News
It is no secret that remote sensing can be considered one of the greatest advances ever produced by science and technology related to the study of the Earth's surface. Earth observation satellites provide data covering the most varied portions of the electromagnetic spectrum at different spatial, temporal, and spectral resolutions. That is why remote sensing has been successfully applied in several areas of knowledge for decades. In recent years, interest in this technology has increased significantly. Its exploitation has moved from developments led by government intelligence agencies to those carried out by companies and general users. In addition, recent innovations in drones have also been providing a powerful solution for the market. These aerial systems cost substantially less and offer potential gains in spatial resolution due to the lower altitudes at which surveys can be performed.
Speak with one of our specialists and discover how Pix Force can transform your business
A relatively new concept
Generally speaking, field data collection is a common practice among companies. However, in addition to having limitations, this type of approach consumes a large amount of human and financial resources, which often makes the practice unfeasible. The development of computationally efficient techniques to transform the large amount of existing remote sensing data into useful information is fundamental. And the best part: it has been under the spotlight in recent times, especially regarding the automation of the image analysis process. Much of the demand for satellite or drone imagery requires time-consuming and expensive visual interpretation. This is not possible to do on a large scale without the use of techniques based on artificial intelligence (AI). Although the term AI and its concepts have existed since the 1960s, it is considered new in the market. In a general context, it can be said that these technologies have been widely applied to science and engineering problems for almost two decades. However, their application in the area of remote sensing is relatively new. Within the broad field of AI, different methods based on Machine Learning (ML), or even Deep Learning (DL), have proven to be very efficient. One example is the processing of optical (hyperspectral and multispectral) and radar images in extracting different types of land cover. As well as extracting roads, buildings, among others, thus being able to solve a range of problems related to monitoring the environment as a whole.
Combining artificial intelligence and satellite imagery
With the increasing availability of data, the use of AI has been gaining more attention in different sectors, especially due to its potential to leverage big data. The application of remote sensing techniques, together with those of AI, represents an opportunity to perform advanced management analyses of production systems, thus improving their efficiency. Thanks to advancements, information can now be quantified with precision and big data can be integrated into monitoring and predictive management tools, benefiting various sectors. Classification is the most commonly used technique for remote sensing data processing. In this context, land cover maps, for example, are one of the most essential inputs when working with environmental monitoring. ML-based classification approaches have recently become the main focus of remote sensing literature. ML algorithms are capable of modeling complex spectral signatures, and they can accept a variety of input data. Furthermore, these methods tend to reproduce processes with greater accuracy, especially for complex data with many predictor variables. Several studies demonstrate the success of ML algorithms for the most diverse purposes. For example: crop mapping, species classification, land cover mapping, wetland classification, forest census, crop monitoring, water resources management, mineral exploration, crop water stress estimation, among others. Once classification approaches are perfected, other sectors that previously did not make use of remote sensing can now benefit. The electric power sector is one of those cases.
Technology and the electric power sector
Energy represents an essential input to society and, in Brazil, the electrical system allows the exchange of energy produced in all regions of the country. Among the main challenges faced within this sector are the precise georeferencing of towers, energy theft, invasions, or irregular constructions in the so-called "right-of-way strips" of transmission lines, among others. Until recently, ground surveillance was the only form of monitoring used by companies, even though it generated high costs and had very low efficacy. In addition, it was often proven to be unfeasible due to difficulty of access or the size of the area. In this sense, it becomes evident that finding alternative and efficient ways of monitoring is an urgent need. The right-of-way strips consist of the ground space occupied by power transmission lines. As a territory with the potential to generate risks for the population, these places have numerous restrictions regarding their use, given their purpose: to protect the electrical system and society. The major challenge linked to the use of remote sensing for this type of monitoring is related to technical issues intrinsic to product characteristics versus costs.
How is it done vs. how to do it?
