An automatic detection method for maize tassel traits

Abstract

The invention discloses an automatic detection method for corn tassel traits. The method firstly performs target detection on the collected field corn bottom view images, generates tassel candidate frames to obtain tassel potential areas, and then uses multi-view image features And the Fisher vector encoding method is used to describe the characteristics of the tassel and detect the target, so as to confirm the area to which the tassel belongs. At the same time, on the basis of the detection results, the semantic segmentation is used to further complete the segmentation of the fine shape of the tassel, and finally the image feature is established. The mapping relationship between the length trait, the width trait, the girth trait, the diameter trait, the ear color trait, the branch number trait and the total ear number trait, which has seven physical meanings. The method of the invention can continuously monitor the growth state of corn tassels in real time, has high accuracy of detection results, and has important significance for research on reproductive growth of corn, research on corn genetics and genetics, and yield estimation.

Description

technical field [0001] The invention belongs to the cross field of computer vision and agricultural meteorological observation, and more specifically relates to an automatic detection method for corn tassel traits, that is, the traits of corn tassels in the image are obtained by taking the field corn bottom-view image sequence as the object feature method. Background technique [0002] As one of the three major food crops in the world, corn is widely planted all over my country. Analysis and research on various traits of corn can help to establish the relationship between its traits and yield to obtain greater benefits, and can also provide a basis for the study of corn genetics and genetics. However, for a long time, the detection of various traits of maize has been mainly through manual detection, which is easily affected by subjective factors and inefficient; at the same time, the observation process will inevitably cause damage to the growth environment of maize; and be...

AI Technical Summary

Problems solved by technology

[0003] In 2011, Tang Wenbing and others published the paper "Recognition and Positioning of Maize Tassels Based on Binocular Stereo Vision" in the "Proceedings of the 2011 Academic Annual Conference of the Chinese Society of Agricultural Engineering" using image segmentation technology to locate and detect corn tassels Relevant research has been carried out, but the method in this paper only considers color features, which is suitable for small-scale detection under specific lighting and simple backgrounds. It is not robust to natural lighting and field environment detection, and this method does not obtain The refined traits of spikes; adopted in the paper "An image-based approach for automatic detecting tasseling stage of maize using spatio-temporal saliency" published by Mengni Ye et al. on "Eighth International Symposium on Multispectral Image Processing and Pattern Recognition" in 2013 The target detection framework of HOG-SVM is added and the clues of time and space are added to detect corn tassels, and then the arrival time of corn earing stage is predicted according to the detection results. The shortcoming of the paper is that only the gradient feature information of the gray image is considered. The adaptation to the complex background is not good, and the tassel is not segmented, and the tassel is obtained for a more refined description; 2014 Ferhat et al. introduced a more robust corn tassel detection method in the paper "Detecting corn tassels using computer vision and support vector machines" published on "Expert Systems with Applications". Compared with the previous paper, it considered more robust Rod shape and texture features, but it is still essentially color-based image segmentation, which is not suitable for sequence images with large illumination changes, and does not map image detection results to biological traits with physical meaning; in summary, Although there are currently many detection technologies related to corn tassels, due to the limitations of various methods or strategies, it is difficult to apply them in the actual field environment, and these methods have not established the relationship between image features and actual physical meanings. Biomass mapping relationship

Images