Vehicle Color Recognition With Spatial Pyramid Deep Learning

View Researcher's Other Codes

Disclaimer: The provided code links for this paper are external links. Science Nest has no responsibility for the accuracy, legality or content of these links. Also, by downloading this code(s), you agree to comply with the terms of use as set out by the author(s) of the code(s).

Please contact us in case of a broken link from here

Authors Chuanping Hu, X. Bai, Li Qi, Pan Chen, Gengjian Xue, Lin Mei
Journal/Conference Name I
Paper Category
Paper Abstract This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination. IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS 1 Vehicle Color Recognition With Spatial Pyramid Deep Learning Chuanping Hu, Xiang Bai, Senior Member, IEEE, Li Qi, Pan Chen, Gengjian Xue, and Lin Mei Abstract—Color, as a notable and stable attribute of vehicles, can serve as a useful and reliable cue in a variety of applica- tions in intelligent transportation systems. Therefore, vehicle color recognition in natural scenes has become an important research topic in this area. In this paper, we propose a deep-learning-based algorithm for automatic vehicle color recognition. Different from conventional methods, which usually adopt manually designed features, the proposed algorithm is able to adaptively learn rep- resentation that is more effective for the task of vehicle color recognition, which leads to higher recognition accuracy and avoids preprocessing. Moreover, we combine the widely used spatial pyramid strategy with the original convolutional neural network architecture, which further boosts the recognition accuracy. To the best of our knowledge, this is the first work that employs deep learning in the context of vehicle color recognition. The experiments demonstrate that the proposed approach achieves superior performance over conventional methods. Index Terms—Color recognition, deep learning, convolu- tional neural network (CNN), spatial pyramid (SP), intelligent transportation. I. INTRODUCTION L ICENSE plate [1], [2] has been one of the core research objects in the area of intelligent transportation systems for a long period of time. However, license plates on vehicles are not always fully visible (due to partial occlusion or viewpoint change) and not easy to recognize under certain situations (due to noise, blur, or corruption). In contrast, paints on vehicles oc- cupy a much larger portion of vehicle bodies and are relatively insensitive to interference factors such as partial occlusion, viewpoint change, noise, and corruption. Hence, vehicle color has been used as a valuable cue in a wide range of applications, such as video monitoring [3], criminal detection [4], and law enforcement [5]. This advantage makes automatic vehicle color recognition an important research topic in the field of intelligent transportation systems [6]. Manuscript received January 14, 2015; revised April 10, 2015; accepted April 24, 2015. This work was supported in part by the National 863 Project under Grant 2013AA014601, by the National Science and Technology Major Project under Grant 2013ZX01033002-003, and by the National Natural Sci- ence Foundation of China under Grant 61222308. The Associate Editor for this paper was J. Zhang. (Corresponding author Li Qi.) C. Hu, L. Qi, G. Xue, and L. Mei are with the Third Research Insti- tute of the Ministry of Public Security, Shanghai 200031, China (e-mail quick.qi@foxmail.com). X. Bai and P. Chen are with the School of Electronic Information and Communications, Huazhong University of Science and Technology (HUST), Wuhan 430074, China. Color versions of one or more of the figures in this paper are available online at http//ieeexplore.ieee.org. Digital Object Identifier 10.1109/TITS.2015.2430892 However, identifying vehicle color in uncontrolled environ- ments is a challenging task. The difficulties mainly stem from two aspects 1) Certain color types are very close to other color types and, thus, are very hard to discriminate. For example, cyan is not so distinguishable from green in real-world images. 2) The color of a vehicle is prone to be affected by numerous interference factors, such as haze, snow, rain, and illumination variation. To tackle these challenges, a number of works have been proposed [7]–[10]. With hand-crafted f eatures (e.g., color sift [8], normalized RGB histogram [9], and feature context [10]), these methods obtain excellent performances but are far from producing all-satisfactory results, particularly in complex real- world scenarios. Furthermore, these approaches usually rely on preprocessing techniques [10], to alleviate the impact of interference factors such as haze and strong illumination. In this paper, we propose a deep-learning-based algorithm for vehicle color recognition. Compared with traditional ap- proaches, the proposed algorithm possesses three advantages. 1) We learn features from training data in an automatic manner, instead of adopting manually designed features. The learned features are more effective for vehicle color recognition and robust to variations in real-world sce- narios. Moreover, the proposed algorithm directly runs on raw pixels and requires no preprocessing techniques, which are essential for the success of previous methods. 2) We introduce spatial information into the vehicle color recognition algorithm, by combining the spatial pyramid (SP) strategy [11] with the framework of deep learning [12]. The usage of spatial information further improves recognition accuracy. 3) Experiments on a standard benchmark demonstrate that the proposed algorithm outperforms other competing methods. As shown in Fig. 1, the convolutional neural network (CNN) architecture proposed in [12] is adopted as the feature extractor, which computes a feature vector for each image, and support vector machine (SVM) [13] is employed as the classifier, which predicts the color of a given vehicle. Before training, the vehicle images are rescaled to a fixed size. In the training procedure, the parameters of the network are iteratively updated by backpropa- gation, using the training images and their labels (black, white, red, green, etc.). Once trained, the parameters are stored, and the CNN architecture becomes a feature extractor, which takes images as input and produces features maps (or vectors) in each layer. 1524-9050 © 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http//www.ieee.org/publications_standards/publications/rights/index.html for more information., A new technique for classifying all 56 different characters of the Manipuri Meetei-Mayek (MMM) is proposed herein. The characters are grouped under five categories, which are Eeyek Eepee (original alphabets), Lom Eeyek (additional letters), Cheising Eeyek (digits), Lonsum Eeyek (letters with short endings), and Cheitap Eeyek (vowel signs. Two related works proposed by previous researchers are ... [Show full abstract], Deep leaning has become a hot research topic with the rapid development of big data technology. As an important branch of deep learning, convolutional neural network has been widely used in image recognition, and has achieved great success. Convolutional architecture for fast feature embedding (Caffe) with features like speed, extendibility and openness is currently top popular tool of deep ... [Show full abstract], In this thesis we focus on visual servoing (VS) techniques, critical for many robotic vision applications and we focus mainly on direct VS.In order to improve the state-of-the-art of direct methods, we tackle several components of traditional VS control laws. We first propose a method to consider histograms as a new visual servoing feature. It allows the definition of efficient control laws by ... [Show full abstract], The determination of plant species from field observation requires substantial botanical expertise, which puts it beyond the reach of most nature enthusiasts. Traditional plant species identification is almost impossible for the general public and challenging even for professionals that deal with botanical problems daily, such as, conservationists, farmers, foresters, and landscape architects. ... [Show full abstract], In this paper, a novel classification technique for large data set of mammograms using a deep learning method is proposed. The proposed model targets a three-class classification study (normal, malignant, and benign cases). In our model we have presented two methods, namely, convolutional neural network-discrete wavelet (CNN-DW) and convolutional neural network-curvelet transform (CNN-CT).An ... [Show full abstract]
Date of publication 2015
Code Programming Language Jupyter Notebook
Comment

Copyright Researcher 2022