基于图像识别技术的南疆红枣品种识别研究

通过Matlab图像处理和识别技术,根据不同品种的红枣的形状特征不同,分别对5种不同品种的红枣进行识别。首先对红枣俯视图像预处理提取红枣的表面轮廓,然后利用轮廓计算矩形度,圆形度,偏心率等7个几何特征和8个图像的不变距。利用PNN和BP神经网络作为分类器,对不同品种的红枣图像进行识别。结果表明,两种神经网络能够对不同品种红枣进行识别,PNN网络的平均识别率为90%,BP网络的平均识别率为80%,PNN神经网络比BP神经网络分类效果好。

Recognition of Southern Jujube Varieties Based on Image Recognition Technology

Through the computer image processing and recognition technology, according to the different shape features of different varieties of red jujube, respectively for five different varieties of red jujube for identification. Firstly, the overlooking image preprocessing to extract the surface contours of red jujube, and then the contour is used to calculate rectangular, circular, eccentricity and seven geometric characteristics, such as eight images of the same distance. Using PNN and BP as a classifier to different varieties of red jujube image recognition. Results show that both neural network can carry on the classification of different varieties of red jujube, the PNN network average recognition rate is 90%, BPN network average recognition rate is 80%. PNN neural network is better than BP neural network classification effect.