Sensing Superblocks
Image Analytics for urban planning
The geological era of the Anthropocene is ushering in a new shift in weather patterns affecting climate, biodiversity, and geomorphology, ultimately impacting our cities and built environments. Today, urban planners and administrators are faced with the urgent need to adapt our built environments to the rising levels of air contamination, an invisible enemy that is deteriorating the well-being of our entire population.
Numerous studies have linked air pollution to congenital anomalies, particularly exposure to traffic-related gases. The Air Quality Index (AQI) categorizes the risks associated with air pollutants based on different scales, with warnings beginning at 51–100 and levels between 201–300 causing a significant increase in respiratory effects, while levels above 301–500, could translate into serious heart or lung disease or even premature mortality.
To face the unprecedented threat triggered by pollution, cities must reprogram their urban infrastructures and reshape their urban patterns towards safer, healthier, and more ecological solutions. The extension plan proposed by Ildefonso Cerdà for the city of Barcelona was designed to balance lighting, ventilation, public spaces, greenery, and mobility into a hybrid urban configuration. However, Barcelona is today one of the most polluted cities in Europe, partly due to its geography, high traffic density, and a large proportion of diesel-powered vehicles.
In 2016, Barcelona introduced a new urban planning model called the Superblock, which aims to reduce motorized transportation, promote sustainable mobility and active lifestyles, provide urban greening, and mitigate the effects of climate change. This planning approach regulates access to cars and vehicles in public streets while enabling the extension of walkable areas and cycling paths, threatening the urban fabric as a programmable surface.
Along with the development of new urban strategies, novel data-driven instruments have emerged, providing different approaches to inform spatial planning. This article highlights several techniques triggered by computer vision and machine learning algorithms to analyze image-based data and extract meaningful metrics to inform spatial transformations and estimate CO2 emissions in an urban environment.
Towards Spatial Analytics
As technology continues to advance and expand, it is becoming increasingly important to establish decision-making protocols that prioritize resilient, participatory, and responsive public spaces. In urban environments, this means embracing the emerging opportunities presented by big data analytics and urban informatics to determine metrics and transform urban spaces.
Big data represents a wide spectrum of observational and informal data produced through a variety of activities in the urban environment. Urban informatics is the operational area dedicated to exploring and understanding urban systems by leveraging novel sources of data. This emerging field has tremendous potential to improve dynamic urban resource management, drive theoretical insights and knowledge discovery of urban patterns and processes, increase urban engagement and civic participation, and facilitate innovations in urban management, planning, and policy analysis.
Within the realm of big data analytics, image-based data analytics has become increasingly important. Machine learning and computer vision algorithms have enabled a variety of image-based applications, and image analytics has emerged as a disruptive technology to sense, capture, and describe complex spatial dynamics.
While mobile data has traditionally been the go-to source for spatial dynamics information, it presents limitations such as positioning inaccuracies and an inability to provide useful insights regarding transportation means or individual behaviours. Image analytics, on the other hand, can introduce new workflows and provide information-rich descriptions of behaviour that can support the development of novel design systems.
By generating datasets from video frames using computer vision and machine learning, convolutional neural networks can be used to classify video and image contents and provide valuable insights into spatial dynamics. This emerging technology holds great promise for informing design solutions through spatial-sensing data and enabling the development of an architecture that considers the interactions between occupants and space based on factual observations.
DEEP LEARNING AND SPATIAL ANALYTICS
The rapid increase of publicly available labelled data and the development of GPU computing has significantly enhanced the performance and efficiency of deep learning algorithms. Unsupervised training logic introduced by Hinton and LeCun in the 80s, and later on boosted by AlexNet, marked a major breakthrough, which subsequently led to the development of deep architectures and deep learning algorithms for computer vision applications, such as object detection, image recognition, motion tracking, pose estimation, and semantic segmentation. These algorithms have played a pivotal role in the emerging field of spatial analytics, with a wide range of applications such as Crowd Management, Public Space Design, Virtual Environments, Visual Surveillance, and Intelligent Environments.
In the context of the Superblock, the analysis of mobility and patterns of spatial occupancy are essential parameters for the activation of a new urban model based on restricted accessibility for cars and heavy vehicles. To create such a new public space, a mobility model needs to be developed to ensure accessibility, comfort, safety, and multi-functionality. Indeed, a data-driven protocol can provide citizens with the opportunity to exercise their rights to culture, leisure, expression, demonstration, and movement.
By leveraging these technologies, a responsive approach can be introduced to adapt, reconfigure, and extend public spaces, providing criteria for intervention to regulate mobility, and enabling the creation of more accessible and liveable public spaces.
In the coming articles, we will delve into the methodologies and solutions developed by the author and the Noumena team. Our focus will be on the application of advanced algorithms based on computer vision and machine learning to interpret image data and extract valuable insights from the actual usage of public spaces.
Acknowledgement
Noumena: Oriol Arroyo, Salvador Calgua, Maria Espina, Iacopo Neri, Paula Osés Noguero, Carlo Ciuffarelli
