Perancangan Omni Vision Camera sebagai Sistem Pendeteksi Objek untuk Robot Sepak Bola

Abstract

Robot adalah kumpulan perangkat elektronika terintegrasi yang dapat bergerak dengan kontrol manusia atau otomatis menggunakan program yang telah ditanamkan terlebih dahulu. Salah satu jenis robot yang banyak dikembangkan yaitu robot sepak bola beroda yang dirancang memiliki kemampuan mendeteksi, menggiring, dan menendang bola ke gawang lawan. Kemampuan penting yang harus dimiliki yaitu kemampuan robot mendeteksi posisi bola. Permasalahan yang dihadapi umumnya yaitu kamera hanya memiliki field of view (FOV) 73.98° pada jarak 1,5 meter sehingga ketika bola berada pada posisi titik buta robot, robot harus berputar untuk mendeteksi posisi bola. Untuk menghindari kehilangan citra bola, robot harus dibekali kemampuan penglihatan seluruh sisinya yaitu menggunakan omnidirectional camera, akan tetapi membutuhkan penyesuaian algoritma pada robot serta harganya relatif mahal sehingga tidak efektif. Sistem pendeteksi objek memanfaatkan omni vision camera yang menggunakan hyperbolic mirror yang diusulkan dapat mendeteksi objek di sekitar keliling robot. Omni vision camera mendeteksi objek secara realtime dengan metode color based filtering menggunakan algoritma Color Connected Components (CCC). Sistem pendeteksi objek menggunakan omni vision camera mampu mendeteksi objek secara 360° sudut visual robot ketika intensitas cahaya berada antara 60 hingga 210 lux. Kemampuan waktu pendeteksian lebih efisien dengan waktu rata-rata 6,33 detik lebih cepat dibandingkan dengan single camera vision yaitu 7,65 detik serta dapat mendeteksi objek bola dengan jarak terdekat 10 cm dan jarak terjauh 160 cm.  

Robot is integrated electronic devices that can move with human control or autonomous using a program that has been implanted in advance. One type of robot that has been widely developed is a wheeled soccer robot which is designed to have the ability to detect, dribble, and kick the ball into the opponent's goal. An important ability that must be possessed is the robot's ability to detect the position of the ball. The common problem is that the camera only has a field of view (FOV) of 73.98° at a distance of 1.5 meters, so that when the ball is in the robot's blind spot position, the robot must rotate to detect the ball's position. To avoid losing the object image, the robot must be equipped with the ability to see all sides, using an omnidirectional camera, but requires algorithm adjustments on the robot and the price is relatively expensive so it is not effective. An object detection system utilize omni vision camera that using hyperbolic mirror is proposed which can detect objects around the robot's circumference. Omni vision camera detects objects in real time with the color based filtering method using the Color Connected Components (CCC) algorithm. The object detection system using an omni vision camera is capable of detecting objects 360° from the robot's visual angle when the light intensity is between 60 to 210 lux. The ability of detection time is more efficient with an average time of 6.33 seconds faster than single camera vision, which is 7.65 seconds and can detect image of objects with the closest distance of 10 cm and the farthest distance of 160 cm.