一种基于机器视觉的飞机钣金件跨粒度识别方法

Cross-grained recognition method for aircraft sheet metal parts based on machine vision

飞机钣金件成形后需要进行喷漆处理,大量不同种类钣金件共同批量喷漆后的分类识别一直以来依靠人工完成,是一项繁琐而困难的工作。提出了一种基于图像的飞机钣金件跨粒度识别方法来应对这个难题,设计了一套专门的图像采集平台,可以对钣金件正面及侧面两个角度进行拍摄,样本图像以及由形状因子和不变矩构成的10维特征向量数据一起保存至样本数据库中,以备识别时使用。根据钣金件种类繁多、高相似度钣金件成组存在的特点,设计了粗粒度的识别方法。通过遍历比较钣金件的10维特征向量,从数据库中找出与被识别钣金件具有最高相似度的两个候选钣金件,然后进一步通过人机结合的细粒度识别方法实现对钣金件的最终识别。在针对20种不同类型飞机钣金件的实验中,该方法达到了96.0%的识别精度,且整个识别过程简便高效。

Aircraft sheet metal parts need to be painted after forming. The classification and recognition of a large number of different kinds of sheet metal parts after batch painting has been performed manually, which is a tedious and difficult work. An image-based cross-grained recognition method for aircraft sheet metal parts is proposed to deal with this tough problem. A specific image collection platform is designed and constructed to take the images of the sheet metal parts in both top and side angles. The images of the sample parts, together with the extracted ten-dimensional(10 D) feature vectors composed of shape factors and invariant moments, are stored in a database for the later use of recognition. According to the features of numerous kinds of sheet metal parts and the existence of groups of sheet metal parts with high similarity, the coarse-grained recognition method is designed. Through traversing and comparing the 10 D feature vectors, the 2 candidate targets of the given sheet metal part to be recognized with top similarity are found from the database. Then, the fine-grained recognition method, which is a man-machine cooperation process, is used to achieve the finally recognition of the expected sheet metal part. In the experiments on 20 different kinds of aircraft sheet metal parts, the proposed method achieves a recognition accuracy of 96.0%, and the whole recognition procedure is simple, convenient and efficient.