ABSTRACT

This chapter discusses the feasibility of using transfer learning to identify the embryo surface and variety of maize seeds. The image processing technology has been widely used in the inspection and grading of agricultural products such as cereals, vegetables and fruits based on their color, shape, defects, and many other external appearances. It is of great significance in sowing process of agriculture to judge the position and germination of maize embryo. The chapter identifies the seed surface, embryo surface, non-embryonal surface and longitudinal section of different maize varieties and compares the accuracy of identification in all cases. It is very important to select the best identification scheme for practical application. Transfer learning has been successfully applied to image classification, human disease detection, text emotion classification, software defect classification, multilingual text classification, plant disease identification and human activity classification. The transfer learning method greatly reduces the workload of image acquisition, enabling the small data sets to achieve higher recognition accuracy.