SUMMARYRecently, research on Intelligent Transportation Systems (ITS) has advanced, and its practical application has expanded to the image applications of measuring the traffic flow and detecting unusual events. In this paper, we describe a more advanced vehicle model identification technology, that is, a technology for detecting a particular model (model B of company A). This method does not read license plates or involve communication between the vehicle and the road, but identifies the vehicle model by using only roadside equipment. This algorithm recognizes the model by voting based on the arrangement of local features and has the following characteristics: (1) Detection is possible even if the target vehicle is occluded; (2) There is no dependence on the traveling position of the target vehicle; and (3) The vehicle region does not have to be extracted from the input image. First, we verify the above advantages in outdoor experiments by confirming the ability of the method called the eigenwindow method to recognize a specific model in the traffic flow. Next, we report on the successful improvement in the processing speed without a drop in the recognition accuracy by slightly modifying the implementation and using an image processing board.