Real-time notification system of moving obstacles in vehicle-camera images using SSD and safe zones

• Main Authors: Panick Chiewcharn, Varin Chouvatut
• Format: Conference Proceeding
• Published: 2020
• Subjects: Computer Science; Decision Sciences
• Online Access: https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85084044569&origin=inward; http://cmuir.cmu.ac.th/jspui/handle/6653943832/70451
• Institution: Chiang Mai University

© 2020 IEEE. Road accidents are caused by a variety of reasons. Having something cutting in front of the car is one of the causes of the accident. Although the current hardware technology may be available to detect objects to prevent the accidental cause, such technology is still expensive, comparing to relying upon only a front camera. We thus propose a real-time analysis of images captured from a front camera to detect objects cutting in front of the car and create a warning system for the purpose. Our method of detecting and tracking objects names the single shot multi box detector, used with pre-trained object detection models. We contribute situations where objects moving with x-axis acceleration and how our regions of interest are determined, based on the actual measurement. If the object has an acceleration in the x-axis and is in the regions of interest, the system will alert the driver to make the right decision in order to prevent road accidents.