基于果肉破裂时力声同步响应锯齿化特征分析的香梨脆度评价研究及声振法无损检测

Korla pear (Pyrus bretschneideri rehd) is a native in Xinjiang. The textural quality differences are becoming large with the increasing planting area and new cultivars cultivation. In particular, the crispness of Korla pear, as an important textural index, appears to be the worst fitted sensory attribute. This problem causes adversely affects on pear reputation, grade and commodity rate. So appropriate crispness evaluation conforming to human sensory for Korla pears is being sought by fruit industry to develop on-line detecting technique. The objectives of this research are to perform a puncturing test for pear to measure instrumental acoustic-mechanical response. The jaggness of response curves is to be measured by second differentiation, fractal dimension and power spectrum analysis. Combining with sensorial test for crispness assessment, cell turgor level manipulation for different crispness and microscopic observation of pear tissue, the mechanical and acoustic parameters are to be obtained for characterization of pear crispness. Then, the comprehensive assessment index for pear crispness containing mechanical and acoustic components is to be established by principle component analysis and cluster analysis. Finally, two piezoelectric bend transducers are to be used to test the Korla pears with different level of crispness. The resonance frequency, sound propagation velocity and frequency band magnitude taken from acoustic vibration are to be conducted correlation analysis and regression analysis with crispness parameters and index. Moreover, the prediction model for pear crispness are to be established aiming to nondestructive testing textural attribute of fruit. It can provide base for the commercial-scale grading line for Korla pear. Also, it has important reference value for other fruits to study crispness evaluation method.

新疆特色香梨因产地扩大和品种增多而造成质地差异变大，果肉脆性这一最为关切的口感质地劣变问题尤为突出，导致香梨声誉受损及等级、商品率下降，因此亟待解决香梨果肉脆度更贴近口感的准确评价问题，进而实现无损检测。为此，本项目将进行香梨果肉穿刺测试，采用力声同步法获取响应曲线，应用二阶微分法、功率谱法、分形维数法对曲线锯齿化特征开展测度分析与脆度评价能力比较，并结合细胞膨压调控果肉脆度、感官评价和细胞组织显微观察的方法，找到适宜测度脆度的力声响应特征参数；然后在此研究基础上，采用主成分分析和聚类分析等方法，构建融合力声信息的香梨脆度综合评价指标；最后，采用压电梁式传感器对不同脆度香梨开展声振响应测试，提取共振频率、声速和频带幅值多响应参数，与香梨脆度评价的参数指标进行回归分析，建立脆度预测模型，实现香梨脆性质地的无损检测，为今后香梨在线质地分级检测提供研究基础，也为其它梨果脆度评价研究提供借鉴。

库尔勒香梨酥脆质地是消费者食用时触觉和听觉的综合感知，难以准确测量，因此迫切需要实现更接近口感的香梨脆度准确检测以满足香梨分级需求。为此，本项目对香梨果肉破裂同步采集力声响应信号，然后采用多种方法从力声响应锯齿化曲线中提取与感官脆度相关的特征参数，构建香梨果肉脆度预测和分类模型。在此研究基础上，采用声振法检测技术获取多域谱信息并提取与香梨果肉感官脆度相关的人工特征，经主成分分析降维后，采用多元线性回归方法构建可无损检测香梨果肉脆度的模型。本项目研究结果为香梨脆度质检测仪器和分级装置研发提供研究基础，所取得的重要结果及关键数据如下：.⑴采用峰值法、功率谱法和表观分形维数法对曲线的锯齿度进行度量，提取了29个力学和19个声学的锯齿度特征参数，有24 个力学特征参数和11个声学特征参数与感官脆度显著相关性（P<0.05），可用于香梨脆度的评价，但有207对力学特征参数和声学特征参数之间存在相关性，信息冗余较严重。.⑵在香梨脆度相关的锯齿度特征参数中选取前8 个主成分进行多元回归分析，建立了香梨果肉脆度预测检验模型，检测模型的决定系数Rc2=0.862（R2>0.8），均方根误差RMSEC为0.588，模型检测准确性和稳定性都较高，可以实现对香梨果肉脆度较为准确的仪器检测。.⑶采用力-声参数融合，基于支持向量机（SVM）和人工神经网络（ANN）两种算法分别构建的香梨果肉脆度分类模型都具有较高分类准确率，ANN 模型采用三层隐藏层每层14个节点，分类准确率达96.1%；SVM模型采用二次核函数，分类准确率达93.8%。两种模型不断学习训练，分类准确率基本接近一致，分别趋于97%和96%，均可满足对不同脆度香梨准确分类要求，也可为其他湿脆性果蔬脆度分类检测提供参考。.⑷从声振响应时域、频域和希尔伯特黄变换（HHT）的时频图像中，提取与感官脆度有显著相关的人工特征数量分别为16, 9和15个，各域特征经降维分别构建的多元线性回归模型都可较高精度检测香梨果肉脆度，检测精度从高至低依次为时域模型、时频域模型和频域模型。当三域特征融合构建的多元线性回归模型，校正集决定系数R2c=0.902，验证集决定系数R2p=0.906，与任何单一域模型相比，对香梨脆度检测精度最高。所构建的香梨脆度各检测模型，可为香梨脆度质地在线检测技术开发提供策略。

内容获取失败，请点击重试