Relating the socio-economic conditions of urban neighborhoods through remote sensing imagery has been a major research challenge in remote sensing. It is clear from the literature that conventional per-pixel classification methods are not efficient in accurately classifying urban neighborhoods from high-resolution imagery due to the heterogeneous nature of urban environment. Instead, classification methods that utilize the textural characteristics of the image could dramatically improve the classification accuracy. Whether these textural methods can be used to detect the socio-economic conditions of urban neighborhoods remains a key question that this project will address. The goals of this doctoral dissertation research project are threefold: 1) to determine whether neural networks in combination with wavelet analysis are effective for the characterization and classification of urban areas; 2) to determine whether texture patterns can be associated with the social-economic conditions of urban neighborhoods; and 3) to explore the effects of image resolution on the classification accuracy. Two study areas, Atlanta, Georgia and Baton Rouge, Louisiana, will be examined and compared, and high-resolution IKONOS images of the study areas will be used. Fieldwork will be carried out and census and other socio-economic data will be used. Geographic phenomena operate at different scales; hence a multi-resolution approach will be used. Wavelet analysis is an efficient approach to studying textural patterns at different scales, whereas artificial neural networks can learn very complex patterns in the data. A combined methodology that utilizes both methods is expected to outperform other classification methods.
Although wavelet analysis and artificial neural networks have been employed separately to analyze remote sensing images, studies that bring them together are rare and the synergy is yet to be explored. This research will increase our knowledge of textures, texture methods, and their relationships with real features on the ground. The research will also enhance our appreciation of the scale effect on the classification of heterogeneous urban environment from high-resolution imagery. The identification and characterization of urban neighborhoods of different social-economic statuses holds the promises of opening up a new avenue to link remotely sensed images to the social-economic aspects of human activities. The methodology used in the project can be extended to study other social-economic applications such as detecting the sanitary or health conditions on the ground and to other classification scenarios (e.g., non-urban applications), hence will add significantly to the field of remote sensing, image processing, environmental assessment and monitoring, and urban analysis. As a Doctoral Dissertation Research Improvement award, this award also will provide support to enable a promising student to establish a strong independent research career.
