The work contained in this thesis is toward producing an objective cyclone climatology of the extra-tropical regions of the Southern Hemisphere using observational data. The data identified as being the most appropriate for this task was the high resolution multi-spectral imagery produced by the Advanced Very High Resolution Radiometer (AVHRR) on board the National Oceanographic and Atmospheric Administration (NOAA) series of polar-orbiting satellites. A series of algorithms are presented to perform the automatic detection and tracking of cyclones in the Southern Hemisphere. Cloud detection was performed using a number of spectral and textural measures. Individual cloud objects were segmented using thresholding techniques. These cloud objects were pre-processed using an image normalisation algorithm to create an image (feature vector) which was invariant to rotation, translation, scaling and skew. Identification and tracking of cyclones was performed by finding the maximum cross-correlation between the normalised target images and a set of pre-defined normalised templates. Problems with automatically distinguishing between cloud and the surface near Antarctica still remain, and subsequently segmentation of the imagery into individual cloud objects was limited to certain cases. Identification of cyclones by matching with the pre-defined normalised templates was also shown to have only very limited ability to distinguish between cyclone cloud objects and non-cloud objects. Tracking of the cyclones once they have been manually identified and measured was shown to be promising, with the algorithm successfully tracking the cyclone in 16 out of 22 pairs of satellite images taken from 4 case studies of real Southern Hemisphere cyclones. Limitations of the tracking scheme were identified and improvements have been suggested.