Augmented reality (referred to as AR) technology is to use the virtual world to enhance reality and enhance people's ability to perceive and interact with the real world. Video is one of the important information carriers for perceiving dynamic changes in the environment. As a universal public media resource, video is no longer just a visual product. The characteristics of space, time, rich information, diverse content, and authentic expression of video make it a natural geographical spatial information data.
SuperMap's desktop GIS product SuperMap iDesktopX 10i (hereinafter referred to as "iDesktopX") combines AR, AI, video and other technologies to provide video access, video map (also known as AR map), video target recognition and other functions, and has been used in practice. This article will explain the AR map technology integrated in iDesktopX.
In the field of GIS, it generally focuses on the expression of static scenes, and pays less attention to dynamic targets and scenes. Video, on the other hand, focuses on image space and pays less attention to geographic space. Combining video with GIS can simultaneously express geographic space and real-time dynamic scenes, and manage and analyze video and map data under a unified geographic reference.
Under the goal of simultaneously expressing geographic space and real-time dynamic scenes, iDesktopX mainly realizes five technical points of video access, video positioning, AR map, video measurement, and video AI target recognition .
Video access
Road camera surveillance video is the most important source of traffic information, but the video formats and acquisition methods are very diverse. How to efficiently manage video and better serve it for intelligent traffic monitoring has become an urgent problem to be solved at this stage.
iDesktopX supports access to real-time video streaming data captured by road cameras, drones, video cameras, mobile phones and other devices, and uses SuperMap iMobile App to shoot videos with AI detection results. Check the video monitoring content in real time on the desktop, which is convenient for understanding the traffic flow, driving order, and violations of the road section.
Figure 1 Access to the AI video stream captured by SuperMap iMobile
Video positioning
Both the video and the camera have spatial geographic information. Before the video is combined with GIS data for mapping and analysis, its coordinate information needs to be converted into the coordinate system of the actual geographic space to calibrate the video range. iDesktopX provides the ability of real-time conversion between video coordinates and map coordinates. When the shooting angle of the monitoring dome camera rotates and the focal length multiple changes, it can coordinate the video and geo-fence data in real time according to parameters such as camera azimuth angle, pitch angle, and field of view angle. Conversion, avoiding time-consuming and laborious manual adjustments and conversions.
The implementation technology of video positioning is shown in Figure 2 below: Point O is the camera point, AB is the ground range captured, A'B' is the corresponding photo, and C is the midpoint of the photo. The geographic coordinates (x, y, h), azimuth, pitch angle γ, vertical field of view α, and horizontal field of view β of the camera are known, and the geographic coordinates and A' of point C in the photo can be obtained according to the relationship of trigonometric functions The actual distance of B'. After obtaining the pixel size of the photo, the actual distance corresponding to each pixel can be obtained according to the corresponding relationship between the pixel size and the actual distance, and then the geographic coordinates of any point P' on the photo can be obtained.
Figure 2 The principle of video positioning based on trigonometric functions
AR map
AR map is a new form of map expression, based on real-time video stream data, according to certain mathematical rules, after unifying the video and map coordinate system through video positioning technology, it can superimpose points, lines, surfaces, 3D models, pipelines, terrains , images, aerial landmarks and other GIS vector and raster data for enhanced display and rich map information.
At the same time, iDesktopX supports drawing points, lines, planes, and text vector objects in the video, which is convenient for drawing fenced areas during geofence analysis; it also supports modifying object styles, which helps to distinguish and highlight object types.
Video Metering
Based on video positioning and delivery technology, it provides video distance and area measurement functions with an error accuracy of less than 1%, which can accurately measure the distance in the video, for example: when there is a traffic jam, the traffic jam distance can be accurately measured; toll collection at important expressway intersections At the station, the length of queuing vehicles at each toll window can be measured, and the queuing window can be intelligently recommended for subsequent vehicles.
Video AI target recognition
The combination of video and AI target recognition technology uses the TensorFlow framework to conduct deep learning and model training on the targets in the video. While the video is playing, the type of the target can be automatically identified from the video, and the number, location, and license plate number of the target can be obtained. and other information.
iDesktopX provides 80 types of objects, such as cars, buses, trains, traffic lights, pedestrians, water cups, notebooks and other transportation and daily necessities objects. At the same time, it supports setting the style of different object logo boxes to distinguish object types. Video target recognition can be applied to real-time monitoring of road traffic congestion, such as the congestion of motor vehicle traffic, non-motor vehicle traffic, and human traffic. The following figure shows the recognition results of video cars, buses, and non-motor vehicles.
Figure 3 Intelligent recognition of traffic targets in road cameras
