Integracija stereo kamer za povečanje varnosti vožnje avtonomnega mobilnega robota

Lavrih, Lenart (author), Munih, Marko (mentor), Bizjak, Aleš (co-mentor)

V tem magistrskem delu je predstavljena integracija dveh kamer na avtonomnega mobilnega robota ter razvoj in testiranje algoritma za zaznavanje ovir. Glavni cilji so bili določanje najboljše postavitve kamer za zajemanje vidnega območja med vožnjo ter razvoj učinkovitega algoritma za zaznavanje okolice. Najprej smo analizirali trg in obstoječe rešitve v literaturi, določili sistemske zahteve, preučili značilnosti globinskih kamer in razvili matematični model za izračun vidnih kotov kamer. Ta model smo uporabili za preučevanje različnih postavitev kamer in določanje optimalne postavitve. Večji del naloge se je osredotočal na razvoj algoritma za zaznavanje ovir v okolici. Uporabili smo stereo kamere Intel RealSense D435f za zajemanje okolice. Celotna arhitektura sloni na odprto-kodnem okolju ROS in jeziku Python. Glavna knjižnica za obdelavo podatkov iz kamer je bila Open3d, za vizualizacijo pa smo uporabljali RViz. Poleg tega smo preučili metodo za zaznavanje ravnin z uporabo GPU, kar je omogočalo hitrejše obdelave podatkov. Na podlagi teh rezultatov smo razvili optimalen algoritem za zaznavanje ovir in ga nato preizkusili na AMR. Rezultati raziskave vključujejo analizo zaznavanja ovir v okolju, merjenje standardne deviacije razdalje, ločljivost detekcije ovir na tleh ter ponovljivost detekcije ovir v določeni coni. Vključeno je še testiranje različnih materialov ovir in detekcija ovir med vožnjo. Diskusija se osredotoča na obravnavo težav med delom. Tukaj so izpostavljeni izzivi, kot so problem bleščanja, zaznavanje ovir v zraku, postavitev kamer ter izboljšave.

• Language: Slovenian
• Keywords: avtonomni mobilni robot, stereo kamere, zaznavanje ovir, ROS, GPU
• Work type: Master's thesis/paper
• Organization: FE - Faculty of Electrical Engineering
• Year: 2023
• PID: 20.500.12556/RUL-148808
• COBISS.SI-ID: 165459459
• Publication date in RUL: 31.08.2023

Secondary language

• Language: English
• Title: Integration of stereo cameras to increase driving safety in mobile autonomous robot

Abstract

In this master's thesis, the integration of two cameras on an autonomous mobile robot is presented, as well as the development and testing of an obstacle detection algorithm. The primary objectives were to determine the best camera placement for capturing the visible area during driving and to develop an efficient algorithm for environmental perception. First, we analyzed the market and existing solutions in the literature, defined system requirements, studied the characteristics of depth cameras, and developed a mathematical model to calculate the cameras' field of view. This model was used to study various camera placements and determine the optimal positioning. A significant portion of the thesis focused on the development of the obstacle detection algorithm in the environment. We used the Intel RealSense D435f stereo cameras for environmental capture. The entire architecture is based on the open-source ROS environment and the Python language. The primary library for processing camera data was Open3d, while visualization was handled with RViz. Additionally, we explored a method for plane detection using a GPU, which enabled faster data processing. Based on these results, we developed an optimal algorithm for obstacle detection, which was then tested on the AMR. The research results include an analysis of environmental obstacle detection, measuring the standard deviation of distance, ground obstacle detection resolution, and repeatability of obstacle detection in a specific zone. Testing of different obstacle materials and obstacle detection during driving was also included. The discussion focuses on addressing potential issues that arose during the study. Challenges such as glare problems, detecting airborne obstacles, camera positioning, and possible improvements are highlighted.

• Keywords: autonomous mobile robot, stereo cameras, obstacle detection, ROS, GPU