基于三维生物散斑技术的牛肉质构特性预测

Prediction of Beef Texture Based on Three-Dimensional Biospeckle

三维成像技术能够获得样品的空间信息,具有快速、方便、可实时的特点。为了缩短生物散斑技术的检测时间,减小其应用的局限性,将三维成像技术引入到传统生物散斑技术中,以期得到更好的预测效果。分别从两个不同角度拍摄同一样品的图像信息,利用广义差分法对图像进行预处理,并运用灰度模板匹配法、小波变换法和对比度调制融合法对两个角度的图像进行匹配融合,以时间序列散斑图灰度共生矩阵的参数对比度表示图像的散斑活性,建立其对牛肉质构特性的预测模型。通过传统相机标定法,获得相机的内外参数,并利用相似三角形原理提取图像的深度信息,对因物体摆放位置不同引起的误差进行校正,使获得的结果更加准确。结果表明,相似三角形原理可对样品的深度信息进行校正。三维生物散斑技术能更好地对牛肉的硬度和咀嚼性进行预测,对3种图像融合方法进行比较可知,小波变换法的预测效果最好,对硬度和咀嚼性的预测相关系数分别可达到0.944 4和0.928 8。

Three-dimensional imaging technology is a low-cost, portable and online screening tool for optical non-destructive testing, which allows the obtainment of the space information of samples. Three-dimensional biospeckle was applied in the prediction of beef texture to minimize the detection time and reduce the limitations of its application. The images of samples were acquired at two different angles, pre-proccessed by generalized difference (GD), and fused by gray template matching,wavelet transform, and contrast modulation fusion separately. A temporal history of speckle patterns viz. inertia moment was used as the speckle activity to establish a prediction model for beef texture. The traditional camera calibration method was used and the internal and external parameters of the camera were established. In-depth information from the images was extracted based on the similar triangle principle. The results showed that three-dimensional biospeckle could enable better prediction of beef texture. Among three image fusion methods, wavelet transform provided the best prediction performance,with correlation coefficients of 0.944 4 and 0.928 8 for the prediction of hardness and chewiness, respectively.