Peat moss forest fire is one of damaging disasters. It disturbs people’s activity
and health and also reduces ecological livings; thus, this disaster becomes attention of
large society either nationally or internationally. Peat moss forest fire can be identified
by remote sensing technology. The development of remote sensing technology by using
MODIS imaging satellite so far has been used in many fields and one of them is in
controlling forest fire multitemporally. Hotspots of information obtained from the MODIS
image processing can be obtained by image feature extraction using a wavelet , so that
the results of this extraction can be input in the process of detection of hotspots using
Artificial Neural Networks . In this research data for surface temperature used Terra
MODIS satellite by taking an advantage of canal 31 and 32 and using Coll’s, et.al (1994)
algorithm, The data of forest fire were from forest fire service of Dumai City, for feature
extraction using a wavelet Haar , Coiflet1 and Symlet5, and backpropagation neural
networks to recognize patterns hotspot. This research was conducted in Dumai City
Province of Riau. Result of this research shows that the use of Canal 31 and 32 through
Terra MODIS imaging satellite can be used for detecting points of fire which is found
that there are more then 17 points of fire within temperature ranges of 270C – 320C,
While the results of the Pattern Recognition hotspots using short - wave and
Backpropagation Network with image input in the form of satellite images with 8 bit and
size 512 x 512 derived from satellite data turned out to give good results with a
performance of 100 % on the image of the wavelet decomposition Haar and Coiflet 1 ,
while for wavelet Symlet5 gives performance by 40 % .