Use of certified seeds, increases the quality and quantity of yield. Typically, for the certification, the analysis and classification process is made by experts using visual characteristics of the seeds. These conventional methods are very time consuming, very tedious, costly, and depend on the person. In the determination of the seed properties process, first, identification of the seed types, varieties and identification of diseased and structural deformed seeds operations are performed. Wheat is one of the world's most important grains, as well as being the main source of food is an important industrial raw material. There are many works in the literature about determination of varieties of wheat seeds. Also, it is very important to confirm the variety of the wheat before planting. Because each variety needs its own condition for taking good yield. Previous work of our research group was based on classification of seed images by using image analysis in the computer environment. Classification of images has been a keep-up- to date method for recent years. Same like images that have close textures and colors are to be classified using some scientific approaches via image processing techniques. Depending on our previous work, same approach that run on MATLAB is to be realized in an embedded system which is independent from the computer. This Linux running embedded card stores data to be classified in an external memory. Classification algorithm is constructed by using Linux based image processing libraries. Rest of the paper is going to introduce scientific methods to classify images together with literature review. After, algorithmic details and some results are going to be presented related to system performance metrics and accuracy rates of algorithm based on computer-free embedded card.