Detekcija prometnih znakova heterogenom kaskadom klasifikatora

undergraduate thesis

• Title (english): Traffic sign detection using heterogeneous classifier cascade
• Author: Karlo Čačković
• Mentor: Zoran Kalafatić (mentor)
• Committee member: Zoran Kalafatić (predsjednik povjerenstva)
• Committee member: Alan Jović (član povjerenstva)
• Committee member: Zvonimir Pavlić (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2014-07-14, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing

Abstract

Tema ovog završnog rada je ostvariti sustav za detekciju prometnih znakova heterogenom kaskadom klasifikatora u video zapisima snimljenim iz vozila u pokretu. Takav klasifikator sastoji se od kaskadnog klasifikatora kliznim prozorom temeljenog na Haarovim značajkama te od klasifikatora strojem s potpornim vektorima temeljenog na histogramima orijentiranih gradijenata.
Koncept takvog sustava je detekcija kaskadnim klasifikatorom te moguće prometne znakove predati dodatnom stupnju heterogene kaskade koji odredi je li unutar kliznog prozora prometni znak ili nije te označi detektirani znak na video zapisu. Sustav je testiran na javno dostupnom skupu slika i stopa ispravne detekcije je 83,125%. Problem predstavlja brzina detekcije pa ovako implementiran sustav nije pogodan za real-time detekciju u video zapisima, te velik broj neispravnih detekcija.

Abstract (english)

The subject of this BSc thesis is to implement a program for traffic sign detection using heterogeneous classifier cascade. The input data for this system are video sequences recorded from a moving vehicle. This classifier consists of a sliding window cascade classifier based on Haar – like features and an additional classifier with support vector machines based on histograms of oriented gradients.
The concept of such a program is detecting traffic signs with cascade classifier and forwarding the candidates to an additional stage of heterogeneous classifier cascade. This additional stage detemines whether sliding window contains a traffic sign or not. The program has been tested on a publicly available set of images. The percentage of correct detection is 83,125%. Problem is the detection speed and such a program isn't good for real-time detection in video sequences. Additional problem is high number of false positive detection.

• Keywords: Računalni vid; Haarove značajke; Kaskada klasifikatora; Algoritam Viole i Jonesa; Histogrami orijentiranih gradijenata; Stroj s potpornim vektorima; Video zapis; Trokutasti prometni znakovi
• Keywords (english): Computer Vision; Haar-like features; Classifier cascade; Viola-Jones algorithm; Histograms of oriented gradients; Support Vector Machines; Video capture; Triangular traffic signs
• Language: croatian
• URN:NBN: urn:nbn:hr:168:421606
• Study programme: Title: Computing; Study programme type: university; Study level: undergraduate; Academic / professional title: sveučilišni/a prvostupnik/ prvostupnica (baccalaureus/baccalaurea) inženjer/inženjerka računarstva (sveučilišni/a prvostupnik/ prvostupnica (baccalaureus/baccalaurea) inženjer/inženjerka računarstva)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Closed access
• Created on: 2018-02-05 17:28:37