基于图像分析的丹贝发酵终点判定方法

Determination of Fermentation End-point of Tempeh by Image Analysis

丹贝是一种发源于印尼的发酵食品，其品质主要取决于表面菌丝的生长情况。为了快捷、可靠、无损的自动化判定丹贝的发酵终点，提出了基于图像分析的判定方法。在丹贝发酵过程中使用数码相机定时获取其表面图像，通过分析图像的纹理特征，将图像的梯度信息与亮度信息相结合，制定了新的丹贝表面图像的均匀性评价指标。结果表明：该指标在丹贝发酵终点（由人工感官判定为22h）取得最小值，且在发酵终点附近时段内（21～23h）变化明显，对于像素为2592×3314的丹贝表面图像，分别为0.054（21h）、0.044（22h）、0.055（23h）；而传统的丹贝表面图像的均匀性评价指标（亮度标准差）则无显著变化，分别为8.3（21h）、8.2（22h）、8.2（23h）。提示该方法对于多种像素的丹贝表面图像均能快速、准确的判定出丹贝的发酵终点，具有良好的稳定性且与人工的感官评价一致。

Tempeh is a fermented food originated in Indonesia. Its nutritional quality depends on the extent of mycelial growth on its surface. A method based on image analysis was proposed to determine the fermentation end-point automatically, quickly, reliably and non-destructively. Tempeh＇s surface images were collected by using a digital camera during its fermentation process and a new uniformity evaluation of tempeh’s surface image was formulated through analysis of texture features and combination of gradient and luminance information. The indicator reached the minimum value at the fermentation end-point （22 h, determined by sensory evaluation） and showed a significant change near the fermentation endpoint. When the pixel of tempeh’s surface image was 2592 × 3314, the indicator＇s value was 0.054 at 21 h, 0.044 at 22 h and 0.055 at 23 h. However, no significant change was observed in the traditional uniformity evaluation of tempeh’s surface image （standard deviation of luminance） in which the values were 8.3 at 21 h, 8.2 at 22 h and 8.2 at 23 h. The experimental results showed that the method could quickly and accurately determine tempeh’s fermentation end-point based on surface images with different pixels, had a good stability and was consistent with artificial sensory evaluation.