目的 现有的车标识别方法尽管取得了不错的识别效果,但最终的识别率容易遇到瓶颈,很难得到提升。车标识别是智能交通系统中至关重要的一部分,识别率的微小提升也能带来巨大的社会价值。为了挖掘与分析车标识别中潜在的问题和难点,通过对实验中未能正确识别的车标图像的观察,发现大部分模糊车标图像未能得到正确分类。针对车标图像中存在的成像模糊等情况,提出一种基于抗模糊特征提取的车标识别方法。方法 该方法首先构建车标图像金字塔模型,然后分别提取图像的抗纹理模糊特征和抗边缘模糊特征。其中,抗纹理模糊特征的提取使用局部量化的LPQ模式,增强了原始特征的鲁棒性,抗边缘模糊特征的提取使用基于局部块弱梯度消除的HOG特征提取方法,在描述车标图像边缘梯度信息的同时,提升特征的抗模糊能力。最后利用CCA方法进行两种抗模糊特征的融合,用于后续的降维与分类。结果 在构建的模糊车标数据集(BVL)、公开车标数据集HFUT-VL和XMU上进行实验,均上取得了很好的识别效果,其识别率高于其他一些车标识别算法,且具有很强的鲁棒性和抗模糊性。结论 基于抗模糊特征提取的车标识别方法,通过抗模糊特征的提取和融合,以及车标图像金字塔模型的构建,提高了方法对成像质量欠缺的车标图像的识别能力,从而提升了整体识别效果,更符合实际应用中车标识别的需求。
Vehicle logo recognition based on anti-blur feature extraction
He Minxue,Yu Ye,Xu Jingtao,Lu Qiang(School of Computer and Information, Hefei University of Technology)
Objective Although the existing vehicle logo recognition (VLR) method has achieved good recognition results, the final recognition rate is limited and it is difficult to improve. VLR is a vital part of the intelligent transportation system. Even a small increase in recognition rate can bring great social value. In order to discover the potential problems and difficulties in VLR, we analyze the incorrectly identified samples, and find out that most of the blurred vehicle logo images are not correctly classified. In order to recognize the blurred vehicle logo images, a VLR method based on anti-blur feature extraction is proposed. Method Our method first constructs the car image pyramid model, and then extracts the image"s anti-texture and anti-edge blur features. The localized LPQ mode is used for anti-texture blurred feature extraction, which can enhance the robustness of the original features. The HOG feature extraction based on local block weak gradient elimination method is used for anti-edge blurred feature extraction, which can well describe the edge feature of vehicle logos, and at the same time, improve their anti-blur ability. Finally, the CCA method is used to fuse the two anti-blur features for subsequent dimensionality reduction and classification. Result Experiments are carried out based on the blurred vehicle logo dataset (BVL) constructed in this paper and two other open vehicle logo dataset HFUT-VL and XMU. The results show that our method can achieve very good recognition results. It can achieve higher recognition rate compared to some other VLR methods, and has strong robustness and anti-fuzziness. Conclusion Based on the anti-blur feature extraction method, the recognition and fusion of anti-blur features and the construction of the car image pyramid model, the recognition ability of the vehicle logo images with insufficient imaging quality is improved, thus improving the final recognition rate, which is proved to be more suitable for the recognition of vehicle logos in practical applications.