Fig. 1. Flow chart of K-mRMR fusion algorithm

Fig. 2. Depth image preprocessing. (a) Original depth image; (b) image after joint bilateral filtering processing

Fig. 3. Scoring curves of fusion images with different atomic numbers

Fig. 4. (a) Dictionary atoms of original selection; (b) dictionary atoms after optimized selection

Fig. 5. RGB-D image fusion effect of a hat used as the identification target. (a) Depth image; (b) gray image; (c) PCA-KSVD method; (d) NSCT-KSVD method; (e) K-mRMR method

Fig. 6. RGB-D image fusion effect of a bowl used as the identification target. (a) Depth image; (b) gray image; (c) PCA-KSVD method; (d) NSCT-KSVD method; (e) K-mRMR method

Fig. 7. Joint detection of RGB-D by SIFT+SVM

Fig. 8. Joint detection of RGB-D based on CNN