Textural rather than spectral approach to mapping seagrass on aerial photographs

Mapping change in seagrass beds over long periods of time (50+ years) is dependent on aerial photography, usually either panchromatic or colour. Remote sensing image processing usually relies on spectral techniques or image enhancement techniques to delineate seagrass boundaries, with varying levels of success. New data sources such as hyperspectral imagery or side-scan sonar are useful for current and future mapping, but this still leaves the need for new methods of extracting data from historical aerial photography.
Texture mapping may be a useful tool for helping to understand the content of an aerial photo image. It aids objectivity in identifying texture of a single type. The human eye is poor at recognising texture of a single type over even small distances across a single image let alone among images. In being objective, machine texture mapping is complementing the human eye.
This paper discusses the use of texture mapping techniques developed for side-scan sonar images on historical aerial photographic images. The test site was off the Adelaide, South Australia coastline where large changes in seagrass have occurred in the past 50 years.
In situations where much of the texture is not related to ground material but to other effects this may be of limited value, but in situations where there is control of active sources such as with side-scan sonar, this objectivity in image texture mapping may be
invaluable.