基于多振动测量传感器融合的猕猴桃软腐病果识别方法研究

Soft rot is one of the main diseases of kiwifruit after harvest. The disease caused large flesh rot in kiwifruit, thus affecting the fruit quality and the competitiveness our products in the international market. Therefore, accurate identification of kiwifruit with soft rot is important for postharvest storage, grading of kiwifruit, and the improvement of product competitiveness. The acoustic vibration method is a nondestructive way for fruit quality evaluation based on the vibration characteristics of fruit, and can be used for the evaluation of fruit organization structure. However, a single vibration measurement sensor cannot provide good accuracy and repeatability when used in the fast on-line detection for fruit quality. Therefore, the study intends to establish a testing platform for fast and accurate measurement of vibration characteristics of kiwifruit based on multiple vibration measurement sensors fusion. Then, the factors affecting vibration characteristics of kiwifruits will be investigated, among which the influence mechanism of soft rot will be the focus of the study. In addition, the processing algorithm for multiple sensors fusion will be studied. Finally, the study aims to establish a fast and nondestructive method for identification of kiwifruit with soft rot, and provide the theoretical basis for the research and development of detection equipment and detection line with independent intellectual property rights.

软腐病是猕猴桃采后最主要的病害之一，软腐病的发生导致果实大面积溃烂，降低其商品品质并严重影响我国猕猴桃在国际市场中的竞争力。因此，猕猴桃软腐病果的准确识别对于果实采后保存、品质分级以及提高产品竞争力等都具有重要意义。声学振动检测技术是一种利用水果振动特性进行品质无损分析的方法，可用于水果的组织结构评估。针对声学振动检测技术中单一振动测量传感器进行水果品质在线快速检测时准确度不高、可重复性难以保证等问题，本研究拟采用多振动测量传感器融合的方式，通过传感器的选择与布置方式优化，构建猕猴桃振动特性检测平台，实现猕猴桃振动特性的快速、准确采集；探索猕猴桃振动特性的影响因素，重点研究不同软腐程度对猕猴桃振动特性的影响机制；研究多振动测量传感器融合的数据处理算法，最终建立猕猴桃软腐病果的快速无损识别方法，为研发具有自主知识产权的猕猴桃软腐病无损检测仪器和检测线提供理论依据。

软腐病导致猕猴桃果实大面积溃烂从而降低其商品品质。本研究采用声学振动无损检测方法研究猕猴桃贮藏过程中品质变化规律，并构建自主设计的猕猴桃振动特性检测平台用于猕猴桃软腐病果的识别研究。. 首先，本研究采用多振动测量传感器融合的方式，通过传感器的选择与布置方式优化，构建了3种猕猴桃振动特性检测平台，实现猕猴桃振动特性的快速、准确采集。其次，研究了猕猴桃在不同贮藏温度下品质指标与振动特性变化规律，结果表明：零级动力学模型结合Ball模型可用于模拟猕猴桃质量损失率、硬度、可溶性固形物、维生素C、可滴定酸、色差和感官等品质的变化，误差在6.5%以内；采用一级动力学模型结合Arrhenius模型对不同温度下弹性系数（EI）变化进行拟合，建立了猕猴桃货架期的无损评估模型，EI的模型拟合值与实测值误差在12%以内，表明建立的声学无损检测方法结合动力学模型可用于猕猴桃品质评估。最后，基于自主设计的滚落式猕猴桃振动特性检测平台进行猕猴桃软腐病果判别研究，研究了软腐病对猕猴桃振动特性的影响，采用逐步多元线性回归（SMLR），BP神经网络（BPNN）和主成分—BP神经网络（PCA-BPNN）建立3种猕猴桃软腐病果判别模型。结果表明：两路信号融合的判别结果优于单一路信号的判别结果；PCA-BPNN较优，软腐病果判别率达到86.2%。. 本研究为猕猴桃软腐病检测提供了一种新方法，同时为研发具有自主知识产权的猕猴桃软腐病无损检测仪器和检测线提供理论依据。