A New Gabor Filter-Based Method for Automatic Recognition of Hatched Residential Areas

Abstract

Extracting residential areas from historical raster topographic maps benefits to analyze land type change. The existing algorithms have the shortcomings including easily misidentifying objects and low positional accuracy of the identified boundary, so we have presented a new automatic recognition method based on Gabor filter for extracting residential areas from historical raster topographic maps. First, the method detected the hatched areas using Gabor filter, Gaussian smoothing, binarization, erosion, and an operation. Afterward, an endpoint of a hatched line on the top of the hatched areas was taken as the starting point, then tracing a boundary point of the residential area along a hatched line toward northeast 45 degrees direction using dynamic filling strategy. Second, an adjacent boundary point was traced according to the pixel value relation of eight neighborhoods by taking the first found boundary point as a starting point. In the tracking process, removing the noises using the pixel value relation of the neighboring pixels and the designed strip detector. Ultimately, the residential boundary was obtained. The experiments were carried out on the samples of three typical areas. The results showed that our method was effective and practical, and outperformed the previous methods in integrity and positional accuracy of the residential boundary.

Keywords

• Residential areas recognition
• raster map
• Gabor filter
• data mining
• image processing