pixel based better, the resolution of landsat is too low for object oriented method.

While object-based image analysis (OBIA) has seen to work better for high resolution images, the technique has also been applied to medium coarse resolution images such as the Landsat.  Dorren et al. (2003) used the technique for Landsat to map the forest cover in steep mountainous terrain.  Jobin et al. (2008) also used OBIA for Landsat 7 to identify potential habitat of the grasshopper sparrow.  A much recent study by Campbell et al. (2015) also used the medium-coarse resolution of Landsat for land cover mapping and change analysis in Northeastern Oregon.  I think both high resolution and medium coarse resolution images can benefit in the potential of the OBIA technique for LULC classification.

Dorren et al., 2003. Improved Landsat-based forest mapping in steep mountainous terrain using object-based classification. Forest Ecology and Management, 183 (1–3) (2003), pp. 31–46

Jobin et al., 2008. Object-based classification as an alternative approach to the traditional pixel-based classification to identify potential habitat of the grasshopper sparrow. Environmental Management, 41 (1) (2008), pp. 20–31

Campbell et al., 2015. Optimal Land Cover Mapping and Change Analysis in Northeastern Oregon Using Landsat Imagery. Photogrammetric engineering and remote sensing, 81 (1), pp. 37 - 47

Hi Girish,

Eric explained about the OOC and its accuracy over the pixel based classification system. You can use the moderate resolution images also for the OOC.

Most studies by OOC is done in high resolution images and also for smaller areas.

But in your case you are covering more area and your studies are at regional scale.  SO It is advisable to the pixel based classification.

Good luck.

Vijith

Hi Girish,

                 The classification which one is required is dependent on the purpose of your work. If you are showing little variation, then unsupervised classification is necessary. Please read the following article that might help you.

Best of luck

http://www.isprs.org/proceedings/XXXVIII/4-C1/Sessions/Session1/6711_Proc_pap.pdf

Completely agree with Supriti and Richard.

According to my recent experience it was nice to see per pixel classification for urban area since I was able to see small green areas (two or three pixels) here and there. This being not possible via OOC with larger object size.