A Temperature Monitoring Method for Sensor Arrays Based on Temperature Mapping and Improved Mask R-CNN

Temperature monitoring of sensor arrays is indispensable for ensuring the stable operation of the entire sensor system. This article presents a novel method (TISM) for sensor array temperature monitoring based on temperature mapping and an enhanced Mask region-convolutional neural network (R-CNN) framework. Initially, the method establishes a robust mapping correlation between sensor temperature data and spatial coordinates, thereby facilitating precise data acquisition and strategic rule formulation through a temperature qualification protocol. Subsequently, employing subarray analysis, the temperature data are structured into a matrix and transformed into a temperature heat map. The thermal image is further refined using interpolation techniques to enhance the accuracy and stability of the monitoring system. Additionally, an improved Mask R-CNN model is proposed, enabling effective target recognition and feature extraction from the temperature thermogram, thereby facilitating the extraction of temperature state information. Ultimately, sensor temperature states are determined based on color discrepancy and temperature mapping, thus achieving the objective of sensor array temperature monitoring. The method was compared with artificial neural network temperature prediction (ANNTM), phase-shifted grating, and photoelectric oscillation temperature monitoring (MPTM). Comparison indicators include comprehensive temperature prediction effect, accuracy, stability, and monitoring range. Notably, the proposed method attains a prediction accuracy of 97.13%, showcasing substantial improvements over ANNTM in terms of mean deviation and standard deviation by 25.89% and 1.91%, respectively. Furthermore, compared to MPTM’s limited monitoring range of 490 °C-495 °C, the proposed method offers a significantly broader monitoring scope. Moreover, in terms of integrated temperature prediction for the sensor array, the proposed approach exhibits superior performance with smaller prediction errors, closely aligning with actual temperature values. Experimental validation corroborates the effectiveness of the proposed method, thereby underscoring its promising potential for real-time temperature monitoring of sensor arrays in practical applications.