Affiliations: [a] College of Mechanical and Electronic Engineering, Sichuan Agricultural University, Ya’an, China
| [b] College of Manufacturing Science and Engineering, Sichuan University, Chengdu, China
Correspondence:
[*]
Corresponding author. Z.L. Kang, College of Mechanical and Electronic Engineering, Sichuan Agricultural
University, Ya’an 625014, China. Tel./Fax: +86 0835 2882270; E-mail: [email protected].
Abstract: The intelligent identification and positioning for melons and other fruits are important links for the melon and fruit picking robot to naturalize smooth picking, thus directly affecting the picking efficiency and success rate. Considering that the picking robot has low identification rate and low positioning accuracy for melons and fruits, this paper designed an intelligent watermelon identification and positioning method in a natural scene. This method included the following steps: first, the natural watermelon images shot by the left and right cameras were captured to increase the proportion of the watermelon region in the image; second, erosion, adaptive noise cancellation and filling, as well as other techniques were used for the captured watermelon image to identify the watermelon region and calculate its values of barycentric coordinates; third, the squint binocular positioning algorithm was designed based on the values of watermelon barycentric coordinates in two images to obtain the actual watermelon three-dimensional space coordinates, using the left camera as the origin of coordinates. The experiment verified that the relative positioning errors of this method were within±15% for the watermelon three-dimensional space coordinates in a natural scene, thereby providing a key intelligent identification and positioning method for the watermelon picking robot.
