[ad_1]
Geospatial scientists at Australian University RMIT Melbourne have developed a new program that uses artificial intelligence to monitor signage that requires replacement or repair, using Google's Street View imagery.
Road authorities around the world are currently spending a lot of time and money monitoring and manually recording the geolocation of traffic infrastructure, a task that also exposes workers to unnecessary risks associated with the passage of vehicles. RMIT's fully automated system is trained in object detection with the aid of an artificial intelligence to identify signal boards on freely available images. The results that have just been published in the "Computers, Environment and Urban Systems" show that the system detects panels with an accuracy of almost 96%, identifies their type with an accuracy of almost 98% and can record their precise geographic location from 2D images. During investigations, the research team discovered that mandatory GPS location data in existing road sign databases was often inaccurate, sometimes up to 10 meters.
& # 39;)
}
// ->
& # 39;)
}
// ->
"The proof-of-principle model was developed to be able to see the" Stop "and" Give Way "signs, but can be formed to identify many other inputs and is easily expandable for use by local governments and road authorities. . Boards have an obligation to monitor this infrastructure, but currently have no effective or cheap means, "said Andrew Campbell, senior author of the study and student in Geospatial Sciences at RMIT. "By using free and open source tools, we have developed a fully automated system to do this work and do it more accurately. Manual tracking of these signs by people who may not be qualified geoscientists introduces human error into the database. Our system, once set up, can be used by any spatial analyst. you simply tell the system what area you want to monitor and look after it for you. "
Dr. Chayn Sun, a geospatial scientist and project co-leader at RMIT, said, "The fact that some boards are already fixing cameras on garbage trucks to collect street images shows how valuable visual data is given what technology could do now. These images are essential for local governments in asset monitoring and management and, with the huge amount of geospatial applications in full swing, this information will only gain in value. Our application is one of the first applications of this technology to meet a specific industry need, but many more will emerge over the next few years. Sequences from other sources, such as those from garbage truck cameras or any other georeferenced road network imagery collected by the councils, may also be introduced into the system. Where sequences are already gathered, our research can provide boards with a cost-effective tool for generating ideas and data from this existing resource. "
The project was co-led by Sun and RMIT's geospatial researcher, Dr. Alan Both, of the University's Center for Urban Research. The RMIT research team is working with local governments on heat response strategies using Google Street View imagery to analyze the shadows of street trees.
Source link
