Penerapan Filter Kalman Untuk Sensor Fusion Berbasis Kamera Mono Dan Sensor Ultrasonik Pada Sistem Persepsi Autonomous Car.

Rysmawan, Zeni Anggara (2022) Penerapan Filter Kalman Untuk Sensor Fusion Berbasis Kamera Mono Dan Sensor Ultrasonik Pada Sistem Persepsi Autonomous Car. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 02311840000079-Undergraduate_Thesis.pdf Restricted to Repository staff only Download (2MB)

Abstract

Metode deep learning menghasilkan sistem persepsi yang memiliki akurasi tinggi pada kondisi real time. Namun, performansi yang akurat bergantung pada komputasi yang tinggi. Pada penelitian ini, dilakukan rancang bangun sensor fusion berbasis sensor ultrasonik HC-SR04 dan kamera mono Webcam Xiaovv (1080p). Tujuan pembuatan sistem persepsi ini adalah untuk mengestimasi jarak serta koordinat dari suatu objek dengan komputasi rendah. Sensor ultrasonik menyediakan data pengukuran jarak, kemudian digabung dengan data estimasi jarak dari YoloV4-Tiny. Algoritma filter Kalman digunakan untuk mengurangi komputasi pada proses estimasi jarak dan koordinat algoritma sensor fusion. Uji performansi algoritma sensor fusion dilakukan dengan menggunakan dua skenario, yaitu objek diam dan bergerak. Proses estimasi jarak dan koordinat pada objek diam menghasilkan error rata-rata sebesar 0,556%. Sementara itu, pengujian pada objek bergerak mendapati RMSE sebesar 1,209. Uji pengoptimalan algoritma untuk sensor dengan filter Kalman menggunakan 46% ruang CPU. Sementara itu, pada sensor fusion tanpa filter Kalman yang memakan ruang komputasi sebanyak 68%.
===================================================================================================================================
A perception system with great accuracy under real-time situations is produced by the deep learning technique. High compute is necessary for correct performance, though. In this study, a fusion sensor based on the Xiaovv mono webcam camera (1080p) and HC-SR04 ultrasonic sensor was built. This perception system was created with the intention of quickly and accurately estimating an object's distance and coordinates. YoloV4-Tiny's distance estimation results are merged with distance measurement data from the ultrasonic sensor. The sensor fusion algorithm's estimation of the distance and coordinates uses the Kalman filter algorithm to streamline computations. Both fixed and moving object situations were used to test the sensor fusion algorithm's performance. There is an average error of 0,556 percent when determining the distance and coordinates of stationary objects. Testing on moving objects revealed an RMSE of 1,209, however. Using 46 percent of the CPU, test an optimization technique for sensors using a kalman filter. Meanwhile, 68 percent of the computer resources are used by sensor fusion without a kalman filter. There is an average error of 0,556 percent when determining the distance and coordinates of stationary objects. Testing on moving objects revealed an RMSE of 1,209, however. Using 46% of the CPU, test an optimization technique for sensors using a kalman filter. Meanwhile, 68 percent of the computer resources are used by sensor fusion without a kalman filter.

Item Type:	Thesis (Other)
Additional Information:	RSF 621.367 Rys p-1 2022
Uncontrolled Keywords:	sensor fusion, filter Kalman, kamera mono, ultrasonik, yolov4-tiny. fusion sensor, filter Kalman, mono camera, ultrasonic sensor, yolov4-tiny.
Subjects:	Q Science > QA Mathematics > QA402.3 Kalman filtering.
Divisions:	Faculty of Industrial Technology and Systems Engineering (INDSYS) > Physics Engineering > 30201-(S1) Undergraduate Thesis
Depositing User:	Mr. Marsudiyana -
Date Deposited:	11 May 2026 07:12
Last Modified:	11 May 2026 07:12
URI:	http://repository.its.ac.id/id/eprint/133126

Actions (login required)

View Item