We propose a vehicle detection algorithm for Intelligent Surveillance used in open pit area . The algorithm detects vehicle in the image using Histograms of Oriented Gradients features (HOG) as the image description, combined with support vector machine (SVM) classifier training. In order to improve the detection efficiency and accuracy, Hough transform line detection method to extract the region that contains the target to reduce the search range. The experiments show that the algorithm can be able to recognition the target in complex images, with high degree of accuracy, robustness for the different environments of the experiment.