论文部分内容阅读
为了实现苹果园的快速精确测产,结合可见光与热红外图像,提出了一种基于机器学习和Hough变换的苹果树测产新方法。以成熟期苹果树为研究对象,利用热成像相机同步采集可见光与热红外图像数据,通过仿射变换模型实现了可见光与热红外温度图像的配准;利用温度信息与RGB颜色波段作为4个分类特征,采用支持向量机,完成分类与后验概率的计算;采用Hough变换实现了图像中苹果的识别标注和计数;通过线性回归模型进行了苹果测产估计,并进行了交叉验证。在光照条件非均一而使苹果颜色存在差异的情况下,经过试验验证,与人工记录的测产数据相比,该文提出的新方法苹果测产的准确率达到80%以上,为果园的科学管理提供了有力的技术支持。
In order to achieve fast and accurate measurement of apple orchards, a new method of apple tree test based on machine learning and Hough transform is proposed based on visible light and thermal infrared images. Taking the ripe apple tree as the research object, the thermal imaging camera was used to collect visible and thermal infrared images synchronously. The registration of visible and thermal infrared images was achieved by the affine transformation model. The temperature information and the RGB color bands were used as the four categories Feature, using support vector machine to complete the classification and posterior probability calculation; Hough transform to achieve the identification of the apple in the image of the label and count; measured by the linear regression model of the apple yield estimates, and cross-validation. Under the condition of non-uniformity of lighting conditions and apple color difference, the experimental verification shows that the accuracy of the new method proposed by this paper is over 80% Management provides a strong technical support.