Project
# | Title | Team Members | TA | Documents | Sponsor |
---|---|---|---|---|---|
22 | Self Adjusting Rear-View Mirror |
Adam Magier Derek Wood Thomas Jarosz |
Anthony Caton | final_paper0.pdf other0.pdf other0.pdf other0.pdf photo0.jpg presentation0.pptx |
|
We propose designing a system which tracks driver eye position, and adjusts the orientation of the rear-view mirror such that the full rear window remains visible to the driver through the mirror. Moving your head only a few inches while driving can result in a majority of the rear window becoming obstructed from view. Position adjustments by a driver accumulate over long trips, and the result is significant periods of time where the driver is either driving with the rear window partially obstructed, or driving distracted while re-adjusting the mirror. As there are 40 million injuries annually from automotive accidents, even if this product prevents only a small fraction of accidents the benefit is significant. One or two IR cameras[1] will be used to image the cabin, illuminated with IR LEDs. A microcontroller running a neural network will be used to locate eyes in the image (even if sunglasses are present). An existing neural network can be downloaded[2], or training data can be labeled[3] and a neural network can be built. A spherical wrist robotic arm will be used to orientate the mirror. Open source files[4] can be used as the basis for our motor mounts. We can use DC motors[5] to move the joints, and rotary encoders[6] for feedback on joint angles. Additional hardware includes designing an analog control system for the motors, and a power system (powered by the car battery). Note that Berkley students did a similar project[7]. Their design only had two degrees of freedom, and thus cannot achieve perfect orientation. Additionally, their design obstructs the windshield. Our design has three degrees of freedom, and would not obstruct the wind shield. Finally, using only one camera (directly behind the mirror) would be more cost effective (and possibly perform better) than their design. Names, NetIDs: Thomas Jarosz, Tjarosz2 Derek Wood, Drwood2 Adam Magier, Magier2 Link to original post: https://courses.engr.illinois.edu/ece445/pace/view-topic.asp?id=26983 References- 1: IR Camera https://www.robotshop.com/en/arducam-mt9m001-13-mp-hd-cmos-infrared-camera-module-adapter-board.html?gclid=CjwKCAjwrNjcBRA3EiwAIIOvqzWkb9C8rrUAjbolDv2cOUpXW6jARzQ-nufIlwaRA7SPCLzTrTpFIhoCT5MQAvD_BwE 2: Downloadable eye tracking software https://imotions.com/blog/free-eye-tracking-software/ 3: AWS Rekognition to label images https://aws.amazon.com/rekognition/ 4: Open Source Motor mount https://dxarts.washington.edu/wiki/pan-tilt-servo-mount 5: DC motor https://www.digikey.com/product-detail/en/dfrobot/FIT0492-A/1738-1268-ND/7087165 6: Rotary Encoder http://p3america.com/products/index.php?main_page=product_info&cPath=479_480_694&products_id=948&gclid=CjwKCAjwrNjcBRA3EiwAIIOvq2cjAa8ZGr_cbFoAeaNoLpMxagrXV_0D9_26JNrMySFAJEDjs9K8ExoCxoIQAvD_BwE 7: Berkley’s Self Adjusting Rear View Mirror (note, this is the only documentation we could find regarding this project) http://courses.me.berkeley.edu/ME102B/Past_Proj/s06/15/index.htm |