基于大数据分析的植物标本取样方法研究

Research on Sampling Methods of Plants Based on Big Data Analysis

植物识别是利用植物的前提。随着图像处理与识别技术的迅速发展,植物种类识别App应运而生。虽然这些App能够识别植物,但存在一些误差,导致识别结果不够准确。为了解产生误差的可能原因,本文以花伴侣作为识别App,通过不同的取样方法,探讨植物标本取样因素对识别率的影响。测试结果表明,基于大数据分析的标本取样,以全株局部含花的样本识别效果最佳,不同植物器官的样本照片在识别上存在显著差异,不同拍摄设备、不同背景条件下拍摄的植物样本照片识别效果也有一定差异。

Plant identification is the precondition for plant application.As the rapid development of image processing and image identification, identification Apps for plants appeared. Though plant species can be identified by these Apps,it still has some errors.These errors account for inaccurate results. In order to find the possible causes of these errors,this paper using I-Iua Banlv as the chosen identification App,the effect of sampling factors on the identification ratio was discussed through different sampling methods. Based on the big data analysis of specimen sampling ,test results indicated that the samples using the whole plant with flowers had the highest identification ratio.Meanwhile ,various specimen photos of different plant organs differentiated from each other.There were also some differences in the recognition ratio of the plant sample photos taken under different shooting equipment and different background colors.