Journal of Modern Classical Physics & Quantum Neuroscience

Abstract

The relevance of this research stems from the limitations in bandwidth and energy efficiency inherent in traditional electronic systems for industrial computer vision. This work substantiates the use of spiking emission dynamics in semiconductor quantum dots (InAs/GaAs) as a physical foundation for an optical neuromorphic system. We experimentally observed a transition of photoluminescence into a pulsed generation mode upon exceeding a threshold pump density(P_th≈1.12 kW/cm^2 ) The duration of an optical spike was determined to be 135±12 пs, with an inter-spike interval jitter not exceeding 12 ps within a temperature range of 293–313 К. kinetic model was developed, linking Auger recombination parameters to the neuron's excitability threshold. The system was demonstrated to perform processing tasks on industrial test images—specifically defect detection and classification-with an accuracy of 93,7–96,2 % and a latency of 280–410 mu s. The energy consumption was 10,8–15,1 nJ/frame, which is 3–8 times lower than that of electronic neuromorphic platforms (such as Intel Loihi) or GPU clusters. The proposed architecture enables matrix-vector multiplication within the optical domain, eliminating the need for intermediate analog-to-digital conversion; this ensures high robustness against electromagnetic interference and suitability for deployment in real-world industrial production environments.