SBIR Phase I: Smartphone-Based Machine Learning and Computer Vision for Cost-Effective Verification of Forest Carbon Offsets

SBIR 第一阶段：基于智能手机的机器学习和计算机视觉，用于经济高效地验证森林碳抵消

• 批准号: 2212767
• 负责人: Patrick Leung
• 金额: $ 25.6万
• 依托单位: EARTHSHOT Labs PBC
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2022
• 资助国家: 美国
• 起止时间: 2022-09-15 至 2023-05-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=2212767&HistoricalAwards=false
• 关键词: SBIR; Phase; Smartphone; Based; Machine

The broader impact of this SBIR Phase I project is to make the gathering of tree measurements and other field-based nature observations dramatically easier. The project involves building a mobile app that uses computer vision and augmented reality to make measuring trees much easier than using the tape measures of today, in fact it will be as simple as scanning a barcode. This will result in much more nature imagery and other training data for use with ecological models, which will yield a better understanding of how ecosystems will change in the coming years. One important prediction this project will help with is forest growth which is very useful for generating carbon offsets that can finance the regeneration of large amounts of land. Better ecological models can also help communities prepare for changes in climate, sea level, soil quality and other ecological shifts in the coming decades. Millions of people will be potentially affected by climate change and its accompanying shifts in ecosystems, so arming communities and governments with better insights about these impacts will be critical. Furthermore, catalyzing many millions of hectares of nature regeneration projects will help mitigate the worst effects of climate change and other ecological challenges.This project involves the unique combination of computer vision with augmented reality in order to quickly and accurately measure trees. There are multiple computer vision neural networks being developed for the app. One of them uses bark and tree imagery in order to classify the tree species, an important feature needed for ecological models. Another model analyzes surrounding scenery and quickly identifies the closest tree trunk. The app then uses the phone’s augmented reality capabilities to gauge the distance and orientation of the phone from the trunk and combines these two to yield a diameter at breast height measurement. The research plan involves collecting data from specific regions in Panama and Brazil in conjunction with active nature restoration projects and gathering a critical mass of leaf and bark imagery so that species in those areas can be classified accurately. The scope will then increase to additional regions and later, measurements such as birdsong and other biodiversity markers. Over time the app could become a platform for many such nature observations, and it is expected to evolve to become more game-like to attract large numbers of people having fun experiences together and advancing the cause of nature science.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

SBIR第一阶段项目的更广泛影响是使树木测量和其他基于实地的自然观测的收集变得更加容易。该项目涉及构建一个移动的应用程序，该应用程序使用计算机视觉和增强现实技术，使测量树木比使用今天的卷尺更容易，事实上，它将像扫描条形码一样简单。这将产生更多的自然图像和其他训练数据，用于生态模型，从而更好地了解生态系统在未来几年将如何变化。这个项目将帮助的一个重要预测是森林的增长，这对于产生碳抵消非常有用，可以为大量土地的再生提供资金。更好的生态模型还可以帮助社区为未来几十年气候、海平面、土壤质量和其他生态变化做好准备。数百万人将受到气候变化及其伴随的生态系统变化的潜在影响，因此让社区和政府更好地了解这些影响至关重要。此外，促进数百万公顷的自然再生项目将有助于减轻气候变化和其他生态挑战的最坏影响。该项目涉及计算机视觉与增强现实的独特结合，以快速准确地测量树木。有多个计算机视觉神经网络正在为该应用程序开发。其中一个使用树皮和树木图像来分类树种，这是生态模型所需的一个重要特征。另一个模型分析周围的风景，并迅速确定最近的树干。然后，该应用程序使用手机的增强现实功能来测量手机与躯干的距离和方向，并将这两者结合起来，以产生乳房高度测量的直径。该研究计划涉及从巴拿马和巴西的特定地区收集数据，结合积极的自然恢复项目，并收集大量的树叶和树皮图像，以便对这些地区的物种进行准确分类。然后，范围将扩大到其他地区，随后将扩大到诸如鸟鸣和其他生物多样性标志等测量。随着时间的推移，该应用程序可能成为许多此类自然观察的平台，预计它将发展成为更像游戏的平台，以吸引大量的人一起享受有趣的体验，并推动自然科学事业。该奖项反映了NSF的法定使命，并通过使用基金会的智力价值和更广泛的影响审查标准进行评估，被认为值得支持。