Implementasi Canny Edge Detection dalam Menentukkan Arah Pergerakan Robot untuk Menghindari Obstacle

Abstract

In this study, realtime obstacle avoidance based on image processing was developed. The obstacle shape used still has a pedestal surface from various related studies. The use of obstacle shapes in the form of gaps is still rarely found from various studies. The obstacle that will be avoided in this study is that there is a pedestal surface and a gap. If the height of the robot is shorter than the gap-shaped obstacle, the greatest possibility will be able to crash because of its inability to detect. Based on this, this research uses the canny edge detection method, which can produce an pixel value averaging to generate the direction of movement. Trash cans and chairs are obstacles that represent pedestal surfaces and gaps. In the research conducted, namely determining the direction decisions of various robot positions against obstacles. Then it shows the difference in frame reading speed when avoiding trash cans and chairs is 1 fps. The frame reading speed when avoiding the trash can is 8 fps and the chair is 7 fps. The delay in frame reading causes a delay in detection, thus affecting the decision of movement direction. The speed value on a robot greatly affects the length of time for the avoidance process. The length of time to avoid the trash can at a speed of 35 PWM speed is 5.07 seconds and 40 PWM speed is 4.58 seconds. And the length of time to avoid the chair at 35 PWM speed is 6.11 seconds and 40 PWM speed is 4.71 seconds. Then, there is an experiment that compares the success rate accuracy between the canny edge detection method and the HC-SR04 sensor from various robot positions towards the obstacle. The accuracy of the success rate using the canny edge detection method against the trash can test was 90.47% and the chair was 85.71%. While in the use of the HC-SR04 sensor, the trash can is 80.95% and the chair is 71.42%. The difference in the percent accuracy of the success rate makes the canny edge detection method very suitable for use as an obstacle avoidance method.