THREE-DIMENSIONAL POINT CLOUD LABEL LEARNING DEVICE, THREE- DIMENSIONAL POINT CLOUD LABEL ESTIMATION DEVICE, METHOD, AND PROGRAM
A clustering unit ( 101 ) divides an input three-dimensional point cloud into a plurality of clusters and outputs cluster data, a surrounding point sampling unit ( 102 ) extracts, for each of the plurality of clusters, a surrounding three-dimensional point cloud present within a predetermined distance of the cluster based on the three-dimensional point cloud and the cluster data, a learning unit ( 103 ) receives, as inputs, extended cluster data including information on a three-dimensional point cloud included in each cluster obtained by the division and information on the extracted surrounding three-dimensional point cloud and a correct answer label indicative of an object to which the three-dimensional point cloud included in each cluster belongs, and learns a parameter of a DNN for estimating a label of each cluster from the extended cluster data, and an estimation unit ( 104 ) inputs the extended cluster data related to the cluster of which the label is unknown to the DNN of which the parameter is trained to estimate the label of each cluster.
1 . A three-dimensional point cloud label learning device comprising a processor configured to execute a method comprising:
dividing an input three-dimensional point cloud into a plurality of clusters;
extracting a surrounding three-dimensional point cloud present within a predetermined distance from a cluster of the plurality of clusters; and
receiving, as inputs, cluster information including information on a three-dimensional point cloud in the cluster and information on the surrounding three-dimensional point cloud, and a label indicative of an object to which the three-dimensional point cloud in the cluster belongs; and
learning a parameter of a deep neural network for estimating the label from the cluster information.
2 . The three-dimensional point cloud label learning device according to claim 1 , wherein,
in the deep neural network, the receiving further includes:
inputting the information on the three-dimensional point cloud in the cluster and the information on the surrounding three-dimensional point cloud collectively to a partial structure for extracting a feature amount of the input three-dimensional point cloud, and
extracting the feature amount.
3 . The three-dimensional point cloud label learning device according to claim 1 , wherein,
in the deep neural network, the learning further includes:
inputting information on the three-dimensional point cloud in the cluster to a first partial structure for extracting a feature amount of the input three-dimensional point cloud;
extracting a first feature amount, inputs the information on the surrounding three-dimensional point cloud to a second partial structure;
extracting a second feature amount; and
acquiring a feature amount, wherein the feature amount corresponds to a combination of the first feature amount and the second feature amount.
4 . The three-dimensional point cloud label learning device according to claim 1 , the processor further configured to execute a method comprising:
inputting the cluster information related to the cluster of which the label is unknown to the deep neural network of which the parameter is trained by the learning to estimate the label of the cluster of which the label is unknown.
5 . A three-dimensional point cloud label estimating device comprising a processor configured to execute a method comprising:
dividing an input three-dimensional point cloud into a plurality of clusters;
extracting a surrounding three-dimensional point cloud present within a predetermined distance from a cluster of the plurality of clusters; and
inputting cluster information related to the cluster of which a label is unknown to a deep neural network for estimating the label of each cluster from the cluster information which is trained by using, as inputs, the cluster information and information on the surrounding three-dimensional point cloud, and the label indicative of an object to which a three-dimensional point cloud belongs to estimate the label of the cluster of which the label is unknown.
6 . A three-dimensional point cloud label learning method comprising:
causing division an input three-dimensional point cloud into a plurality of clusters;
causing extraction of a cluster of the plurality of clusters, a surrounding three-dimensional point cloud present within a predetermined distance of the cluster; and
causing a receipt of, as inputs, cluster information including information on a three-dimensional point cloud in the cluster and information on the surrounding three-dimensional point cloud, and a label indicative of an object to which the three-dimensional point cloud included in each cluster belongs, and learn a parameter of a deep neural network for estimating the label of each cluster from the cluster information.
