Research On Nondestructive Detection Of Fruit And Vegetable Quality During Drying

Fruit and vegetable are necessary for maintaining the normal physiological functions of human body and keeping our body in good health, it’s also indispensable in our daily life and the second world’s most important agricultural products right after grain. However, there are problems in the seasonal production and concentrative ripening period of fruit and vegetable,which are easy to arise overproduction in time and space and cause the phenomenon such as the prices drop dramatically and rot loss. The fruit and vegetable drying should be a feasible method to solve the aforesaid problems. Moisture content, moisture content uniformity and color are three important indexes of dried fruit and vegetable product quality, which affect the purchase desire of consumers, the product shelf-life and the storage stability. Meanwhile, they are also the right kind of important indicator which reflects whether the drying process route and choosed parameters are appropriate. Obtaining those indexes quickly is beneficial to improve process route and optimize process parameters, thereby improve economic benefit.Aiming to implement the nondestructive detection for moisture content, moisture content uniformity and color, hyperspectral imaging technology was used for fruit and vegetable during drying in this paper. This study includes below three aspects.1. The quality detection method during drying based on the multiple model fusion was studied to improve the detection precision of color and moisture content for the dried soybeans. Hyperspectral re?ectance images were acquired from fresh and dried soybeans over the spectral region between 400 and 1000 nm. Firstly, mean, entropy, relative divergence and standard deviation feature were extracted; then PLS sub-models were built by using these four features, respectively; finally, the final prediction was obtained by the weighted fusion of the prediction sub-models. The results showed that the multiple model fusion method achieved consistently better results comparing with PLS sub-models.2. Moisture content uniformity was one of critical parameters to evaluate dried product quality and drying technique. Two methods, using prediction value of moisture content to calculate the content uniformity uniformity(indirect) and using hyperspectral imaging technique predicting the moisture content uniformity directly, were investigated. The test results showed that using direct prediction method of the moisture content uniformity had significantly better result than the indirect method.3. The hyperspectral image preprocessing method based on orthogonal signal correction was studied and applied to the nondestructive detecting of moisture content uniformity. The mean feature and standard deviation feature were extracted; then orthogonal signal correction(OSC) was applied to pre-process the mean feature and standard deviation feature, finally the prediction model was built combining with PLS. The results showed that on the premise of guarantying the accuracy of model prediction, the redundant spectral information could be deleted and the effective spectral information can be retained after that the mean and standard deviation are pre-processed with OSC, and the model could also be simplified and improved.