Prediction of Ear Weight, Kernel Weight and Viability in Maize Using Image Analysis

Abstract

In maize breeding studies, it is becoming common to determine the ear and kernel characteristics by image analysis. While current methods focus on measurements that can be obtained directly by image analysis, it has not been adequately addressed whether different parameters such as weight and viability can be estimated using these measurements. This study aimed to determine whether it is possible to estimate the ear weight (g), kernel weight (g), single kernel weight (g) and viability (1/0) status of maize with the help of features (area, perimeter, width, length) extracted from images of the ear and kernel samples. In this study, 233 ear and 1242 grain samples belonging to 13 maize genotypes were used as material. Digital images of the ear samples were taken with a 5 MP camera and from the kernel samples with a desktop scanner. The ear weight reference data (DV1) and the kernel weight reference data (DV2) were obtained by weighing each sample on a precision balance. Single kernel reference data (DV3) was obtained with the measurements of single kernel weights. Kernel samples underwent paper germination test and reference data (DV4) related to viability was created. Regression models were developed by using the features obtained from image analysis (area, perimeter, width, height) for each reference data set as the predictor variable. As a result of the study, it was seen that the ear weight and kernel weight can be estimated with the help of the parameters extracted from the image analysis. While moderate success was achieved in the determination of single seed weight, it was difficult to determine the viability status based on the morphometric measurements of a single kernel in maize.