一种基于LC-BoVW模型的建筑垃圾物料识别算法

An Algorithm of Construction Waste Material Recognition Based on LC-BoVWModel

针对建筑垃圾物料的种类多、形貌易混淆等问题,构建了一种基于局部约束的视觉词袋(local constraint-bag of visual words,LC-BoVW)模型的建筑垃圾物料识别算法。首先,对建筑垃圾物料图像分块,分别提取局部颜色特征和局部二值模式特征;考虑到图像分块特征的局部相似特性,构建LC-BoVW模型分别对目标图像的显著特征进行统计。然后,基于信息融合思想对特征统计量进行融合,形成图像的判别性特征并输入到分类器中进行物料的精确识别。最后,利用自建的5类建筑垃圾物料图像数据集进行实验,实验结果表明,所提算法能够快速有效地实现建筑垃圾物料识别,平均识别准确率可达到97.92%。

To solve the problems of many types of construction waste materials and easy confusion of appearance,an algorithm of construction waste material recognition based on local constraint-bag of visual words(LC-BoVW)model is constructed.Firstly,the image of construction waste material is divided into blocks,and local color features and local binary pattern features are extracted respectively.Considering the local similarity of image block features,a LC-BoVW model is constructed to carry out statistics on the significant features of the target image respectively.Then,based on the idea of information fusion,the feature statistics are fused to form the discriminant features of the image and input into the classifier for accurate material recognition.Finally,the experiment is carried out by using the self-built image dataset of 5 types of construction waste materials.The experimental results show that the proposed algorithm can quickly and effectively realize the recognition of construction waste materials,and the average recognition accuracy can reach 97.92%.