GEO 420: Digital Image Processing
The size of different land cover usage fluctuate as time progresses due to a number of different variables such as urban sprawl, an increase in population, and the amount of rainfall. This research project analyzes the land use changes throughout the study area which encompasses Los Angeles County and small portions of adjacent counties. The County of Los Angeles is the largest county in the United States based on population which is at 10,017,068 according to a 2013 estimate by the US Census Bureau. In this study, the reader will find and see that these three very important variables have impacted the land cover usage of the Los Angeles County Region.
INTRODUCTION & BACKGROUND
The Los Angeles County region is one of the most populous in the United States and as a result, the amount of urban land cover usage has increased dramatically over such a short time-span and this is due to rapid increase in population which in result creates a higher demand of urban area to accommodate living space.
The chosen study area is 50 x 70 miles (3500sqmi), which stretches from the eastern portion of Ventura County and then 70 miles eastward capturing the western portion of San Bernardino County. The two years that were selected for this study are 1989 and 2005. The reason why this study area was designated in particular is because it is expected to have a lot of different land cover changes primarily due to the sheer amount of people.
Digital images were captured by sensors capable mounted on satellites (which can also be mounted on aerial vehicles as well); which do so by measuring wavelengths of the electromagnetic spectrum. Using data captured from these sensors, a digital image can be composed in a variety of different ways. The two composites used for this study are "True Color," which produces the image in a color you would see things naturally, and a "near-infrared" composite, which makes analyzing vegetation much easier because vegetation appears as red.
The program used to process this data is called ERDAS IMAGINE, a remote sensing application that is capable of data processing and allows the user to perform a variety of different operations and enhance/display digital images.
DATA & METHODS
Classification is the process of sorting image pixels into a number of categories according to their spectral values. The three type of classifications used for this analysis are: unsupervised, supervised, and NDVI (Crisp).
An unsupervised classification analysis (figure 2a and 2b) was done for the first analysis of the study area. An unsupervised classification is done purely through a computational algorithm called the isometric clustering method. The algorithm uses the spectral distance formula to form clusters. Cluster means are then chosen and each time the clustering repeats; the means of these clusters fluctuate accordingly. This type of classification requires minimum effort from the user and the result provides a quick, but not absolutely accurate, representation of land cover usage. This phase of the study is to get a brief idea of all the different land covers.
A supervised classification analysis was then performed in order to obtain a more accurate result of land cover usage. This type of classification requires the input of the user. The methods of the user include “ground truthing,” which is visiting specific sites in person to confirm their land use type, and manually picking common "training sites" to categorize them into a specific class. The algorithm used in a supervised classification will go through all the signatures and match similar them to the classes that the user has created. Physical groundtruthing sites (figure 7) were chosen based on discrepancies seen in the unsupervised classification. Online groundtruthing, which was done using Google Earth, was performed to…