| CPC G06T 7/11 (2017.01) [A01B 39/18 (2013.01); A01B 69/008 (2013.01); G06T 7/521 (2017.01); G06V 10/764 (2022.01); G06V 20/188 (2022.01); G06V 20/70 (2022.01); G06T 2207/20081 (2013.01); G06T 2207/20084 (2013.01)] | 8 Claims |

|
1. A method for planning a weeding path for a weeding robot, comprising:
acquiring an image segmentation model based on neural network model training, wherein the image segmentation model is configured to identify and segment a weed target feature, a soil target feature and a crop target feature; and
based on the weed target feature, acquiring the weed target feature by intra-camera tracking, and acquiring the weed target feature by inter-camera tracking, and planning a weeding path for a weeding robotic arm of the weeding robot, to enable the weeding robotic arm to weed according to the weeding path,
wherein a first camera, . . . , an i-th camera, . . . , an N-th camera are arranged sequentially at intervals in an opposite direction of a heading direction of the weeding robot, and the first camera to the N-th camera move with a movement of the weeding robot;
wherein based on the weed target feature, acquiring the weed target feature by intra-camera tracking, and acquiring the weed target feature by inter-camera tracking, and planning a weeding path for a weeding robotic arm of the weeding robot, comprises:
acquiring a first image captured by the first camera, identifying and segmenting the first image by the image segmentation model, to generate a first weed target feature and a weed label of the first weed target feature;
when the i-th camera is a second camera, acquiring an i-th image captured by the i-th camera, extracting a target feature in the i-th image according to a density-based spatial clustering method, and matching the target feature in the i-th image with an (i−1)-th weed target feature, to obtain an i-th weed target feature corresponding to the (i−1)-th weed target feature and a weed label of the i-th weed target feature; according to the i-th weed target feature and the weed label of the i-th weed target feature, planning the weeding path for the weeding robotic arm of the weeding robot;
when the i-th camera is a third camera to the N-th camera in sequence, extracting target feature in the i-th image according to the density-based spatial clustering method, matching the target feature in the i-th image with an (i−1)-th weed target feature to obtain an i-th weed target feature corresponding to the (i−1)-th weed target feature, and a weed label of the i-th weed target feature; according to the i-th weed target feature and the weed label of the i-th weed target feature, correcting the weeding path for the weeding robotic arm of the weeding robot;
wherein, N≥2, 1<i≤N, i and N are both positive integers; time for the N-th camera to move with the weeding robot to an original position of the first camera and a sum of time for image processing of the first camera to the N-th camera are greater than or equal to time for the weeding robotic arm of the weeding robot to swing to a position of a weed corresponding to the first weed target feature.
|