The conventional pixel-oriented classification is the most commonly used approach in remote sensing for land use product extraction. The object-oriented classification based on image segmentation is an alternative, which uses pixel context, texture and shapes, in addition to their spectral characteristics. This paper reports on a comparative study between supervised pixel-oriented and object-oriented classifications in a precision agriculture context using three hyperspectral images on our first study site, and a set of hyperspectral and multispectral images for a second site. The images for the first site, owned by the horticulture research and development centre (Agriculture Canada) at L'Acadie in southern Quebec, were acquired with the Compact Airborne Spectrographic Imager (CASI) sensor at three different altitudes, providing three different spatial resolutions: 1, 2 and 4 m. For the second site, located at the Indian Head Agriculture Research foundation in Saskatchewan, a Probe-1 hyperspectral image was acquired as well as a multispectral IKONOS image. After calibration and correcting the imagery, pixel-oriented classifications were carried out using the maximum likelihood algorithm and object-oriented classifications with a nearest neighbor classifier after region growing hierarchical segmentation. After segmentation, statistical comparison on the mean difference to neighbor objects confirmed that the segments had minimum mixing effects in respect to other segmentation levels and neighboring ground entities. After accuracy analysis on the classifications for the first site, the segmentation process allowed the use of a spatially coarser hyperspectral image (4 m with kappa of 0.8268) to achieve better results than pixel oriented classification of a spatially finer hyperspectral image (1 m with kappa of 0.7730), in the task of delineating agricultural classes. For the second site, results are still consistent. Object oriented results of the hyperspectral Probe-1 image (kappa of 0.9628) significantly exceed the pixel oriented results (kappa of 0.9217). Similarity is observed with IKONOS multispectral imagery (kappa of 0.9371 for object oriented and kappa of 0.8926 for pixel oriented). Image segmentation is therefore an important technique to achieve high accuracy in classification of land cover classes. Hyperspectral imagery also has a strong power of discrimination between many agricultural classes.
| Identifer | oai:union.ndltd.org:uottawa.ca/oai:ruor.uottawa.ca:10393/28097 |
| Date | January 2009 |
| Creators | Lanthier, Yannick |
| Publisher | University of Ottawa (Canada) |
| Source Sets | Université d’Ottawa |
| Language | French |
| Detected Language | English |
| Type | Thesis |
| Format | 107 p. |
Page generated in 0.0143 seconds