基于卷积神经网络的车型识别方法及装置

Vehicle type recognition method and device based on convolutional neural network

Abstract

本发明提供了基于卷积神经网络的车型识别方法,该方法包括:选取已标签车型的样本图像,对卷积神经网络进行训练,获取训练好的车型识别模型;采集或者输入彩色场景视频图像;采用加权平均法对每帧彩色场景图像进行灰度化处理,获取每帧灰度场景图像;采用均值漂移背景更新背景法更新背景图像,计算每帧灰度场景图像与背景图像的差分图像,对差分图像进行中值滤波、连通区域标记、连通区域筛选处理,获取每帧灰度场景图像的感兴趣区域图像;采用训练好的车型识别模型对每帧灰度场景图像的感兴趣区域图像进行分类识别,输出识别结果。与现有技术相比,本发明能快速地对场景图像中的车型进行识别,且识别准确率高。
The invention provides a vehicle type recognition method based on a convolutional neural network. The method includes the steps: selecting a sample image of a labeled vehicle type and training the convolutional neural network to obtain a trained vehicle type recognition model; acquiring or inputting color scene video images; performing gray-scale treatment on each color scene image by a weighted average method to obtain each gray scene image; updating a background image by a mean shift background updating method, calculating a differential image of each gray scene image and the background image, filtering a mid-value of the differential image, marking and screening a communication area and acquiring an area-of-interest image of each gray scene image; classifying and recognizing the area-of-interest image of each gray scene image by the trained vehicle type recognition model and outputting recognition results. Compared with the prior art, the vehicle type recognition method has the advantages that vehicle types in the scene images can be rapidly recognized, and recognition accuracy is high.

Claims

Description

Topics

Download Full PDF Version (Non-Commercial Use)

Patent Citations (0)

    Publication numberPublication dateAssigneeTitle

NO-Patent Citations (0)

    Title

Cited By (0)

    Publication numberPublication dateAssigneeTitle