CodePudding user response:
From a technical point and the other: based on the qualcomm SDK, direct call existing apis can complete recognition and tracking, you just need to get track of POSE matrix can be used to draw, draw, can use Unity3D, OpenGLES or can be directly used for parsing model rendering and animation, of course with Unity3D can simplify a lot of learning cost, are the benefits of using OpenGLES also need to cost many time, and the existing use of rendering, animation, open source code much, much to write their own, the beginner is not recommended, so the advantage is that can quickly start solutions, based on qualcomm SDK development application there will be a watermark, if you want to remove the watermark, or want to use the cloud recognition version is to pay, based on OpenCV, Markless recognition and tracking of all have no ready-made comparing with open source code, can be found in most of the open source code is based on the Marker, and the efficiency is very low, do the AR part recognition and tracking are the need to develop, and the cost of this part of the study, development costs are high, think it is very difficult to achieve high efficiency/accuracy/robust, although this part of the thesis also many can be found, but to implement will find all sorts of problems, advantages, is to improve the ability of self-study, but want to reach the level of business, the difficulty is very high,