Summary of three-dimensional point cloud target extraction

three-dimensional point cloud target extraction 1. Preface

Unconsciously, has been to graduate student's third year, next year will graduate to participate in the work, offer to take a few, basic work is almost set up, after all, graduate students two years of time the main research direction is three-dimensional point cloud target recognition and three-dimensional reconstruction (think is the research direction bar), as the domestic graduate student situation, A few years of glorious time have been dedicated to the mentor of several projects, do things are miscellaneous, proficient skills are not, riding the season to graduate, is now dedicated to the three-dimensional point cloud target recognition of some views (should only be suitable for just contact to make three-dimensional point cloud of the small Shuo see, but seemingly domestic engaged in this piece of the university is not 2. Three-dimensional laser point cloud application domain Three-dimensional laser point cloud acquisition principle and the basic characteristics are not introduced, compared to the comparative professional knowledge, talk about the application of three-dimensional laser point cloud. Personally, three-dimensional laser application prospects are very wide, especially with other sensors, such as image sensors. Applications include, but are not limited to: intelligent driving, smart home, three-dimensional reconstruction, digital earth, urban planning, disaster prevention and mitigation, marine mapping and so on. Only two examples are discussed: Smart driving and three-dimensional reconstruction. 2.1 Smart Driving




Smart driving is the need for high-precision 3D map, what is called high-precision 3D map, is able to express the lane level map, including lane lines, markings, road parameters and so on information.
2.2 Three-dimensional reconstruction



Three-dimensional reconstruction is generally based on the three-dimensional point cloud mesh mesh structure, of course, can also be reconstructed by structural.
Well, from the above applications, three-dimensional point cloud research is very meaningful. But in fact, the three-dimensional point cloud we get from the sensor has no objects, just a bunch of chaotic three-dimensional points.


However, in the above application, the chaotic point cloud is impossible to deal with. For example, three-dimensional reconstruction, if there is no object of the target, the entire point cloud reconstruction, the end result is that there is a good-looking overall model, but also limited to good-looking, you do not know how many buildings in this model, how many trees, terrain, you do not know the specific location of concrete buildings and so on information. Therefore, the object of three-dimensional point cloud is very important, is the basis of other applications. Well, that's all, so let's talk about goal-picking in the next section.