Technically, the extraction of surface information using optical satellite images is directly dependent on the presence of clouds. This is even more pronounced in tropical regions. An alternative to bypass this limitation is the use of radar images which, in general, require more complex processing and interpretation than optical sensors. When talking about low-cost continuous monitoring, the spatial resolution of the images is one of the main constraints. A coarse spatial resolution can prevent the identification of the targets of interest and/or cause errors in image interpretation. On the other hand, high spatial resolution satellite images usually come with a high cost. Depending on the application, the same can occur with drone images. Despite their high spatial resolution, they are only capable of covering small areas, often requiring multiple flights to image the entire region. It can be said that the greatest obstacle remains finding the ideal balance between cost and the reliability of the results obtained.
What to expect from the future?
Fortunately, with the advent of new sensors, it is increasingly possible to achieve high accuracy in image classifications at reduced costs. The CBERS program is a great example. Born from a partnership between Brazil and China, it has been providing satellite images since 1999 that are used in monitoring deforestation in the Amazon, as well as studies and mapping in the areas of water resources, agriculture, vegetation, urban planning, among others. CBERS products have been the subject of many studies and technical papers over the years, and, very recently, CBERS 04A images have begun to be distributed. In this way, Brazil became the distributor of the best free satellite product, with up to 8 meters of spatial resolution in the WPM camera (reaching up to 2 meters in fusion with the panchromatic band). Behind the rise of the use of AI and satellite imagery are three main trends: the increased availability of remote sensing data, the advancement of artificial intelligence (particularly ML and DL), and the availability of massive computing power. The geospatial data revolution offers a tantalizing promise: nearly limitless opportunities when combined with imagination. Since the possibilities are endless, it is difficult to think of any business sector or government entity that cannot benefit from the combined use of remote sensing and AI. Article produced by Pix Force developer Pâmela Käfer. References • Eskandari, R., Mahdianpari, M., Mohammadimanesh, F., Salehi, B., Brisco, B., & Homayouni, S. (2020). Meta-Analysis of Unmanned Aerial Vehicle (UAV) Imagery for Agro-Environmental Monitoring Using Machine Learning and Statistical Models. Remote Sensing, 12(21), 3511.
• Aneel Website
• Paul Hahn, (2019). The European Business Review https://www.europeanbusinessreview.com/three-trends-driving-the-geospatial-ai-revolution/
• Jung, J., Maeda, M., Chang, A., Bhandari, M., Ashapure, A., & Landivar-Bowles, J. (2021). The potential of remote sensing and artificial intelligence as tools to improve the resilience of agriculture production systems. Current Opinion in Biotechnology, 70, 15–22. doi:10.1016/j.copbio.2020.09.003
• Kussul, N., Lavreniuk, M., Skakun, S., & Shelestov, A. (2017). Deep Learning Classification of Land Cover and Crop Types Using Remote Sensing Data. IEEE Geoscience and Remote Sensing Letters, 14(5), 778–782. doi:10.1109/lgrs.2017.2681128
• Lary, D. J., Alavi, A. H., Gandomi, A. H., & Walker, A. L. (2016). Machine learning in geosciences and remote sensing. Geoscience Frontiers, 7(1), 3–10. doi:10.1016/j.gsf.2015.07.003
• Lee, C. A., Gasster, S. D., Plaza, A., Chang, C.-I., & Huang, B. (2011). Recent Developments in High Performance Computing for Remote Sensing: A Review.
• Martin, P. G., Connor, D. T., Estrada, N., El-Turke, A., Megson-Smith, D., Jones, C. P., … & Scott, T. B. (2020). Radiological Identification of Near-Surface Mineralogical Deposits Using Low-Altitude Unmanned Aerial Vehicle. Remote Sensing, 12(21), 3562.
• Maxwell, A. E., Warner, T. A., & Fang, F. (2018). Implementation of machine-learning classification in remote sensing: an applied review. International Journal of Remote Sensing, 39(9), 2784–2817.
• Valente, S. (2015). Remote sensing in the study of irregular occupation in power transmission line easement (Master’s thesis, Universidade Federal de Pernambuco).
Fabio Caraça
Fábio Caraça is the Chief Growth Officer at Pix Force. He leads Pix Force's transformation into a scalable SaaS operation, combining strategic vision, culture, and high-impact execution.
