Individual Tree Crown Detection using GAN and RetinaNet on Tropical Forest

Roslan, Zhafri Hariz; Long, Zalizah Awang; Ismail, Roslan

doi:10.1109/imcom51814.2021.9377360

Cited by 6 publications

(3 citation statements)

References 30 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Commonly, forestry inventory is estimated by detecting the trees. Several works proposed this approach as a way of assessing quantitatively the forest yield, forest biomass, and carbon dynamics from high-resolution remote sensing or UAV-based imagery [15][16][17][18][19][20][21][22]. The inventory from a certain ecosystem can also be estimated by mapping it through satellite images, as was made in [23] for a mangrove ecosystem.…”

Section: Vision-based Perceptionmentioning

confidence: 99%

Unimodal and Multimodal Perception for Forest Management: Review and Dataset

et al. 2021

View full text Add to dashboard Cite

Robotics navigation and perception for forest management are challenging due to the existence of many obstacles to detect and avoid and the sharp illumination changes. Advanced perception systems are needed because they can enable the development of robotic and machinery solutions to accomplish a smarter, more precise, and sustainable forestry. This article presents a state-of-the-art review about unimodal and multimodal perception in forests, detailing the current developed work about perception using a single type of sensors (unimodal) and by combining data from different kinds of sensors (multimodal). This work also makes a comparison between existing perception datasets in the literature and presents a new multimodal dataset, composed by images and laser scanning data, as a contribution for this research field. Lastly, a critical analysis of the works collected is conducted by identifying strengths and research trends in this domain.

show abstract

Section: Vision-based Perceptionmentioning

confidence: 99%

Unimodal and Multimodal Perception for Forest Management: Review and Dataset

et al. 2021

View full text Add to dashboard Cite

show abstract

“…Deep learning (DL) a field that is currently trending in machine learning and it focuses on fitting large models with millions of parameters for a variety of tasks [17]. DL often employs Convolutional Neural Networks (CNNs) to detect objects based on image classification and anchor box regression [18]. Faster R-CNN, which is a well-known object detection model, gives high-recall region proposals at low cost through Region Proposal Network (RPN), can significantly improve the efficiency of object detection.…”

Section: Introductionmentioning

confidence: 99%

“…The results reported a mean average precision of 0.86. ROSLAN et al [18] integrated a GAN based model and a RetinaNet model to detect individual tree crowns; the results showed excellent F1 scores. However, datasets are the foundation of deep learning, which need large amounts of labeled training data.…”

Section: Introductionmentioning

confidence: 99%

Individual Tree Detection Based on High-Resolution RGB Images for Urban Forestry Applications

2022

View full text Add to dashboard Cite

Urban forests play an important role in urban ecosystems. They can not only beautify the urban environment but also help protect biodiversity and maintain ecological balance. Effective urban forest management is a basic requirement to ensure sustainable development. Traditional urban forest management usually requires the investment of a lot of materials and labor to conduct field research. RGB high-resolution aerial images have emerged as an efficient source of data for use in the detection and mapping of individual trees in urban areas. In recent years, there has been impressive progress in the field of deep learning methods for use in object detection. Semi-supervised learning is an effective way to deal with the problem that deep learning requires a large amount of labeled data. In this paper, we proposed an improved faster region-based convolutional neural network (Faster R-CNN) with Swin transformer method. Based on existing datasets, the model was trained and then transferred to new datasets. The method was evaluated within three distinct urban areas: a green space, a residential area and a suburban area. The experimental results indicate that our method achieved higher performance than other Faster R-CNN models. This method provides a reference in automated individual tree detection based on high-resolution images in urban areas for urban forestry managers.

show abstract