Semantic segmentation is used to recognize a collection of pixels that form different categories. The purpose of semantic segmentation is classifying a certain image class and separating it from other image classes by covering it with a segmentation mask. Semantic segmentation results can be obtained using different methods. Two popular segmentation methods used in extracting leaves from complicated backgrounds are: the watershed method and the cut graph method. The watershed method is to place water sources in each regional minimum and then flood the relief. The graph cut algorithm is widely applied to computational problems that can be formulated in terms of energy minimization. Effective segmentation of plant markers is one of the main requirements for precision agriculture. A significant amount of research has been devoted to the classification of plant species through the examination of leaf images. The video image segmentation system for fruit trees in an orchard in natural conditions is mostly used in agriculture and industry. While semantic segmentation has been used for various robotic applications such as robot manipulation, autonomous navigation, and camera localization, its use in segmenting natural structures, such as trees, into their components is still an open area for research. The measurements in this paper, that is, the processing of the collected photos, was based on the diversity of the canopy and the way of photographing depending on the proximity, appearance and density of the canopy. The final results were obtained using the graph cut method.