<p>Computer vision algorithms for indoor mobile robot navigation often face challenges under suboptimal or uneven lighting, leading to reduced contrast, colour distortion, and image noise that compromise visual measurements. Existing literature has proposed positioning methods using ceiling-facing cameras to address these issues. However, these approaches often overlook illumination interference from ceiling-mounted lamps, which causes local brightness variations and observation uncertainties. As a result, localization accuracy decreases and error rates increase, requiring advanced sensors or computationally intensive image processing, thereby limiting their practicality in real-world applications. To address these challenges, this paper proposes a novel technique for positioning indoor mobile robots based on identifying 2D data from images captured in unevenly lit environments. Firstly, the positioning method using 2D data image identification is presented. Secondly, the proposed reference object identification model is based on a corresponding threshold adjustment to mitigate illumination noise. Consequently, it provides an adaptive threshold to match data accurately to minimize positioning errors for robots operating in indoor environments. Lastly, experiments are conducted in our laboratory using a wheeled mobile robot prototype equipped with a ceiling-facing CMOS camera, operating in a long corridor under uneven lighting conditions, to validate the effectiveness of the proposed novel technique. Experimental results show a position error of less than 10&#xa0;cm for a 5500&#xa0;cm trajectory. This ensures that the proposed method enhances the self-locating ability of mobile robots using computer vision, offers the advantage of cost reduction, and eliminates the need for sophisticated sensors or intricate image processing techniques.</p>