To make full use of the color feature of a vehicle, a color transform model based on Karhunen-Loeve(KL) transform is proposed. Utilizing the model, all pixels can be transformed from RGB color space to a 2-D color space. On this 2-D color space, vehicle pixels and non-vehicle pixels will respectively concentrate on detachable areas. On this basis, vehicles can be detected from static images by using a classifier. Testing results demonstrate that when the model is applied the vehicle detection can be used with high accuracy.