Object-Oriented Canopy Gap Extraction from UAV Images Based on Edge Enhancement

Abstract

Efficient and accurate identification of canopy gaps is the basis of forest ecosystem research, which is of great significance to further forest monitoring and management. Among the existing studies that incorporate remote sensing to map canopy gaps, the object-oriented classification has proved successful due to its merits in overcoming the problem that the same object may have different spectra while different objects may have the same spectra. However, mountainous land cover is unusually fragmented, and the terrain is undulating. One major limitation of the traditional methods is that they cannot finely extract the complex edges of canopy gaps in mountainous areas. To address this problem, we proposed an object-oriented classification method that integrates multi-source information. Firstly, we used the Roberts operator to obtain image edge information for segmentation. Secondly, a variety of features extracted from the image objects, including spectral information, texture, and the vegetation index, were used as input for three classifiers, namely, random forest (RF), support vector machine (SVM), and k-nearest neighbor (KNN). To evaluate the performance of this method, we used confusion matrices to assess the classification accuracy of different geo-objects. Then, the classification results were screened and verified according to the area and height information. Finally, canopy gap maps of two mountainous forest areas in Yunnan Province, China, were generated. The results show that the proposed method can effectively improve the segmentation quality and classification accuracy. After adding edge information, the overall accuracy (OA) of the three classifiers in the two study areas improved to more than 90%, and the classification accuracy of canopy gaps reached a high level. The random forest classifier obtained the highest OA and Kappa coefficient, which could be used for extracting canopy gap information effectively. The research shows that the combination of the object-oriented method integrating multi-source information and the RF classifier provides an efficient and powerful method for extracting forest gaps from UAV images in mountainous areas.

Forest Gap Extraction Based on Convolutional Neural Networks and Sentinel-2 Images

“…It was confirmed that the transferability of the CNN models was better than that of the DF models. We speculate that this is because although the model construction idea of deep forest is similar to that of deep learning model with multiple levels, it is still a tree-based machine learning model in essence, its performance in transfer learning cannot be comparable to that of deep learning model, even though it was proven to perform well in the case of few-shot learning [59], and other studies have also only demonstrated the effectiveness of machine learning algorithms in a single region [18,27,32,35,36]. However, considering the results shown in Figure 11 and Table 7, the performance of DF models in a single region was comparable to that of CNN models, but the transferability was very poor, we assumed that there may be overfitting in DF models [72], undermining the accuracy of transfer learning.…”

“…Most of the forest gap extraction studies based on airborne LiDAR and high-resolution multi-spectral data have used the object-based image analysis (OBIA) approach to segment then classify forest gap [27,34,35]. These studies are premised on accurate segmentation and extraction of forest gap boundaries, and therefore rely heavily on high-quality CHM data derived from high-accuracy LiDAR data, thus, their classification accuracy of forest gaps may be reduced in the absence of high-accuracy CHM data [36].…”

“…In terms of classification algorithms, most of the current studies have used only machine learning algorithms, which have been demonstrated to be effective for forest gap classification based on both high-and medium-resolution multi-spectral data [18,27,32,35,36]. Additionally, a couple of deep learning algorithms including models such as VGG16, ResNet152V2, long short-term memory (LSTM) and 2D-CNN have been widely used in multi-spectral image classification [46][47][48].…”

Research on Grape-Planting Structure Perception Method Based on Unmanned Aerial Vehicle Multispectral Images in the Field

“…The field grape validation set was classified using SVM [45], ML [46], RF [47], and ISO DATA, and the classification results are shown in Figure 8 and Table 6. The results showed that the cultivation of grapes in the field was complex, containing not only non-crops, such as barns and land, but also green vegetation, such as maize and grape weeds.…”