Research On Super Resolution Of Tactile Image Based On Sensor Array

Tactile information such as pressure and temperature generated by the interaction of the robot,wearable devices,medical equipment are usually collected by the tactile sensor.The tactile image is formed by converting tactile information into digital image.At present,tactile sensor array is the main way to obtain the tactile information.Different from the visual sensing unit(such as CCD / CMOS),the sensing unit of tactile sensor array has the characteristics of diverse working principles,large size,complicated reading circuit and difficult integration.In order to improve the resolution of tactile image,this paper conducts a super-resolution study of tactile image based on sensor array.First of all,the tactile imaging system is studied,and it is pointed out that tactile image has the characteristics of diverse acquisition principles,complicated tactile interaction,and low resolution;afterwards,the super-resolution reconstruction algorithms of single and sequential tactile images are studied separately.Then,a super-resolution reconstruction system for tactile image based on mixed programming of MATLAB and MFC is built;finally,experiments of tactile image such as noise detection,motion estimation,and super-resolution reconstruction are implemented.The main research contents of this article are listed as follows:1.The tactile imaging system based on sensor array is studied,the characteristics of tactile image are summarized,the similarities and differences between tactile image and visual image are summarized.In addition to the above,the feasibility challenges of super-resolution reconstruction algorithm of tactile image are analyzed.2.Aiming at the problem that the collected tactile images may contain noise,a noise detection algorithm based on neighborhood quotient is proposed.Experimental results show that the cost of the noise detection algorithm is better than the cost of the local standard deviation noise detection algorithm,and the running speed is better than that of the neighborhood difference noise detection algorithm.Then,the super-resolution reconstruction algorithm of single tactile image is studied,and the interpolation algorithm is applied to the super-resolution of single tactile image.Because interpolation will cause blurred edges and reduced contrast,image enhancement algorithms are explored,and the histogram equalization is used for tactile image enhancement.3.The super-resolution reconstruction of the sequential tactile images needs to estimate the sub-pixel displacement first.In view of the problem that the motion estimation algorithm in the field of digital image cannot be fully applied to tactile image,a motion estimation algorithm suitable for sequential tactile images is proposed,namely the ARTC(Adaptive Reference Template Cross-correlation)algorithm.In the ARTC algorithm,the appropriate reference template in the reference tactile matrix is adaptively selected first,then the correlation search algorithm is used to search for the most similar subregion in the tactile matrix to be estimated,and the whole pixel displacement is obtained.Finally,the surface fitting method is used to estimate the subpixel displacement.The experimental results show that the displacement estimation results of the ARTC method is superior to the results of phase correlation algorithm,affine algorithm,and SIFT algorithm,when the resolution of the sequential tactile images is 8×8.4.Based on the existing 16×16 resistive sensor array in the laboratory,a super-resolution reconstruction system of tactile image based on mixed programming by MFC and MATLAB is developed,which can be used to realize super-resolution reconstruction of single and sequential tactile images.Experimental results show that the designed super-resolution reconstruction system can significantly increase the resolution of low-resolution tactile image,and can improve its contrast and sharpness.