Pamukkale Univ Muh Bilim Derg. 2020; 26(2): 371-384 | DOI: 10.5505/pajes.2019.69649

Vehicle detection in urban areas from very high resolution UAV color images

Müslüm Altun, Mustafa Türker
Department Of Geomatics Engineering, Faculty Of Engineering, Hacettepe University, Ankara, Turkey

It is very essential and important in urban areas for the automatic detection of geographical objects such as buildings, trees, and vehicles by using remotely sensed images. In this study, the stationary vehicles were detected from very high spatial resolution stereo color (Red, Green, Blue) unmanned aerial vehicles (UAV) images in urban areas. In the first step of the approach used, digital surface model (DSM) is generated from the stereo images. Then, digital terrain model (DTM) is generated from the DSM, and by using the DSM orthophotos are generated from IHA images. Next, the normalized digital surface model (nDSM) is calculated by taking the difference between the DSM and DTM to obtain only the ground objects.
After that, using the obtained nDSM data as an additional band, the multi-resolution segmentation and then object-based classification of the orthophoto are carried out. The approach was applied on two areas with different characteristics at Hacettepe University, Beytepe Campus. After comparing the results with the reference data, the vehicle detection accuracy was computed as 78.53% for the first test field (Field # 1) and it was computed as 92.15% for the second test field (Field # 2). The results show that the detection of stationary vehicles from very high spatial resolution UAV images can be performed with high accuracy using the proposed approach.

Keywords: Unmanned Aerial Vehicle (UAV), Orthophoto, Digital Surface Model (DSM), Digital Terrain Model (DTM), Segmentation, Detection of Stationary Vehicle

---

Çok yüksek çözünürlüklü renkli ÝHA görüntülerinden kentsel alanlarda araç tespiti

Müslüm Altun, Mustafa Türker
Hacettepe Üniversitesi Mühendislik Fakültesi Geomatik Mühendisliði Bölümü, Ankara

Kentsel alanlarda uzaktan algýlama görüntülerinden bina, aðaç, araç, vb. coðrafi nesnelerin otomatik olarak tespiti oldukça gerekli ve önemlidir. Bu çalýþmada, çok yüksek konumsal çözünürlüklü renkli (Kýrmýzý, Yeþil, Mavi) stereo insansýz hava aracý (ÝHA) görüntülerinden kentsel alanlarda sabit araçlarýn tespiti yapýlmýþtýr. Kullanýlan yaklaþýmýn ilk adýmýnda stereo ÝHA görüntülerinden sayýsal yüzey modeli (SYM) oluþturulmaktadýr. Sonra, SYM verisinden sayýsal arazi modeli (SAM) ve SYM kullanýlarak ÝHA görüntülerinden ortofoto oluþturulmaktadýr. Ardýndan, yalnýz yer üstü nesneleri elde etmek için SYM ve SAM verilerinin farký alýnarak normalize edilmiþ sayýsal yüzey modeli (nSYM) hesaplanmaktadýr. Daha sonra, elde edilen nSYM verisi ek bant olarak kullanýlmak suretiyle ortofotonun çoklu çözünürlük segmentasyonu ve ardýndan nesne-tabanlý sýnýflandýrmasý yapýlmaktadýr. Yaklaþým, Hacettepe Üniversitesi, Beytepe Yerleþkesi’nde farklý özelliklere sahip iki alan üzerinde uygulanmýþtýr. Oluþturulan referans veriyle yapýlan karþýlaþtýrma neticesinde, araç tespiti doðruluðu birinci test alaný (Alan#1) için %78.53 ve ikinci test alaný (Alan#2) için %92.15 olarak hesaplanmýþtýr. Elde edilen sonuçlar, önerilen yaklaþýmla sabit araçlarýn çok yüksek konumsal çözünürlüklü ÝHA görüntülerinden tespitinin yüksek doðrulukla yapýlabildiðini göstermiþtir.

Anahtar Kelimeler: Ýnsansýz Hava Aracý (ÝHA), Ortofoto, Sayýsal Yüzey Modeli (SYM), Sayýsal Arazi Modeli (SAM), Segmentasyon, Sabit Araç Tespiti

---