Fig. 1. Fully-connected neural network model with a hidden layer

Fig. 2. Deformation modes of simulated speckle image. (a) Uniaxial tensile deformation; (b) shear deformation; (c) combined deformation of tension and shear

Fig. 3. Deformation of one dimension embedding each other

Fig. 4. Based on the random deformation field u and v, the simulated speckle images before and after deformation are obtained. (a) Simulated speckle image before deformation; (b) cloud diagram of given field v; (c) simulated speckle image after displacement; (d) cloud diagram of given field u

Fig. 5. CNN model for displacement field recognition of digital speckle image

Fig. 6. Simulated speckle image displacement field test results by CNN method. (a)--(d) Given displacement field; (e)--(h) measured displacement field by CNN

Fig. 7. Displacement field calculated results by CNN method and NIG-GA method. (a)--(d) Given displacement field; (e)--(h) measured results by CNN method; (i)--(l) measured results by NIG-GA method

Fig. 8. Uniaxial tensile image of silica gel. (a) Image when L=0.35 mm, F=0.23 N; (b) image when L=0.4 mm, F=0.24 N; (c) selected calculation area of displacement field

Fig. 9. Displacement field u for experiment image calculated by CNN method