With the tremendous increase in the number of air passengers in the past years, aviation safety has been of
utmost importance. At any given point of time, there will be several flights lining up for landing. Landing in
good visibility conditions is not a problem. However, the problem arises when we have poor visibility conditions,
especially foggy conditions. The pilot finds it difficult to land the flight in poor visibility conditions because
of the difficulty to spot the runway clearly. This paper presents a novel method for detecting the runways and
hazards on it in poor visibility conditions using image processing techniques.
The first step is to obtain the images of a runway on a clear day and compute the smoothness coefficient
followed by edge detection, using the SUSAN edge detection algorithm and then finally develop a database of
the smoothness coefficients and edge detected images. Now, for the foggy images we compute the smoothness
coefficient. Typically, foggy images have low contrast. Hence, before we perform edge detection, we enhance
the image using Multi-Scale Retinex (msr). msr provides the low contrast enhancement and color constancy,
required to enhance foggy images, by performing non-linear spatial/spectral transforms. After enhancement,
the next step is to run the same edge detection algorithm with appropriate thresholds. Finally we determine a
hazard by comparing the edge detected images of images taken under clear and foggy conditions. The paper also
compares the results of the SUSAN edge detection algorithm with the state of art edge detection techniques.