智慧农业导刊2025，Vol.5Issue(20)：13-16,4.DOI:10.20028/j.zhnydk.2025.20.004
荔枝劣变的多模态传感感知与智能识别
摘要
Abstract
In response to the deterioration problems such as browning of the peel and softening of the flesh during postharvest storage and transportation of Lychee,a multi-modal sensing and detection system integrating machine vision,gas sensor array and self-made pressure testing device has been built.By extracting visual features such as color,shape,and texture,detecting changes in volatile compound components,and measuring textural and mechanical parameters to obtain multi-dimensional characteristic data of Lychee deterioration.Establish a CNN-LSTM fusion network model to realize multimodal data fusion and intelligent recognition.Experimental results showed that Lychee began to deteriorate significantly on the third day of storage.The peel brightness value L* dropped from 72.5 to 45.2,and the hardness dropped from 856 N to 312 N.The accuracy rate of the constructed multi-modal fusion identification model reaches 96.8%,which is 12%～18%higher than that of a single sensing mode,providing an effective technical solution for rapid and non-destructive testing of Lychee quality.关键词
荔枝劣变/多模态传感/CNN-LSTM网络/特征融合/智能识别Key words
Lychee deterioration/multimodal sensing/CNN-LSTM network/feature fusion/intelligent recognition分类
园艺学与植物营养学引用本文复制引用
朱启明,罗剑斌,杨图信,郭少英,滕丽丽,潘朝勃..荔枝劣变的多模态传感感知与智能识别[J].智慧农业导刊,2025,5(20):13-16,4.基金项目
广东省普通高校青年创新人才类项目(2023KQNCX265) （2023KQNCX265）
茂名市科技计划项目(2023S011) （2023S011）
