论文部分内容阅读
目的建立针对淡水鱼整鱼鱼体新鲜度的快速无损检测方法.方法通过比较不同的光谱与相应挥发性盐基氮(TVB-N)值的建模结果,以及对比分析竞争性自适应重加权算法(CARS)、遗传算法(GA)及连续投影算法(SPA)三种特征波长选择方法对模型的优化结果,对鱼鳞及光谱采集部位等影响因素进行研究。结果鱼体有鳞时的尾部为最佳新鲜度检测部位。CARS法较优且鱼体新鲜度检测的最优波段为800～1100nm,采用CARS特征波长选择方法选择出23个波长变量重新建立PLS模型,模型预测集相关系数RP=0.957,预测均方根误差RMSEP=0.589mg/100g。利用CARS方法选择的23个波长变量对淡水鱼进行新鲜度评价,准确率达96.67%。结论该方法为淡水鱼整鱼新鲜度快速无损检测提供了一种有效的方法。
OBJECTIVE To establish a rapid nondestructive detection method for the freshness of freshwater fish whole body fish.Methods By comparing the results of different spectra with the corresponding modeling results of TVB-N values and comparative analysis of competitive adaptive weighting (CARS), genetic algorithm (GA) and continuous projection algorithm (SPA) were used to optimize the model, and the influencing factors of fish scales and spectral collection sites were studied. Results tail when the scales of fish for the best freshness detection site. The optimal CARS method was used to detect the freshness of fish and the optimal wavelength band was 800 ~ 1100nm. The PLS model was established by selecting 23 wavelength variables using CARS wavelength selection method. The correlation coefficient of model prediction set was RP = 0.957. The root mean square error of prediction RMSEP = 0.589 mg / 100 g. The freshness of freshwater fish was evaluated using 23 wavelength variables selected by CARS method, with an accuracy rate of 96.67%. Conclusion This method provides an effective method for rapid and non-destructive testing of whole fresh freshwater fish.