Intelligent Visual Grass Quality Detection System

Green spaces, an integral part of a public space system, are urban lands covered with all parts of vegetation, that promote the liveness, comfort, and integrity of public places, as stated in “Urban green spaces: a brief for action” by WHO (2016). Grasses constitute the major element in the green space environment and their quality can be assessed visually by detecting the color, patches, and unevenness of the grass-covered areas. This paper focuses on developing a green spaces quality detection system that will help to monitor, maintain and assess the quality of the urban green space areas based on the local ecosystem. The detection system aims to monitor the status of the grass and to detect any aberrant grass conditions that will influence the general outlook of the public spaces. Initially, the image data on various grass patches are collected and labeled as healthy and unhealthy and other distinguishing factors are identified, and they are assessed with a machine learning model that is a deep convolutional neural network based on the architecture of the fully convolutional network (FCN). The model detects, identifies, and classifies the labeled images based on their distinguishing factors and determines whether the blocks of grass identified are healthy or unhealthy. Once the quality of the patch of grass is obtained, the evaluation report can be generated. As the proposed solution offers a customized solution based on the local climatic conditions, it is an area-specific solution that can be implemented by the various local authorities and municipalities to assess, maintain and monitor various green spaces. This solution will in turn enhance the overall quality of public spaces and aid in realizing the vision of a cleaner, safer, well-nurtured, and maintained ecosystem.