7 The Intelligent Classification and Extraction of Point Clouds - LiDAR360 MLS

everything hey everyone today I will introduce to you the intelligent classification and extraction of Point clouds this function uses deep learning methods to classify Point cloud data

and for places where the classification effect is not good manual editing can be used to modify the classification and the point cloud data can be extracted according to the elevation

intensity return number time and classic requirements Etc deep learning classification is to use deep learning methods to classify Point cloud data

click the Deep learning category to pop up the setting dialog window parameter setting of this function mainly includes the following points

Point Cloud file check the point cloud data that needs to be classified [Music] mode the software currently provides two

deep learning models we can choose different modes according to actual needs and data conditions foreign

types GPU and CPU if the performance of the computer graphics card is high we can choose the GPU mode and the software will list the GPU model

and video memory information and the GPU sitting interface if the computer has multiple graphics cards you must pay attention to manual

switching to the graphics card that can run the Deep learning environment it is recommended to use Nvidia graphics card for GPU mode classification [Music]

the classification efficiency of GPU will be more than twice that of the CPU [Music] classified mapping the software will be divided into 11

categories by default which are unclassified such as noise ground low vegetation High vegetation building wire

static cars Dynamic cars car rails poles pedestrians such as moving people bicycle electric vehicles Etc

the labels of the default categories and the software basic categories are the same and of course we can reset them if needed the batch size indicates the number of

Point Cloud samples processed each time the larger the computer memory and the higher the performance of the graphics card the larger the value that can be set and the faster the overall

processing speed will be the software has set the default value according to the performance of the computer we can modify it ourselves

but it is still recommended to use the value recommended by the software click ok to start the classification when completed we can see that our original data has been classified into

these 11 categories [Music] since the effect of deep learning is not

100 accurate we can modify the data by manual editing also [Music] click classify selection area

in the edit window and classify settings window will pop up we can choose polygon rectangle sphere or Circle to frame the area to be

classified or choose an online area an offline error or in the playing distance settings set the maximum value minimum value

plane thickness and robust fitting Etc of the distance to select moving objects above the ground roads below the plane or moving objects on the ground [Music]

you can also reverse the selection by subtracting the area if the box is wrongly selected it can be

canceled directly or by pressing Ctrl Z the software provides a variety of frame selection methods which can be applied to a variety of point-cloud data scenarios

here we take the polygon as an example click to frame the area to be classified check the category of the framed area in the source category

select the category to be assigned in the Target category [Music] and click the classified button then we can see the frame selection area

has been classified into the target category in the extraction panel the software provides five extraction

methods elevation intensity return number time and class to extract by elevation first select the point cloud data to be extracted

and then set the maximum and minimum values according to our needs the extracted Point cloud data will be saved in the folder under the same path as the original Point cloud data by

default we can also set A New Path it is recommended to set it under the same path here to facilitate subsequent data management

click ok to start the extraction when completed a pop-up window will prompt whether to add to the current project or create a new project we can see that the point cloud data

value at this time shows the corresponding elevation range after extraction in the operation of extracting by intensity is the same as the above steps

to extract by return first select the point cloud data to be extracted by return number set return number from one to seven times and then click ok to start the

extraction should be noted here that if the original Point cloud data does not exist the return number selected by the user cannot be extracted

to extract by time it is also necessary to select the extracted Point cloud data and then input the extracted maximum and minimum values as well as the start time

and in time if we want to extract the point Cloud at a specified interval just enter the interval time in the text box and then click this button

the start time and end time will increase as is set in interval we can also add the input time range to the range list

[Music] or import a data file in the dot txt format externally and it should be noted that the time

range in the file needs to correspond to the maximum and minimum values of the point cloud data if the time range is Miss set it can be deleted

when completed all Point cloud data within this range can be extracted [Music] to extract by classic

we can select the point cloud data to be extracted according to our actual needs choose a certain category or multiple categories to be extracted

and then click the OK button to extract [Music] when the extraction is complete it will also be added to the current project

[Music] choose the display Point classification here and we can see that there are only the categories we need in the point cloud data

this concludes the video about Point Cloud intelligent classification and extraction thank you for watching