Traditional capillary melting point meters rely on human observation, and there is subjective error in determining the timing of initial melting (appearance of the first droplet of liquid) and full melting (disappearance of solid). Based on image processing algorithms, automatic recognition technology achieves objectivity and precision in measurement by accurately segmenting key visual features during the melting process. The core algorithm process is as follows:

1. Image preprocessing and sample area localization:

The algorithm first preprocesses the collected video sequence, including Gaussian filtering denoising and contrast enhancement, to highlight the samples inside the capillary. Subsequently, edge detection or template matching algorithms are used to accurately locate and segment fixed capillary regions (ROIs), eliminate background interference, and lay the foundation for subsequent analysis.

2. Accurate detection of initial melting point - based on texture and shape mutations:

This is the first key to segmentation. Solid powders have specific texture characteristics in the initial stage. The algorithm monitors changes in ROI by calculating its grayscale statistical features (such as standard deviation) or texture features (such as entropy based on grayscale co-occurrence matrix). When initial melting occurs, droplets appear in local areas, causing a sudden change in texture uniformity, and the above characteristic values will undergo significant transitions. By combining the frame difference method to detect local change areas, the corresponding moment is determined as the initial melting point. This step overcomes the disadvantage of delayed response of the human eye to subtle initial melting phenomena.

3. Accurate determination of full melting point - based on contour and transparency analysis:

This is the second key to segmentation. As melting progresses, the solid contour continues to shrink. The algorithm accurately depicts the shape of solid columns through edge extraction and contour tracking (such as Canny operator and active contour model). The criterion for complete melting is the disappearance of the solid contour. More advanced algorithms simultaneously monitor the average grayscale or transmittance of the entire ROI. At the moment when the solid disappears, the transparency at the bottom of the capillary increases sharply, resulting in a step change in the overall grayscale value. The algorithm determines the moment corresponding to this inflection point as the full melting point, avoiding the uncertainty of the human eye's judgment of the moment when the last small particle disappears.

Conclusion: By converting the visual information of the melting process into quantifiable texture, contour, and grayscale features, image processing algorithms have achieved accurate and automatic segmentation of the two physical stages of initial melting and full melting. This study not only significantly improves the repeatability and accuracy of melting point measurements, but also provides a solid foundation for the digitization and intelligent analysis of experimental data.