【摘要】 针对汽车VIN码的识别与应用,研发出一款具备VIN码识别与信息解析可视化展示能力的手机APP软件系统。该系统实现了一种基于逻辑回归神经网络模型的识别算法,并具有一种可扩展的系统框架,可以很方便地集成适应性更强的识别模型。实验表明:该APP软件系统面向标准字符图像集识别率达到95%以上,但受拍摄质量和VIN码铭牌污损等影响,实际识别率在90%左右;与模板匹配的方法相比,该算法对于VIN码字体和轻微背景污损具有更好的适应性。更多还原

【Abstract】 For recognition and application of Vehicle Identification Number(VIN),this paper develops a mobile phone APP software system with the ability of VIN code recognition and information analysis and visualization display.The system realizes a recognition algorithm based on logistic regression neural network model,and proposes a scalable system framework,which can easily integrate more adaptable recognition models.Experiments show that the recognition rate of the APP software system for standard character image set can be achived more than 95%,and the recognition rate decreases to about 90%due to the quality of shooting and the stain of VIN code nameplate.Compared with the method of template matching,it has better adaptability to variousVIN code fonts and slight background fouling.更多还原