To solve the low accuracy of image feature matching in horticultural robot visual navigation, an innovative and effective image feature matching algorithm was proposed combining the improved Oriented FAST and Rotated BRIEF (ORB) and Lucas–Kanade (LK) optical flow algorithm. First, image feature points were extracted according to the adaptive threshold calculated using the Michelson contrast. Then, the extracted feature points were uniformed by the quadtree structure, which can reduce the calculated amount of feature matching, and the uniform ORB feature points were roughly matched to estimate the position of the feature points in the matched image using the improved LK optical flow. Finally, the Hamming distance between rough matching points was calculated for precise matching. Feature extraction and matching experiments were performed in four typical scenes: normal light, low light, high texture, and low texture. Compared with the traditional algorithm, the uniformity and accuracy of the feature points extracted by the proposed algorithm were enhanced by 0.22 and 50.47%, respectively. Meanwhile, the results revealed that the matching accuracy of the proposed algorithm increased by 14.59%, whereas the matching time and total time decreased by 39.18% and 44.79%, respectively. The proposed algorithm shows great potential for application in the visual simultaneous localization and mapping (V-SLAM) of horticultural robots to achieve higher accuracy of real-time positioning and map construction.

To solve the problems of the low target-positioning accuracy and weak algorithm robustness of target-dosing robots in greenhouse environments, an image segmentation method for cucumber seedlings based on a genetic algorithm was proposed. Firstly, images of cucumber seedlings in the greenhouse were collected under different light conditions, and grayscale histograms were used to evaluate the quality of target and background sample images. Secondly, the genetic algorithm was used to determine the optimal coefficient of the graying operator to further expand the difference between the grayscale of the target and background in the grayscale images. Then, the Otsu algorithm was used to perform the fast threshold segmentation of grayscale images to obtain a binary image after coarse segmentation. Finally, morphological processing and noise reduction methods based on area threshold were used to remove the holes and noise from the image, and a binary image with good segmentation was obtained. The proposed method was used to segment 60 sample images, and the experimental results show that under different lighting conditions, the average F1 score of the obtained binary images was over 94.4%, while the average false positive rate remained at about 1.1%, and the image segmentation showed strong robustness. This method can provide new approaches for the accurate identification and positioning of targets as performed by target-dosing robots in a greenhouse environment.