加工条件下酱用番茄关键成分漫透射光谱图像特性及检测机理研究

To tackle the detection problem of key quality components in processing tomato caused by soaking under the process conditions, we proposes a testing strategy based on spectral diffuse transmittance imaging. The modern testing methods were used to study th varying pattern of the key quality components of processing tomato of different varieties, maturity stages, temperatures and immersion time ranges.The acquired spectral images in different positions of processing tomato will be processed using segmentation method to extract regions of interest, for the sake of obtaining the characteristic information of spectral and spatial texture, then the information was preprocessed by means of integration, normalization and baseline correction.Partial least squares, neural network and genetic algorithm method will be used to establish the quantitative predictive models that would be optimized by analying the predictive effect for the key quality components of processing tomato.Multiple linear regression, interval partial least squares and interval partial least squares will be used as wavelength selection algorithm to screen the optimal wavebands for the detection of quality indices.On these basis, the effect of light intensity, exposure time, lighting angle parameters, relative position of light sources,the optical aperture and the object distance on the effects of processing tomato diffuse transmission image characteristic and the detection effect of processing tomato for different varieties, maturity stages, temperatures and immersion time ranges will be studied by orthogonal test. Finally,the detection mechanism for the key quality components of processing tomato will be revealed. All these achievements would lay theoretical basis for the development of online detection technology and equipment deployment to capture diffuse transmittance multispectral images for detecting of the key quality components of processing tomato under the process conditions.

针对酱用番茄关键成分加工环节因浸泡所引起的检测难题，提出基于番茄漫透射光谱图像特性的检测思路。.利用现代测试手段对不同品种、成熟度、水温和浸泡时间的酱用番茄关键品质参数变化规律进行研究；利用背景分割方法对采集的多姿态下番茄漫透射光谱图像目标区域进行精确提取，获取光谱和空间纹理信息；然后对信息进行整合、归一化和基线校正等预处理；分析偏最小二乘、神经网络、遗传算法等方法对各关键成分的建模效果，优选建模算法，并利用逐步多元线性回归、间隔偏最小二乘和连续投影等算法筛选有效检测波段；通过正交试验，研究光源功率、曝光时间、打光角度、光源相对位置、通光孔面积和物距等参数对不同品种、成熟期、水温和浸泡时间的酱用番茄漫透射光谱图像特性及检测效果的影响关系，最终探明水浸酱用番茄关键成分漫透射光谱成像检测机理，为进一步研发加工条件下酱用番茄关键成分漫透射多光谱成像在线检测技术和装备提供理论依据。

本研究对不同成熟度、水温和浸泡时间的酱用番茄关键品质参数变化规律进行了分析，并搭建了高光谱漫透射成像系统，研究了水浸酱用番茄的漫透射光谱图像特性，建立了酱用番茄SSC含量的预测模型。主要研究成果如下：.（1）利用破坏性测试方法对不同浸泡、温度的酱用番茄SSC含量进行了测试分析，得出酱用番茄浸泡时间和浸泡水温对酱用番茄SSC含量均存在显著性的影响关系。并且酱用番茄在相同浸泡时间下，随着温度增加SSC含量降低；在相同浸泡水温下，随着浸泡时间的增加，SSC含量降低。.（2）测试分析了不同成熟度酱用番茄硬度、可溶性固形物含量和水分含量。得出半熟期酱用番茄硬度最大，在1.8~3.5 MPa之间，SSC含量平均值为3.9 °Brix。成熟期硬度其次，在1.7~2.8 MPa之间。完熟期硬度最低，在1.1~2.5 MPa之间。成熟期和完熟期的酱用番茄SSC含量平均值分别为4.8 °Brix和5.1 °Brix，两者相差不大。酱用番茄的含水率在半熟期，成熟期和完熟期没有明显的变化，均为94%左右。.（3）测试分析了酱用番茄的基本物理参数（横纵径和质量），发现酱用番茄横径在40~60 mm之间，纵径在50~75 mm之间。酱用番茄的质量在50~125 g之间，平均质量为85 g。 .（4）搭建了高光谱漫透射成像系统，研究了不同打光角度下的光谱图像特性。得出随着卤素灯与水平面之间夹角的减小，酱用番茄光谱图像的亮度逐渐减小，平均光谱曲线数值减小。.（5）证明了利用可见-近红外高光谱漫透射成像技术测量酱用番茄内部可溶性固形物含量的可能性。利用CARS算法选取了47个有效波长，分别建立了不同的多参数补偿线性PLS模型和非线性LS-SVM模型，并进行了系统地比较分析。结果表明，由于加工番茄的尺寸和重量等特征参数与SSC之间呈非线性关系，因此对构建的线性PLS模型的预测能力无明显提高。对比所有的PLS模型和LS-SVM模型，基于47个有效波长的多参数（光谱、面积和重量）补偿CARS-Area-Weight-LS-SVM模型获得了最优的预测结果，其预测相关系数rpre、预测均方根误差RMSEP和剩余预测偏差RPD分别为0.94、0.17和2.94，该模型对加工番茄内部可溶性固形物含量SSC具有良好的预测能力。

内容获取失败，请点击重试