Robust real time road detection is a crucial task for autonomous transport systems and driver assistance systems. Several state of the art road detection techniques make use of the illuminant invariant feature space for classifying pixels. One of the assumptions used in the derivation of illuminant invariant image from the RGB image is that the surface being captured is a lambertian surface. The specular reflection of sunlight from the road surface, which is a common daytime occurrence violates the lambertian assumption. Thus, the methods based on illuminant invariant fail to detect the road region containing specularities. Another aspect where the algorithms need to be improved upon is the detection of white markings painted over the roads. These markings are not built into the road model. Hence, the road detection algorithms fail to classify them as part of road region. Here, we propose a novel specularity detection and removal method devised specifically for road detection which inherently eliminates the white markings present in the road image. This modification improves the road detection accuracy in the presence of specular reflections as shown by the results. Apart from the improvement in accuracy of the algorithm, it needs to function in real time on mobile platforms with low power consumption. For this, we implemented the algorithm using OpenCV on two low power development boards- BeagleBone Black and Raspberry pi-2, the latter being chosen for its capability of multiprocessing using 4 cores. These experiments provide a proof of concept of real time implementation.