Visual measurement method of crop height based on color feature in harvesting robot

Abstract

Focusing on the problem of inefficiency and labor waste in contact and sensing crop altimetric methods, a method is proposed for crop height measurement using machine vision. Firstly, a crops height measurement model was established based on aperture image principle, and then segmentation lines extraction of crops and background was implemented by a series of morphological operations. Secondly, the foreground information is segmented according to the H channel in the HSV color space. To help extract the complete crop area, a global scan was performed on the image using multiple images to obtain a segmentation threshold with a certain fault tolerance, and the inter-frame sum method and large-scale filter template are also adopted for image enhancement. Considering of the uneven height lines and porous holes in the crop area, the boundary line between the crop region and the background is strengthened through morphological operations such as dilation and erosion. Finally, the Sobel operator is applied to detect the horizontal line. The parameters of the height line could be calculated, and thus achieving crop height measurement. Experiment results show that the inter-frame enhanced image is filtered using the median and morphology of the large-scale window, and the complete upper boundary line of rice and wheat is obtained through hole filling to realize height measurement. Average error of crop height is less than 1.6%, and processing time per frame within 50 ms.