chromoCorn: Zea mays Chromosome Classification Using Zero Order Fuzzy Inference System Based on Kernel’s Haralick Grey Level Texture Phenes

The selection of haploid kernels is one of the critical stages in the doubled haploid method. Various vision-based methods have been developed through visual color contrasts that is limited to human eye. In sowing and germination, thousands of kernels are being separated manually which is time-consuming and prone to errors. Hence, development of automatic sorting systems is essential. Currently, detection of colored genetic markers on kernel’s embryo, called R1-nj are being developed. However, color attributes of R1-nj in some embryo kernel images brought by different genetic backgrounds, morphological, and environmental dynamics became unreliable. Also, hyperspectral imaging is too expensive for low-income farmers who are also the intended end users of this technology. This premise laid the groundwork for the current study’s goal of developing chromoCorn model which is a reliable and sensitive model in classifying the chromosome of maize whether it is haploid (1 complete set of chromosomes) or diploid (2 complete sets of chromosomes) using Sugeno fuzzy inference system (FIS). The extracted textural features from RGB turned grayscale images, namely energy, contrast, homogeneity, and correlation, serves as the four inputs in the fuzzy inference system. The Gauss method was used for the fuzzification process, and the weighted average approach is used for defuzzification. The 120 set of IF-THEN statements formed the basis of the fuzzy model to obtain the defuzzied output. Based on training and testing evaluation, the detection system’s accuracy in identifying haploid and diploid kernels is 96.04% and 91.23%, respectively.