การตรวจหาสิ่งกีดขวางที่มีการเคลื่อนไหวในภาพที่ถ่ายโดย กล้องติดยานพาหนะ

Abstract

This paper proposes analysis of images from the front camera to detect and create a warning system that are likely to cross the front of the car without a sensor. Using the Single Shot Multi Box Detector (SSD) to detect object and introduced a new method of object tracking by storing mean values of objects in the x-axis and axis. y, the type of object detected and similarity percentage of detected object compare with Common Object in Context Dataset (COCO) in matrix. Assigning 1 image to hold 100 data, the matrix object will be 4 x 100. Compare the matrix of current frame with the previous frame by focusing only on objects that move in the x-axis, as well as determining the region of interest (ROI) from the actual measurement. If the object moves inward on the x-axis and is in the area of interest then program will alert the driver to make the right decision to prevent road accidents. The results show that in the experiment, the test of object detection and alert while the vehicle is not moving has an accuracy rate of 93.33% and the test of the object detection and notification while the vehicle is moving at a speed of 20 - 40 kilometers per hour has an accuracy rate of 76%