THE DEVELOPMENT OF IMAGE PROCESSING METHOD ON THE AUTOMATIC READING SYSTEM OF THE GLASS THERMOMETER USING A DIGITAL CAMERA

The reading performance of an analog thermometer, Liquid in Glass Thermometer (LiGT), can be improved using a digital camera. The aim is to minimize the human error on the reading of LiGT and increase the accuracy of temperature measurement results. In order to achieve an accurate result, a robust image processing method is required in the measurement. In this work, the LiGT image generated using a digital camera is analyzed using the technique in HSV color space which consists of some image processing methods (e.g., thresholding, morphology filter). The type of LiGT used is the glass thermometer with the colored liquid. There are three main parts to this developed technique process, i.e., identifying the scale of LiGT to calculate the pixel per temperature unit value (ppt), segmentation of the liquid column, and calculate the temperature based on the ppt value. Through simulation with a synthetic image, we demonstrate that the developed technique in this work has successfully read (measured) the temperature value of the LiGT (having a scale unit of 1C) with a measurement error of 0.04C. In the experimental results, we also report the developed technique performed on a real image of LiGT.


INTRODUCTION
Glass thermometer/Liquid in Glass Thermometer (LiGT) is an analog temperature measuring instrument, which works based on the principle of the thermometric properties of the liquid column in the glass rod whose volume (the liquid column height) is proportional to the temperature value [1,2]. This type of temperature measuring instrument is one of the earliest forms of the thermometer, whose measurement range capabilities can reach the range between -170C and 330C [3]. Currently, LiGT is still widely used, including in the fields of industry, health, and research laboratories. The Glass thermometer is easy to use. In terms of reading, it can be performed with the naked eye without requiring other auxiliary equipment. LiGT material is also made of glass, which is inert that it does not react to chemicals. In addition, the price of LiGT is relatively inexpensive, but still has good long-term stability at the same level of uncertainty [4].
Because the glass thermometer is an analog instrument, its reading is very simple visually, which can be performed with the naked eye or with the aid of a magnifying glass [5]. In terms of construction, LiGT does not have a communication interface with a computer, so that the reading can only be done manually. It is not possible to directly read the temperature measurement of LiGT automatically. The choice of the reading method is one of the most influential contributions to the LiGT calibration performance.
In metrology activities, the calibration of measuring instruments, including LiGT, has a significant role in maintaining the traceability of temperature measurements (traceability). Based on several literature sources, it is known that in addition to repeatability, a significant contribution of the uncertainty sources in the LiGT calibration is the scale division and the method of reading the scale of the LiGT [4,5]. The results of reading LiGT with the naked eye can be different from each other, depending on the ability of the calibration technician (subjective). The observation results of some technicians in reading the LiGT scale can reach up to ½, ⅓, or ¼ of a scale division. In order to increase readability, the magnification devices, such as a magnifier, can be used to allow interpolation up to smaller-scale divisions [4,6,7,8].
Several factors, such as the technician's ability, the parallax error, and the fatigue on reading, can influence the uncertainty of the reading of the scale on LiGT. This certainly affects the accuracy and the precision of temperature measurement results. In order to minimize the possibility of reading errors, in this study, a vision system for reading the LiGT analog measuring instrument is developed. The aim is to improve the performance of temperature calibration in terms of simplification of measuring instrument readings and to increase the objectivity of the LiGT reading.

DESIGN OF VISION SYSTEM
The vision system includes digital image processing technology generated from a digital camera using a computer, which is then processed to obtain specific information [9,10]. Since the nature of the LiGT temperature reading is visual, the automation method of reading the | 3

SPEKTRA: Jurnal Fisika dan Aplikasinya Volume 6 Issue 1, April 2021
LiGT scale can be performed using a digital camera, as shown in FIGURE 1. In FIGURE 1, it can be shown that when the temperature measurement using the LiGT was taking place, the LiGT image data were acquired using a digital camera. Then, it is needed an image processing and analysis of the image generated by the digital camera to calculate the temperature using a computer. In this study, the Matlab program was used to process the data.
The performance of the developed vision system depends on the optical capabilities of the digital camera and the design of the image acquisition system. The optimal design of the vision system prototype can make it easier to develop the image processing method required for the LiGT scale reading vision system. Meanwhile, the image processing method relies heavily on the performance of the segmentation process in identifying the scale line and liquid column in the glass rod.

Image Processing and Analysis of LiGT Scale Reading
In order to obtain precise and accurate results, it requires a robust image processing method in this measurement system. There are three main processes in the developed method, i.e., scale line and liquid column segmentation, identifying the LiGT scale to calculate pixels per temperature unit value (pixels per temperature -ppt), and calculating temperature based on the ppt value. In the image processing process in the developed method, the LiGT image is processed in the HSV (hue saturation value) color space technique for segmentation purposes for identifying the scale line and liquid column in a glass rod [11,12]. shown in FIGURE 2, is carried out to make it easier to separate the object (scale line and liquid column on the glass rod) from the background. The segmentation process begins with converting the image format from a color image to a grayscale image, then followed by converting to be a binary image (black and white). In this developed image processing method, the morphological techniques, i.e. closing and dilation, will also be used to improve the quality of segmentation results, both in grayscale and binary images [13,14]. After the image segmentation in identifying the position of the liquid column height to the scale line has been performed, then the LiGT temperature is calculated based on the ppt value as the used standard.

SIMULATION RESULT
As an initial step to test the performance of the developed image processing method, as described in section 2.2, a simulation is carried out using a synthetic LiGT image (FIGURE 3). By using the naked eye, the temperature value of the LiGT synthetic image can be read in the range of 22C.   study. Based on the processed synthetic image, the following LiGT temperature reading is obtained.
of LiGT = −3°C + (24 × 1°) + 0.0385 × (30 − 5)°= 21.96° By the limitation of reading LiGT with the naked eye, the synthetic image in this simulation is read at 22C, a difference of 0.04C with the reading using the developed method (21.96C). The difference probably comes from the uncertainty of the top position of the liquid column that coincides with its scale line, which is about 4 pixels wide (equivalent to 0.16C).

EKSPERIMENT RESULT
After being performed on a synthetic image, the developed method is then tested using a real LiGT image generated from the digital camera. In this study, 3 (three) different LiGT samples were used to test the robustness of the performance of this method. These three LiGT samples have the same scale unit, which is 1C. However, the image quality characteristics of them vary on the sides of the column or scale line. The LiGT-1 image in FIGURE 6 shows the quality of the scale line object is very good, but in the case of the column line object at a value between 6C and 8C, there is a column line discontinuity. This is clearly seen in the grayscale image segmentation results. On the other hand, in the LiGT-2 image, as shown in FIGURE 7, the column line object is very good, but the scale line object is not well (looks thin and unclear). The quality of the column line and scale line objects in the LiGT-3 image (FIGURE 8) is not clearer than LiGT-1 and LiGT-2.

CONCLUSION
In this research, image processing and analysis methods have been developed for scale line reading on an analog temperature measuring instrument, LiGT. Through a simulation using a synthetic image, it is shown that the developed technique has succeeded in reading the temperature value of LiGT (which has a 1C scale unit) with a measurement error of 0.04C. This deviation probably comes from the uncertainty of the top position of the liquid column coinciding with the scale line. Then in the experimental results using some real LiGT images, it was found that the developed method can be applied to the real images (three different LiGT samples). The used morphological techniques (closing and dilation) in this method effectively improve the reliability of the LiGT reading performance. This technique can improve the poor segmentation results due to an unclear segment of the column or scale line object (looks thin) on the old LiGT.