The automatic detection of vehicles over urban areas is studied based on airborne decimeter resolution polarimetric Synthetic Aperture Radar(SAR) data. Urban areas are characterized by complex ground objects, which makes the detection of vehicles particularly challenging. Firstly, several classes of polarization information are extracted, including those from Freeman-Durden decomposition, Polarization Whitening Filter(PWF) and similarity parameters. Then, deep convolutional neural network algorithm is adopted to classify vehicle targets and other objects to realize the detection of vehicle targets in urban areas. Using experiments based on airborne decimeter resolution polarimetric SAR data, high detection rate under low false alarm rate can be obtained, demonstrating the effectiveness of the proposed method. When the three polarization characteristics are combined, the detection rate of 95.65% can be reached when the false alarm rate is 2.82%.