摘要 针对报废汽车企业对不同型号汽车换挡面板自动分类的需求,提出了一种基于机器视觉的面板自动分类方法。该方法首先使用Blob分析对图像进行预处理,分割出图像中的换挡面板目标对象,接着提取被分割图像的11个形状特征作为特征向量,利用特征向量训练多层感知器(Multi-Layer Perceptron,MLP),最后使用完成训练的MLP神经网络模型对不同类型的汽车换挡面板进行分类预测。实验结果表明,该方法的分类准确率达到99%,每幅图像的分类时间小于100 ms,验证了该方法的有效性与高效性,能够作为报废汽车换挡面板自动分类的一种手段。 For the needs from scrapped car companies to automatically classify the car gear shifting panels of different models,an automated classification method of car gear shifting panel based on machine vision was proposed.The Blob analysis was used for preprocessing,when image was segmented to locate the target of gear shifting panel.Then shape feature extraction was carried out to find 11 shape feature as characteristic vector,which was the input of trained MLP model.Finally,by using complete trained MLP model,different types of car gear shifting panel were classified.Experimental results showed that the classification accuracy of this method is 99%,and the classification time of each image is less than 100 ms,which verifies the feasibility and efficiency of this method,and can be used as an effective approach of automatic classification of shifting panel of scrapped cars.
机构地区 浙江科技学院信息与电子工程学院
出处 《现代制造工程》 CSCD 北大核心 2021年第4期104-108,共5页 Modern Manufacturing Engineering
基金 浙江省基础公益研究计划项目(LGG21F010003,LGN19B050002) 浙江省科技特派员项目 2017年度教育部人文社会科学研究一般(规划基金)项目(17YJA880004)。
关键词 机器视觉 汽车换挡面板 自动分类 形状特征 多层感知器 machine vision car gear shifting panel automatic classification shape characteristics Multi-Layer Perceptron(MLP)
