Evaluation Metrics :-

Qualitative Precision (QP): In general, Precision measures the accuracy of the predicted position of the tracked object. QP is a performance metric specifically designed to evaluate object-tracking models operating in low-light conditions (nighttime). It combines two key factors: 

• Image Quality Assessment (IQA): This metric assesses the visibility of individual frames within the video sequence. It typically outputs a score ranging from 0 (perfect visibility) to 1 (worst visibility). 

• Center Location Error (CLE): This metric calculates the Euclidean distance between the center of the ground truth object (actual location) and the center of the predicted bounding box by the tracking model.

where xgti and ygti are the coordinates of the center of the ground truth bounding box, and xpredi and xpredi are the coordinates of the center of the predicted bounding box. 

A frame is considered a "positive detection" when the combined effect of low visibility and tracking error falls below a predefined threshold. This threshold considers both the IQA score and the CLE value. Mathematically, QP is expressed as: 

The number of positive frames is the count of frames where the product of IQA and CLE falls below the specified threshold (typically a distance threshold (Th) in pixels). This is mathematically represented as:

A threshold value of TH = 15 has been chosen, after careful inspection of the dataset.

QP is a performance metric used to assess the effectiveness of an object tracking model at night by combining the quality of a video frame and the CLE in object tracking at that particular frame. The value of QP will vary between zero and unity. The unit value of QP indicates that the tracker algorithm is completely successful in tracking the object, whereas the zero value of QP indicates that the algorithm failed to track the object. This metric provides a more comprehensive evaluation of tracking performance than precision alone, as it considers how the visibility of the frame degrades at night.