Abstract—In industrial factories, many measuring instruments are used to display, for instance, pressure, voltage, temperature or humidity. Human errors are the main problem and often occur in many processes mostly done manually, such as data acquisition. Therefore, the problem of how we obtain such data automatically and correctly in real-time is important. In this paper, a numeral recognition system (NRS) is proposed based on an optical character recognition (OCR) method. The NRS embedded industrial Internet of things (IIoT) is used to serve a real-time service. Moreover, digital image processing (DIP) together with the multi-layer perceptron (MLP) is applied to efficiently recognize the numeral data. Furthermore, it is very common that the instruments' screens can face the rotation problem. This problem can be solved using the histogram of oriented gradients (HOG) and Hough transform (HT) techniques. In addition, realistic conditions under various noise types are considered such as salt and pepper (SP) noise, Gaussian noise, and Speckle noise. The system performances are evaluated in terms of confusion matrices and accuracies. The strong contribution of our proposed NRS system is that it works excellently in any situations and achieve up to 95.13 percent accuracy. From the actual experiments, we achieve an average about 95 percent accuracy. Although the NRS with the HOG and HT technique takes a bit longer computation time and more memory usage to process the images than another NRS, the system provides better results. Our proposed system is suitable for a real-time service due to low computation time.