Fig.1

Three-dimensional point cloud object recognition algorithm framework based on feature matching"

Fig.2

Target oriented bounding box diagram"

Fig.3

Three-dimensional point cloud mapping flow chart based on PCA"

Fig.4

Flowchart of contour information extraction"

Table 1

Comparison of shape and skeleton features"

Fig.5

Shape and skeleton feature coding diagram based on centroid"

Fig.6

Original three-dimensional point cloud diagram of eight furniture models"

Fig.7

Schematic diagram of three-dimensional point cloud after density drop sampling"

Fig.8

Schematic diagram of three-dimensional point clouds with density of 3% and signal-to-noise ratio of different positions"

Fig.9

Two-dimensional map of the three-dimensional point cloud after rotation at different angles"

Fig.10

Shape and skeleton feature recognition accuracy comparison under different sampling points"

Fig.11

Processing time of shape features under different sampling points"

Fig.12

Identification accuracy under different down sampling point cloud density conditions"

Fig.13

Recognition accuracy under different position signal-to-noise ratio conditions"

Table 2

Correlation coefficient table between ${T_x}$ and ${T_s}$"

Fig.14

Polynomial fitting diagram"

Fig.15

Exponential fitting diagram"

Fig.16

Confusion matrix of the skeleton matching algorithm result"

Fig.17

Confusion matrix of the shape content matching algorithm result"

Fig.18

Confusion matrix of the proposed algorithm recognition result"

Fig.19

Comparison of recognition accuracy between the proposed algorithm and shape context algorithm under different position signal-to-noise ratios conditions"

Fig.20

Original three-dimensional point cloud illustrations for the remaining five models of guitar categories"

Fig.21

Original three-dimensional point cloud illustrations for four indoor object models"

Fig.22

Confusion matrix of intra-class sample recognition results of the proposed algorithm"

Fig.23

Confusion matrix of cross-class sample recognition results for the proposed algorithm"