Study Of Nondestructive Test Of Apple Internal Quality By Computer Tomography Technology

In this paper we use the Computed TomograpHy(CT), for the purpose of the apple quality nondestructive, this research include the sugar content, titratable acidity and moisture content. The two main contents as follows:The first part,on the base of our laboratory’s research, we found the differences among the prediction models established by the CT numbers of different varieties apples, and their diversity all change as the time. So we use the apple include the Fushi planted in Shandong and Shanxi, the other apple is Xinjiang Tangxi.First of all, a good correlation of each prediction model was confirmed though the multiple-variety apple experiment, With the R" generally above0.8. In order to analyze and quantify the difference between each model, we first conducted a cross prediction between the CT numbers and the linear models of every two apples, and then mixed the data of the two apples to obtain a unified model to predict the content by sampling. Error analysis was also taken in both steps above.The results showed that the difference of the prediction model of sugar decreased as the storage time increased, the error rate of cross-prediction lowered from10%to less than5%, and the R2of the unified model mounted up from about0.5to more than0.8; the models of titrable acidity had great difference continuously, the error rate of cross-prediction amounted to more than50%, and the R2of the unified model were all below2.5%during the whole storage time; the difference in the model of moisture content were small, reaching2%generally, and the R2of the unified model were all greater than0.75.The second part, we attention the whole apple’s quality, not only the part of the apple. In this research include three steps:First, unified the window/level number of CT image as430/-210, then build the linear model of CT numbers and gray level values; After that, an adaptive threshold method named Otsu was used to segment the BMP gray image, so extract the pulp area, then calculate the weighted mean of pixels in this area and to CT average value; Finally the model of the relationship between CT numbers and the internal quality in the area was developed. It can be conclude that the sugar content, titrable acidity and moisture content all showed good linear relationships with average CT number and the R2reach0.8464,0.8233,0.9075respectively, and the prediction error can be controlled within5.0%,7.4%,3.8%.