Semantic segmentation for partially occluded apple trees based on deep learning

Optical flow-based branch segmentation for complex orchard environments

“…Amatya and Karkee [5] train a Bayesian classifier to segment and reconstruct branches on images of sweet cherry trees taken at night, while [6] uses an encoderdecoder network to segment out branches, wires, and fruit in a kiwifruit orchard. Depth information is a common modality for filtering out unwanted background noise: [7], [8] use depth information to filter out all points beyond a specified threshold before feeding the RGB image through a neural network, while [9], [10] feed the depth channel directly into the neural network to filter out background noise. Yang et al [11] produce a mask with a raw RGB image but postprocess the mask with depth data to create an accurate tree model for localization.…”

“…We utilize the pix2pix [17] framework, a Generative Adversarial Network (GAN) [18] designed to convert images in one domain into another one, to perform the segmentation. Our previous work [1], as well as others [10], have shown that the pix2pix network is capable of performing robust segmentation.…”

“…Trained using the manually labeled RGBD data described in Section V-A. This setup is comparable to the one described in [10]. To further examine the effects of the use of depth and flow data, we train three other networks for an ablation study:…”

HOB-CNN: Hallucination of occluded branches with a convolutional neural network for 2D fruit trees

“…Fruit orchard tasks, such as thinning, pruning, and harvesting, require an accurate tree skeleton model. However, fruit trees are often occluded by artifacts in the environment, such as foliage and fruits, which makes it a challenge for 1 Washington State University https://research.libraries.wsu.edu/xmlui/handle/2376/17719 current computer vision technology (Chen et al, 2021). Furthermore, branch detection currently lacks a publicly available annotated datasets, unlike the more studied tasks of fruit detection, weed detection, and disease detection.…”

“…Branch detection is commonly done using semantic segmentation Zhang et al, 2018;Majeed et al, 2020c;Chen et al, 2021;Liang et al, 2020;Granland et al, 2022). Majeed et al (2020c) used SegNet (Badrinarayanan et al, 2017) and Liang et al (2020) used U-Net (Ronneberger et al, 2015) to segment the visible region of both branches and trunks of a tree.…”