Smart Colorimetric Temperature Sensor with Thermoresponsive Polymers and Automatic Image-Based Identification

Temperature monitoring is essential across multiple sectors, including healthcare, logistics, and safety. Achieving accurate measurement under diverse conditions in a cost-effective, and sustainable manner remains a key challenge for both the scientific community and industry. In this context, we present the development and characterization of colorimetric sensors based on a polydimethylsiloxane (PDMS) matrix that incorporates commercial, food-grade, organic thermochromic pigments, thus enabling temperature sensing without electronic components. These sensors consist of six distinct modules containing pigments with different activation temperatures to enable continuous optical monitoring over a broad thermal range. The colorimetric response of each module was quantitatively analyzed as a function of temperature using image-based processing techniques. The results demonstrate good thermal reversibility and stable, reproducible behavior over multiple thermal cycles and under different ambient conditions. Once the colorimetric response of the sensor was characterized, we implemented a precise color quantification method through digital image processing to overcome the purely qualitative nature of traditional thermochromic materials. Moreover, the application uses a random forest model with a high-precision correlation between colorimetric data and temperature, thus confirming sensor reliability. Overall, the results highlight the system's potential as a scalable, electronics-free solution for temperature monitoring in diverse applications, such as smart packaging or intelligent surfaces. • Fully passive colorimetric temperature sensors using PDMS and thermochromic pigments for broad-range thermal monitoring are presented. • Six thermochromic pigments with different activation temperatures are used to fabricate a modular sensor design, enabling continuous, reversible, and reproducible color changes across multiple thermal cycles. • An automated image-based color quantification with a random forest regression model is also presented, ensuring accurate and scalable temperature readings without subjective interpretation.