HOG (Histograms of Oriented Gradient) is a feature which is a basis of the modern object detection algorithms. I applied HOG-SVM (Support Vector Machine) detector to our ship detection system and evaluated it quantitatively. Firstly, I gathered large number of navigational images from ship mounted camera and manually annotated the area of ships in the images as the ground truth for training and evaluation. Then I carried out k-fold cross validation to evaluate HOG-SVM for various parameters. The result shows that HOG-SVM achieves 99% true positive and 0.0% false positive for ships with resolutions larger than 128x64.