Browsing by Author "Ponce-Donoso, Mauricio"

Now showing 1 - 2 of 2
  • Thumbnail Image
    Item
    Challenges for computer vision as a tool for screening urban trees through street-view images
    (2024) Arevalo-Ramirez, Tito; Alfaro, Anali; Figueroa, Jose; Ponce-Donoso, Mauricio; Saavedra, Jose M.; Recabarren, Matias; Delpiano, Jose
    Urban forests play a fundamental and irreplaceable role within cities through the ecosystem services they provide, such as carbon capture. However, inadequate management of urban trees can heighten the risks they pose to society. For instance, mechanical failures of tree components, such as branches, can cause harm to individuals and property. Regular assessments of tree conditions are necessary to mitigate these tree-related hazards, yet such evaluations are labor-intensive and currently lack automation. Previous studies have proposed utilizing street view images to alleviate tree inspection and shown the feasibility of visually inspecting trees. However, only a limited number of studies have addressed the automatic evaluation of urban trees, a challenge that can potentially be addressed using deep learning networks. Particularly in urban environments, there is a pressing need for increased automation in unresolved computer vision tasks. Therefore, this research presents a comprehensive analysis of neural networks and publicly available datasets that can aid arborists in automatically identifying urban trees. Specifically, we investigate the potential of deep learning networks in classifying tree genera and segmenting individual trees and their trunks. We emphasize the utilization of transfer learning strategies to enhance tree identification. The results demonstrate that neural networks can be considered practical tools for assisting arborists in tree recognition. Nevertheless, there are still gaps that remain and require attention in future research endeavors.
  • Thumbnail Image
    Item
    Exploring the Potential of Reconstructed Multispectral Images for Urban Tree Segmentation in Street View Images
    (2024) Arévalo Ramírez, Tito; Alfaro, Analí; Saavedra, José M.; Recabarren, Matías; Ponce-Donoso, Mauricio; Delpiano, José
    Deep learning has gained popularity in recent years for reconstructing hyperspectral and multispectral images, offering cost-effective solutions and promising results. Research on hyperspectral image reconstruction feeds deep learning models with images at specific wavelengths and outputs images in other spectral bands. Although encouraging results of previous works, it should be determined to what extent the reconstructed information can lead to an advantage over the captured images. In this context, the present work inspects whether or not reconstructed spectral images add relevant information to segmentation networks for improving urban tree identification. Specifically, we generate red-edge (ReD) and near-infrared (NIR) images from RGB images using a conditional Generative Adversarial Network (cGAN). The training and validation are carried out with 5770 multispectral images obtained after a custom data augmentation process using an urban hyperspectral dataset. The testing outcomes reveal that ReD and NIR can be generated with an average structural similarity index measure of 0.93 and 0.88, respectively. Next, the cGAN generates ReD and NIR information of two RGB-based urban tree datasets (i.e., Jekyll, 3949 samples, and Arbocensus, 317 samples). Subsequently, DeepLabV3 and SegFormer segmentation networks are trained, validated, and tested using RGB, RGB+ReD, and RGB+NIR images from Jekyll and Arbocensus datasets. The experiments show that reconstructed multispectral images might not add information to segmentation networks that enhance their performance. Specifically, the p-values from a T-test show no significant difference between the performance of segmentation networks.