Land Use/Cover Classification over Small Areas Using Conventional Digital Camcorder Imagery Based on Frequency-Based Contextual and Neural Network Classification Techniques

Abstract

An airborne survey was conducted to produce land cover/use maps. The feasibility of using a conventional digital camcorder to acquire remotely sensed data was investigated, and the imagery for land cover mapping using remote sensing technique was evaluated. The study area was the Universiti Sains Malaysia campus, Penang, located in Peninsular Malaysia. Digital images were taken from a low-attitude light aircraft, Cessna 172Q, at an average altitude of 2.4384 km above sea level. The use of a digital camcorder as a sensor to capture digital images is more economical compared with other airborne sensors. This technique is designed to overcome the problem of obtaining cloud-free photographs from a satellite platform in equatorial regions. Digital video imageries were taken in the red, green, and blue bands. A comparison between frequency-based contextual and neural network classification techniques for analyzing digital camcorder imagery is presented. Frequency-based contextual and neural network classification techniques were applied to the digital camera spectral bands (red, green, and blue) to extract the thematic information from the acquired scenes. The classified map was compared with the ground truth data, and accuracy was evaluated by an error matrix. Results indicate that a conventional digital camcorder can be used to acquire digital imageries for land cover/use mapping of a small area of coverage.