, including face detection, eye detection, blink detection and color tracking. AForge.NET 3 is a C# framework designed for developers and researchers in the fields of Computer Vision and Artificial Intelligence. It includes image processing, neural networks, genetic algorithms, machine learning, and much more. A DrowsyDriver C++ library was created to interface with the different image processing libraries and the microcontroller. When the application launches, DrowsyDriver initializes both MPT and the lane departure application. DrowsyDriver then monitors the image processing results and communicates its status with the microcontroller through PC RS-232 serial port. The code running on the microcontroller was written in MCU12 assembly. The microcontroller reads the serial port and interprets the data to manipulate specific output ports and track the warning system’s status. The microcontroller also reads the input values from the driver control box in order to inhibit or enable the proper alerts. SoftwareHardware Cameras: The Logitech QuickCam® Pro 9000 was selected for it’s premium autofocus and ultra-wide field of view in hopes of providing the best possible input to the image processing libraries. The IR blocking filter on the driver facing camera was physically removed to allow for IR illumination. When a hazardous situation is detected, the warning system will provide both a tactile and audible response in order to wake the drowsy driver. The audible response is driven by a Piezo buzzer operating in the 70 - 80 dB range. The tactile response is provided by a USB massage pad that has been modified to accept control signals from the microcontroller. Infrared LEDs are illuminated when MPT is unable to detect the driver’s face using ambient lighting. This enables face monitoring at night or under low light situations. The IR lighting also helps reduce glare on eyeglass lenses. A Freescale HCS12 microcontroller is responsible for activating the warning system, enabling the IR LEDs, and interfacing with the driver control box. A driver control box is provided to allow the driver to inhibit face and lane detection independently and simultaneously. Status LEDs are also mounted in the control box to notify the driver if blink or lane departure detection are not functioning reliably. ComponentQuantity Total Retail Cost Total Cost to Students HCS12 Board1$270.00Free Massage Pad & Buzzer 1$21.00 10 Gauge Jumper Cables 1$12.00 Power Inverter1$32.00 Driver control box LEDs and switches 1$10.00 Project Board & Enclosure 1$5.00 Total Retail Cost: $15763 Total Cost to Students $261 Imaging SystemWarning & Support System Screen capture of the Lane Departure Detector after Sobel edge detection (not shown) and Hough line transform. The Hough lines are represented by the green lines drawn from the origin, which are then decoded into the lane edges and drawn in red. To eliminate extraneous data, only edges in the area between the two horizontal lines are analyzed. The crossbar (pink) shows the region which is considered a dangerous lane departure. If a Hough line passes through the cross bar it will trigger an alarm message to be sent. Screen capture of the blink detection application supplied by the Machine Perception Toolbox. The red box detects a face found by the application with white boxes around the eyes. The red bar to the left shows the blink threshold which will grow or shrink based on how likely it is that the subject is blinking. The bar will turn blue if a blink is detected. After a specified frequency of blinks, the blink detector will activate the warning system.