Fig. 5: Pattern recognition and target location.

a Optical image of the 6 × 6 array. The array was divided into 3 × 3 pixels, with each pixel containing 4 subpixels. The array acted as a single-layer perceptron, where the output currents I₁, I₂, I₃, and I₄, were the total sum of the photocurrents generated by the corresponding subpixels under the respective incident light intensities. b Schematic illustration of the optical setup and circuit diagram of the custom printed circuit board for pattern recognition tasks. When different letters were projected onto the device array through masks, the corresponding output LEDs lit up to display the recognition results. c Schematic diagram of input letter images ‘D’, ‘L’, ‘K’, and ‘J’ with Gaussian noise standard deviations varying from 0.1 to 1. d Noise-level-dependent recognition accuracy histogram. e Target location and discharge mode classification of corona discharge in power transmission lines. f Partial results of the target location. The red dots represent the experimental output, and the black line represents the ideal location results. g Classification accuracy of different discharge modes.