基于机器视觉的兰花分级方法研究

打开文本图片集

中图分类号：S566.1；TP391.4 文献标识码：A 文章编号：2095-5553（2025）10-0107-06

Abstract：Aiming at the problem of thelowdegreeof mechanization in poted orchid grading，agrading method fororchid basedon machine vision is designed.Anorchid visual inspection room is built andorchid imagesare obtained by using overhead and sidecameras.Normalized template matching algorithm，flood filing algorithm，Laplace algorithmand MER algorithmareused forplant height measurement，MERalgorithm isused forcrown width measurement，MERalgorithm andfine segmentation algorithm are used for uniformity measurement，and HSValgorithm is used for flowercoverage measurement.To test the performanceof the algorithm，thevisual recognition algorithm measurement acuracy tests and grading judgment tests were conducted．Theexperiment showed that therewasa high correlation between the measurement results of visual recognition algorithms and manual measurement results in a linear regression. The R2 values of plant height，crown width，uniformity，and flower coverage were0.9974，O.997，0.980 2，andO.9821， respectively，allapproaching1.O.Theoverallrecognitionrateof thealgorithmbasedonmanual grading resultsisabout 88.3% ，but manual grading is highly subjective，and the visual algorithm's grading measurement accuracy is higher when usingquantitative indicators for orchid grading.

Keywords：orchid grading；machine vision；normalized template matching；quadrant method；afine segmentation

0 引言

我国是花卉生产和消费大国，2024年花卉种植总面积达 1389.9khm2 ，年产值达5200多亿元1]。（剩余7772字）