Estimation of the quantity of jelly in a tomato fruit based on the external appearance information
2003
Kawano, T. (Kochi Univ., Nankoku (Japan). Faculty of Agriculture)
Abstract Methods for estimating jelly quantity in tomato fruit by the external appearance have been developed and its estimating accuracy was confirmed. This method consists of two major processes, first the process of identifying the number of locules in a tomato fruit and then the process of estimating its jelly quantity. The former process was conducted by analyzing a fruit surface image with a novel neural network and the latter by the integration of the jelly area distribution along the major axis of the fruit. The neural network had two input neurons, six hidden neurons and one output neuron, and a tangent sigmoid-type function was adopted as the input-output transfer function in all neurons. It took hue and saturation information from the fruit surface image and generated a pixel image of the locule partitioning lines. After that, the image was passed through a locule counting process, identifying the number of locules in the fruit with 97% of accuracy. Jelly area distributions along the major axis of the fruit were obtained through multiplication of the section area distribution by the approximating function of the jelly occupation ratio distribution to the section area. As the result of estimation by this method, jelly quantity of tomato fruits was estimated with a relative error of only 18. 4% on average. This method was sufficiently accurate for con-sumers at retail markets to select tomato fruits according to their characteristic taste, and therefore it suggests the possibility of development of non-destructive tomato fruits selectors based only on the external appearance.
显示更多 [+] 显示较少 [-]