The temperature of metal surfaces at normal temperatures under sunlight can be measured by combining visible images with machine learning (ML). But this method is susceptible to illumination conditions, which causes large errors. To solve this problem, a method of temperature difference measurement based on visible images’ differential chromatic features (DCFs) between two metal surfaces, which are at different temperatures but the same illumination, is proposed in this paper. Through experiments, temperature difference image libraries of brass surfaces under varying illumination are established. DCFs of each image are extracted, and temperature difference between two surfaces is used as the label. Models that can predict the temperature difference between two surfaces are trained by ML. Prediction results show that the mean absolute error of this temperature difference measurement method can be as low as 1.6, decreasing by 42% at most compared with those of simply measuring the surface temperature. The effect of temperature and illumination on visible images’ gray levels is analyzed. And it is concluded that the introduction of a temperature-constant surface provides a reference item for illumination, which helps to weaken the influence of illumination changes and filter out the interference caused by them. This is why the prediction accuracy can be improved by the temperature difference measurement method.
This article proposes an improved hardware implementation method for phase-based frequency estimation algorithm. Compared with the original algorithm, this paper uses spatial Gabor filtering instead of frequency domain Gabor filtering to avoid complex spatial frequency domain conversion and reduce resource requirements for hardware implementation. The entire algorithm is designed as a pixel stream processing architecture with pipeline blocks, enabling it to easily adapt to common image sensing components with streaming pixel output, while achieving higher performance at lower system costs. This article implements a prototype based on Field Programmable Gate Array (FPGA), which can process up to 100M pixels per second and calculating vibration frequency in real-time.