Multi-channel compressive sensing-based object recognition

A method for capturing an image using compressive sensing, the method comprising:

• capturing an image and projecting the image onto a digital micromirror device (DMD) array comprising a plurality of micromirrors;
• controlling the DMD array to generate a first set of samples and a second set of samples of the image using a sampling rate lower than that required by the Shannon-Nyquist sampling theorem;
• projecting spatial information about the first set of samples of the image from the DMD array onto a first detector array through a first optical channel;
• projecting spectral and polarization information about the second set of samples of the image from the DMD array onto a second detector array through a second optical channel;
• and
• reconstructing the images using the samples from the first and second channels.

An optical system for capturing an image using compressive sensing comprising:

• a digital micromirror device (DMD) array comprising a plurality of micromirrors;
• means for capturing an image and projecting the image onto the DMD array;
• a first means for image detection;
• a first optical channel including a first means for projecting spatial information about the image from the DMD array onto the first means for image detection;
• a second means for image detection;
• a second optical channel including a second imaging means, means for capturing spectral information and means for capturing polarization information and for projecting the spectral and polarization information about the image from the DMD array onto the s second means for image detection;
• means for controlling on/off states of the micromirrors in the DMD array to direct a first sample of the image from the DMD array to the first optical channel and direct a second sample of the image from the DMD array to the second optical channel;
• means for controlling the DMD array to generate the first set of samples and the second set of samples of the image using a sampling rate lower than required by the Shannon-Nyquist sampling theorem, and
• means for reconstructing the image by applying inverse compressive sensing based on Hadamard approach.