Selective Multi-Convolutional Region Feature Extraction based Iterative Discrimination CNN for Fine-Grained Vehicle Model Recognition

With the rapid rise of computer vision and driverless technology, vehicle model recognition plays a huge role in the common application and industry field. While fine-grained vehicle model recognition is often influenced by multi-level information, such as the image perspective, inter-feature similarity, vehicle details. Furthermore, pivotal regions extraction and fine-grained feature learning have become a vital obstacle to the fine-grained recognition of vehicle models. In this paper, we propose an iterative discrimination CNN (ID-CNN) based on selective multi-convolutional region (SMCR) feature extraction. The SMCR features, which consist of global and local SMCR features, are extracted from the original image with higher activation response value. As for ID-CNN, we use the global and local SMCR features iteratively to localize deep pivotal features and concatenate them together into a fully-connected fusion layer to predict the vehicle categories. We get better results and improve the accuracy to 91.8% on Stanford Cars-196 dataset and to 96.2% on CompCars dataset.

I. Introduction

Fine-grained vehicle recognition is an important subject in computer vision field, which indicates that different subtypes in the same large class and recognize the category information of vehicles should be recognized, such as vehicle make, vehicle model or manufacture year. For example, we need to retrieve “Rolls-Royce Phantom” from numerous images of different vehicle models, as Fig. 1 shown. There are two difficulties of fine-grained vehicle recognition. One is the subtle granularity of images. Another is the huge differences in morphology, posture, color and background in fine-grained images. With the continuous improvement of the living standard of modern society and the rapid growth of vehicles' number, traffic supervision is facing great challenges. As an important point of traffic supervision, video surveillance system has been widely used in various fields of modern traffic [1], [2]. However, the traditional way of relying on manual interpretation has been unable to meet the needs of mass traffic video processing nowadays. Building intelligent recognition system to automatically process all kinds of traffic video information is an inevitable trend. The identification of vehicle types in traffic video images, as one of the key technologies in the construction, has been widely noted by researchers at home and abroad for a long time. Fig. 1.

Fine-grained vehicle image.