State of the Art

Literature Review

Since meat is closely related to people’s daily diet, much research has been done on meat quality assessment. The evaluated characteristics include color, texture, pH, tenderness, etc. The implementation of computer vision techniques in the food industry can be dated back to the 1990s. The followings are related researches we found.

Beef-related imaging techniques

It has become increasingly popular because it is objective, non-destructive, and has high reliability and efficiency [1].

Computer vision segmentation

In [2],[3], the image processing algorithm was used to evaluate the intramuscular fat, also called marbling, in beef. Color parameters and variations were intensively investigated.

Support vector machine segmentation

In [4], the support vector machine was applied to classify the grades of beef-fat color.

Fuzzy ARTMAP segmentation

In [5], Fuzzy ARTMAP and ANNs were used so that the color of beef can reflect the bacterial colonies formed in the sample.

The lighting condition is too ideal.

However, most of the researches mentioned above are from the perspective of meat manufacturers. Most studies use pictures of beef taken on the production line or taken in a standard environment, eliminating the influence of different lighting conditions. We plan to acquire steak pictures from online sources for our project, which look more similar to what we see in grocery stores.