基于机器视觉小麦叶片含水量检测研究

发布时间：2018-02-11 11:25

本文关键词： 机器视觉叶片含水量特征提取 Matlab　出处：《山东农业大学》2017年硕士论文　论文类型：学位论文

【摘要】：随着机器视觉技术的蓬勃发展,国内外专家更倾向于应用机器视觉技术进行信息的诊断、研究。本文利用机器视觉技术,结合当地小麦墒情,对小麦叶片的含水量进行预测研究。结果表明,该方法用于对叶片水分进行预测研究是切实可靠的,可以为后续的研究提供重要依据。本文采用机器视觉技术将在田间采集到的100幅小麦叶片样本中与含水量相关性较大的叶片特征进行检测提取出来,实现了提取特征的无损化和模块处理的快速化,用BP神经网络建立含水量评判模型,通过该模型就可以实现叶片含水量的预测。为了提高该预测系统的精度,首先对图像的预处理、图像的分割以及图像的特征提取等关键算法的选择进行理论性的研究。在对图像进行预处理的过程中,图像增强部分选用了中值滤波算法,然后直接增强图像与背景的对比度,达到目标边缘清晰同时目标与背景的对比度明显的效果。在图像分割部分,由于图像目标与背景间的灰度级差别比较明显,所以选用了Ostu最大类间方差法的图像二值化运算。然后运用Matlab编写程序对图像进行特征提取,本研究分别提取了叶片图像的颜色、纹理及外形算法作为评价的综合指标,减少单个参数对判定的影响,提高综合判定的精度。最后建立BP神经网络,对该网络进行训练和输出仿真,并对该网络进行预测验证,预测结果的精度已经达到了96%,达到预期的预测目标。这一结果表明,基于机器视觉小麦叶片含水量的检测是可行的,并且可以应用到叶片含水量的实际预测中。
[Abstract]:With the rapid development of machine vision technology, experts at home and abroad tend to use machine vision technology to diagnose and study information. The prediction of water content in wheat leaves is studied. The results show that the method is effective and reliable for the prediction of water content in wheat leaves. In this paper, 100 wheat leaf samples collected in the field were extracted by machine vision technology, which had a high correlation with water content. In order to improve the accuracy of the prediction system, the model of water content evaluation can be established by BP neural network, and the prediction of leaf water content can be realized by the model. Firstly, the selection of key algorithms, such as image preprocessing, image segmentation and image feature extraction, is theoretically studied. In the process of image preprocessing, the median filter algorithm is used in the image enhancement part. Then the contrast between image and background is enhanced directly to achieve the effect that the edge of the target is clear and the contrast between the target and the background is obvious. In the image segmentation part, because the gray level difference between the image object and the background is obvious, Therefore, the binarization operation of the maximum inter-class variance method of Ostu is selected, and then the feature extraction of the image is carried out by using the Matlab program. In this study, the color, texture and shape algorithm of the leaf image are extracted as the comprehensive evaluation indexes, respectively. The influence of single parameter on the decision is reduced, and the accuracy of comprehensive judgment is improved. Finally, BP neural network is established, the network is trained and simulated, and the network is forecasted and verified. The accuracy of the predicted results has reached 96%, which indicates that the water content detection of wheat leaves based on machine vision is feasible and can be applied to the actual prediction of leaf water content.
【学位授予单位】：山东农业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41;S512.1

【参考文献】