Transparent object detection using single-pixel imaging and compressive sensing

Opaque object detection and image acquisition are perfectly done with the techniques collecting reflections from the object. However, transparent object detection is a challenge due to the complex nature of light interacting with it. The characteristics of light such as absorption, transmission, reflection, and refraction in transparent objects makes its acquisition challenging. Single-pixel imaging (SPI), which uses a single pixel sensor to capture the image, is an evolving technology. The proposed study focuses on designing a 2D single-pixel transparent object inspection system which utilizes compressive sensing and 11 minimization approach. A novel camera architecture based on Digital Micromirror Device (DMD) along with Compressive Sensing (CS) algorithm is adopted for image reconstruction. Two-dimensional image reconstruction of transparent objects is performed by collecting transmitted light intensity from the object with a single-pixel detector. With the CS algorithm, good quality images are reconstructed with few measurements in contrast to large data required for Nyquist criteria. The digitized input for CS algorithm is based on the inner product between the transparent object and a set of random patterns projected from the DMD. The transparent object detection was achieved using only 30% of the total pixels in the image with the reconstructed images showing around 60 percent similarity to the real object. Our experimental setup has been compared to a conventional imaging system to prove its efficiency in obtaining accurate results. Furthermore, our technique is nondestructive, does not require raster scanning, not time-gated detection and benefits from compressive sensing.