Embryo feature extraction and dynamic recognition method for maize haploid seeds

Haploid breeding is an efficient new way in breeding, but the natural generating possibility for maize haploid seeds is too low. However, haploid seeds are generally sorted by hand, which reduces the sorting efficiency. A sorting way of haploid that is labour-saving, timesaving and with high accuracy should be found. Nowadays, machine vision technology has been more improved and widely applied to the processing, identification, and classification of agricultural products, which makes it possible to realize the automatic sorting for maize haploid seeds based on machine vision. Maize seeds with genetic marker gene within a same variety are mainly classified into haploid and hybrid according to the distribution of Navajo genetic markers on different parts of each seed, especially the embryo. So the characteristics of maize embryo are important to distinguish between haploid and hybrid seeds. In this paper, the embryo feature extraction and dynamic recognition method for maize seeds with genetic markers was studied. An image segmentation method to extract the characteristics of embryo referring to the average pixel values of B channel was proposed. And the embryo images of the maize seeds with Navajo genetic markers were extracted. The Navajo genetic markers in the acquired embryo region were extracted in RGB color space, and the number of Navajo pixels for each seed was counted to judge which classification it belonged to. A rapid recognition algorithm of maize haploid seeds was obtained after embryo feature extraction and Navajo marker extraction. The dynamic sorting test for LC9124-UH400 maize seeds was performed on the maize haploid sorting platform, and the result of the test showed that the recognition rate for haploid seeds was 98.04%, and that for hybrid seeds was 94.44%. The main factors that affected the recognition result in the test were analyzed. A dynamic sorting system for maize haploid seeds including the algorithm and the maize haploid sorting platform has been built up, which has a helpful significance for the realization of maize haploid automatic sorting.