Real-time 3D point cloud obstacle discriminator apparatus and associated methodology for training a classifier via bootstrapping
First Claim
1. A controller for training a classifier, the controller comprising a processor configured to:
- classify point cloud data with a first classifier;
infer a first terrain map from the classified point cloud data;
reclassify the point cloud data with the first classifier based on the first terrain map;
train a second classifier based on the point cloud data reclassified with the first classifier based on the first terrain map;
compare a first classification result of the first classifier with a second classification result of the second classifier;
determine whether the first result is sufficiently close to the second result; and
in response to determining the first and second results are not sufficiently close, classify the point cloud data with the second classifier and infer a second terrain map from the point cloud data classified by the second classifier.
2 Assignments
0 Petitions
Accused Products
Abstract
Training a strong classifier by classifying point cloud data with a first classifier, inferring a first terrain map from the classified point cloud data, reclassifying the point cloud data with the first classifier based on the first terrain map, and training a second classifier based on the point cloud data reclassified with the first classifier based on the terrain map. The point cloud data is then classified with the second classifier, and the procedure followed with the first classifier is iteratively repeated until a strong classifier is determined. A strong classifier is determined when a probability of a terrain map matching a given terrain for the strong classifier is approximately equal to a probability of a terrain map matching the given terrain for a prior trained classifier.
-
Citations
19 Claims
-
1. A controller for training a classifier, the controller comprising a processor configured to:
-
classify point cloud data with a first classifier; infer a first terrain map from the classified point cloud data; reclassify the point cloud data with the first classifier based on the first terrain map; train a second classifier based on the point cloud data reclassified with the first classifier based on the first terrain map; compare a first classification result of the first classifier with a second classification result of the second classifier; determine whether the first result is sufficiently close to the second result; and in response to determining the first and second results are not sufficiently close, classify the point cloud data with the second classifier and infer a second terrain map from the point cloud data classified by the second classifier. - View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17)
-
-
18. A method of training a classifier with a processor, the method comprising:
-
classifying point cloud data with a first classifier; inferring a first terrain map from the classified point cloud data; reclassifying the point cloud data with the first classifier based on the first terrain map; training a second classifier based on the point cloud data reclassified with the first classifier based on the first terrain map; comparing a first classification result of the first classifier with a second classification result of the second classifier; determining whether the first result is sufficiently close to the second result; and in response to determining the first and second results are not sufficiently close, classifying the point cloud data with the second classifier and inferring a second terrain map from the point cloud data classified by the second classifier.
-
-
19. A non-transitory recording medium storing instructions that when executed by a processor cause the processor to perform a method of training a classifier, the method comprising:
-
classifying point cloud data with a first classifier; inferring a first terrain map from the classified point cloud data; reclassifying the point cloud data with the first classifier based on the first terrain map; training a second classifier based on the point cloud data reclassified with the first classifier based on the first terrain map; comparing a first classification result of the first classifier with a second classification result of the second classifier; determining whether the first result is sufficiently close to the second result; and in response to determining the first and second results are not sufficiently close, classifying the point cloud data with the second classifier and inferring a second terrain map from the point cloud data classified by the second classifier.
-
Specification