COMPUTER SCIENCE
and ENGINEERING

Image change detection, a technique in distinguishing “change/no change†area, can be regarded as a process of decision which can be solved by constructing hypothesis tests. Co-registered paired images (before image and after image) were studied in this report by exploring the standard steps of hypothesis test. Differencing of biomass index (dNDVI) is selected to denote the change variable considering its advantage of maximizing the spectral difference between vegetation and manmade features. Since this was a before-and-after study, t-test for paired observations was established. Adopting statistical test as inferential tool, land change decisions were made to incorporate both land change signals and noises. However, under complex spatial circumstance, many assumptions can be violated. After achieving the change decisions, spatial correlation and phenology problem were further checked as extreme outlier might lead to a false decision. Results indicate that phenology problems can pollute the change decisions and need to be further isolated in future studies.

Land change, Binary image change detection, Hypothesis test, Phenology problem.Text