Every day, motorists navigate busy roads where split-second decisions can mean the difference between a safe journey and a tragic collision. Among all road users, motorcyclists face unique hazards because their smaller physical profile makes them far more difficult for other drivers and automated traffic surveillance cameras to spot. In fact, a significant number of traffic accidents occur simply because of human interpretation failures, where drivers fail to notice an unexpected object located right in plain sight. Furthermore, teaching automated computer systems to recognize motorcycles is surprisingly complex compared to detecting larger, more uniform vehicles like cars and trucks. To tackle this vital road safety challenge, a team of dedicated engineers and scientists at the Institute of Nanoscience and Nanotechnology within Universiti Putra Malaysia embarked on an innovative study. Led by researcher Mohd Ali Mat Nong alongside his collaborative team, the institution set out to evaluate and improve motorcycle image detection using a sophisticated digital technique known as the contrast stretching method. By implementing this technology on specialized hardware platforms, the Universiti Putra Malaysia researchers aim to ensure that digital cameras can clearly see motorcycles under varying weather conditions, ultimately laying the groundwork for smarter and safer roadways.

Teaching a machine to detect a moving motorcycle requires transforming raw light captures into structured digital data that a computer processor can analyze and interpret. Digital image processing generally involves three fundamental steps: importing the visual image through acquisition camera tools, analyzing and manipulating those visual signals, and generating an enhanced output image or reporting feature data based on that analysis. To accomplish this for motorcycle tracking, the research team designed a specialized engineering workflow utilizing simulation software and hardware blocks. Their system begins with a demosaic block, which acts as a color filter targeting raw motorcycle images captured in a specialized camera format known as the Bayer pattern. Next, the signal flows into a color space conversion block that identifies the fundamental three-dimensional color system composed of red, green, and blue light. Each of these basic colors carries an intensity value ranging from zero up to two hundred and fifty-five, and their unique combinations produce all visible color properties. For instance, combining the maximum red, green, and blue values of two hundred and fifty-five generates pure white, whereas a value of zero across all three produces complete blackness. Once the colors are organized, a Sobel edge detector steps in to identify the physical outline of the motorcycle. This edge detection tool classifies the motorcycle image into a Boolean matrix using binary format, representing visual boundaries with simple numerical values of one for white edges and zero for black backgrounds. Finally, a live video display block allows researchers to view the targeted motorcycle in real time for accurate engineering analysis.

While capturing the physical outline of a motorcycle is a critical first step, raw camera footage often suffers from poor lighting, deep shadows, or washed-out backgrounds that hide important visual details. To overcome these visual flaws, the Universiti Putra Malaysia team applied a mathematical technique known as contrast stretching, which engineers also refer to as image normalization. The primary objective of image enhancement is to modify a picture so that the processed version is significantly clearer and more intelligible to both human viewers and machine detection algorithms than the unprocessed original. Contrast stretching achieves this by expanding the dynamic range of brightness levels found within a digital picture. In a low-contrast image, the individual picture elements, known as pixels, tend to clump together tightly within a narrow tonal scale. When pixels cluster in this manner, visual distinction is lost, making a dark motorcycle blend invisibly into a dark asphalt road. The contrast stretching algorithm takes these clustered pixel values and spreads them out across the entire available tonal scale through grayscale transformation functions. To evaluate how well the pixel intensities are distributed, the researchers analyze digital graphs called histograms. A histogram displays the total number of pixels corresponding to each tonal variation, dividing the image into three distinct zones: the dark area, the mid-tone area, and the light area. By utilizing histogram equalization techniques, the research team can objectively measure and verify image enhancement without losing valuable visual information.

To prove that their hardware platform could perform reliably in the real world, the researchers evaluated motorcycle images across completely different environmental lighting situations. Specifically, they recorded and processed motorcycle visibility under normal daylight conditions as well as challenging, gloomy rainy daylight conditions. Rain and overcast skies introduce severe visual interference, scattering light and blurring the sharp boundaries needed for accurate vehicle recognition. During their experiments, the scientific team cropped the targeted motorcycle images to establish a specific region of interest for detailed analysis. They then scaled and normalized these regions to ensure that every pixel mapped correctly regardless of whether the camera used high- or low-resolution settings. To find the exact setting that produces the clearest picture, the researchers tested various threshold levels ranging from level two up to level ten. At each threshold setting, they carefully observed how the pixel intensities behaved inside the dark, mid-tone, and light histogram zones. Their evaluation revealed striking differences between low and high threshold settings. At threshold level two, the total number of useful pixels dropped, and the image exhibited lower overall intensity. The pixels in low-quality images were scattered randomly across the scale and distributed inconsistently, which blurred the motorcycle outlines. Conversely, when the system operated at threshold level ten, the pixel intensities stabilized and formed a uniform, well-tabulated distribution across the tonal scale.

The experimental results gathered by the Universiti Putra Malaysia researchers conclusively demonstrate that contrast stretching dramatically improves motorcycle visibility on digital screens. In both clear daylight and difficult rainy weather conditions, output images processed at threshold level ten showed superior visual quality, sharper edge accuracy, and brighter tone variations compared to those processed at lower threshold levels. Because the pixel data was distributed uniformly, digital bounding boxes were able to lock onto and detect the motorcycle candidates successfully regardless of atmospheric interference. Furthermore, the study confirmed that feeding high-resolution images into the contrast stretching platform consistently yielded better detection performance than using low-resolution inputs. This research represents a significant technical leap forward for intelligent transportation engineering. By refining how machines process colour, edges, and contrast on hardware platforms, the Institute of Advanced Technology has provided a reliable blueprint for next-generation traffic monitoring and automated driving systems. As automated vehicles and smart traffic cameras continue to evolve, the contrast stretching methods validated at Universiti Putra Malaysia will play a crucial role in preventing road accidents, helping ensure that motorcyclists remain visible and protected on every journey.

Source: https://ieeexplore.ieee.org/document/9703892

Prepared by:
Dr. Ismayadi Ismail (NSCL)

Date of Input: 29/07/2026 | Updated: 29/07/2026 | roslina_ar