Semolina Speck Counting Using an Automated Imaging System 1

Cereal Chem. 73(5):561-566 An objective instrumental method for counting specks in semolina has were used to test the influence of speck size and darkness on the count been developed. The method involves making pellets from semolina and levels. The inclusion of smaller specks in the count method increased the detecting specks using a computerized image analysis (IA) system. The count levels, while excluding lighter specks reduced the speck count. The IA method for speck counting has been compared to visual counting of IA method has proven reliable over a wide range of samples representing the pellets by three experienced technicians and to the visual counting of diverse genotypes and environments. It can be set for specific speck size the original semolina using the standard Grain Research Laboratory and darkness, allowing calibration to relate directly to results by tradi(GRL) procedure. For a linear range of speck counts created by blending tional visual counting. The semolina pellets may be stored for a long semolina samples, a linear response was obtained by the IA method. The period, providing a permanent record of each sample, and allowing for robustness of the developed IA method was tested using semolina milled standardization of speck counts over time. from a diverse range of plant breeders samples. Eight of these samples The milling of semolina from durum wheat aims to fully separate the starchy endosperm from other components of the wheat kernel. In common with all milling processes, this is an imperfect process due to the complexity in structure of the wheat kernel itself, such that the semolina (starchy endosperm) product is contaminated by bran fragments. In addition, incomplete removal of contaminants from the grain during cleaning can ultimately provide a further source of impurities in the milled product. Bran fragments and ground impurities are visible as specks in semolina. The presence of specks has a negative impact on the value of the semolina because they cause brown or dark flecks in pasta, reducing consumer acceptability of the product. A count of the number of visible specks is usually conducted as part of the quality control process during semolina production, and often is a primary specification that the miller must meet. The exact nature of the counting process varies, but typically a semolina sample is spread on a table, the semolina surface is flattened, and a grid of known dimension is placed on top (Vasiljevic and Banasik 1980, Dexter and Matsuo 1982, El Bouziri and Posner 1988). The number of specks present is visually counted and expressed as the total number within a defined area. Visual identification of specks is subjective; since it is based upon observer experience, it is influenced by factors such as fatigue and subject to observer bias in determining speck size and darkness for inclusion in the count. Visual counting is also tedious. Pouring the semolina sample gives an inconsistent and only temporary surface, providing no permanent record of the sample. It is difficult to generate consistent results, due to these constraints. The limitations of visual speck counting make the development of a rapid objective instrumental procedure desirable for both laboratory and commercial applications. Imaging methods for detecting bran fragments in common wheat flours were reported using white light illumination (Evers 1993, Whitworth 1994) and commercialized using fluorescence methods (Harrigan 1995). The uneven surface of semolina, due to its coarser granulation, limits the use of common wheat flour imaging methods to count specks in semolina. The objective of the experiments reported here was to