7 - 8 . (canceled)
9 . The three-dimensional point cloud label learning device according to claim 1 , wherein the input three-dimensional point cloud indicates geometric information of an object, wherein a three-dimensional point cloud includes a three-dimensional point, and wherein the three-dimensional point includes three-dimensional position coordinates and at least an attribute information associated with the object.
10 . The three-dimensional point cloud label learning device according to claim 1 , wherein the object includes an artificial object with a structure in which a given cross-sectional shape is stretched.
11 . The three-dimensional point cloud label learning device according to claim 2 , the processor further configured to execute a method comprising:
inputting the cluster information related to the cluster of which the label is unknown to the deep neural network of which the parameter is trained by the learning to estimate the label of the cluster of which the label is unknown.
12 . The three-dimensional point cloud label learning device according to claim 3 , the processor further configured to execute a method comprising:
inputting the cluster information related to the cluster of which the label is unknown to the deep neural network of which the parameter is trained by the learning to estimate the label of the cluster of which the label is unknown.
13 . The three-dimensional point cloud label estimating device according to claim 5 ,
in the deep neural network, the receiving further includes:
inputting the information on the three-dimensional point cloud in the cluster and the information on the surrounding three-dimensional point cloud collectively to a partial structure for extracting a feature amount of the input three-dimensional point cloud, and
extracting the feature amount.
14 . The three-dimensional point cloud label estimating device according to claim 5 , wherein,
in the deep neural network, the learning further includes:
inputting information on the three-dimensional point cloud in the cluster to a first partial structure for extracting a feature amount of the input three-dimensional point cloud;
extracting a first feature amount, inputs the information on the surrounding three-dimensional point cloud to a second partial structure;
extracting a second feature amount; and
acquiring a feature amount, wherein the feature amount corresponds to a combination of the first feature amount and the second feature amount.
15 . The three-dimensional point cloud label estimating device according to claim 5 , the processor further configured to execute a method comprising:
inputting the cluster information related to the cluster of which the label is unknown to the deep neural network of which the parameter is trained by the learning to estimate the label of the cluster of which the label is unknown.
16 . The three-dimensional point cloud label estimating device according to claim 5 , wherein the input three-dimensional point cloud indicates geometric information of an object, wherein a three-dimensional point cloud includes a three-dimensional point, and wherein the three-dimensional point includes three-dimensional position coordinates and at least an attribute information associated with the object.
17 . The three-dimensional point cloud label estimating device according to claim 5 , wherein the object includes an artificial object with a structure in which a given cross-sectional shape is stretched.
18 . The three-dimensional point cloud label learning method according to claim 6 , in the deep neural network, the receiving further includes:
inputting the information on the three-dimensional point cloud in the cluster and the information on the surrounding three-dimensional point cloud collectively to a partial structure for extracting a feature amount of the input three-dimensional point cloud, and
extracting the feature amount.
19 . The three-dimensional point cloud label learning method according to claim 6 , wherein,
in the deep neural network, the learning further includes:
inputting information on the three-dimensional point cloud in the cluster to a first partial structure for extracting a feature amount of the input three-dimensional point cloud;
extracting a first feature amount, inputs the information on the surrounding three-dimensional point cloud to a second partial structure;
extracting a second feature amount; and
acquiring a feature amount, wherein the feature amount corresponds to a combination of the first feature amount and the second feature amount.
20 . The three-dimensional point cloud label learning method according to claim 6 , the processor further configured to execute a method comprising:
inputting the cluster information related to the cluster of which the label is unknown to the deep neural network of which the parameter is trained by the learning to estimate the label of the cluster of which the label is unknown.
21 . The three-dimensional point cloud label learning method according to claim 6 , wherein the input three-dimensional point cloud indicates geometric information of an object, wherein a three-dimensional point cloud includes a three-dimensional point, and wherein the three-dimensional point includes three-dimensional position coordinates and at least an attribute information associated with the object.
22 . The three-dimensional point cloud label learning method according to claim 6 , wherein the object includes an artificial object with a structure in which a given cross-sectional shape is stretched.