Pothole in roads constitutes a major problem for both citizen and government. The pothole can create serious damages to the vehicles such as vehicle flat tires, scratches, dents and leaks. Thus, to detect these potholes and provide maintenance is highly time consuming and required lot of man power. Therefore this paper purposes a pothole detection system which is used for detecting the pothole and to analyze the image to determine its dimension. For detecting the pothole captured road images are inserted in the system then feature extraction and classifier performs. Lastly the predictor is done with the detection of pothole based on machine learning. The pothole detection system is derived from that assumptions that any strong dark edges within the extracted surfaces estimated a pothole if it observes certain constraints. Such as size color. Any outlines that meet these conditions are estimated as pothole by the algorithm. On the other hand for analyzing the image of pothole starts by converting the road surface images to gray scale and calculate the SURF points using Manhattan and Euclidean algorithm for calculating the dimensions of the pothole in the MATLAB environment, further comparing the system result obtain by these algorithms with the result calculated manually in order to find the error percentage of the system.

Keywords : Pothole Detection System, Manhattan Algorithm, Euclidean Algorithm, Knearest Neighbor Algorithm, MATLAB.