Abstract:

Ammoniacal odor, hydrogen sulphide odor, hue, saturation and brightness in the collected image, the number of adipocytes and bacterial plaque of pork reflecting its freshness were measured. A backpropagation (BP) neural network structure, in which TVB-N value was the output and the above 7 non-coherent parameters composed the input was established. Based on this, a SOM network structure, in which TVB-N value was the input, was established for the cluster analysis of TVB-N values collected during pork spoilage and 5 TVB-N value based pork freshness rates were obtained. The rating results were in accordance with the national standard and those from sensory evaluation. This demonstrates high reliability of this rating method.

Key words: plaque area, digital image processing, neural network, freshness rating