Hi ShengJu,
The 'MergeThreshold' parameter defines the criteria needed to declare a final detection by requiring multiple overlapping detections in an area. Increasing this threshold can suppress false positives by ensuring the target object is detected multiple times during the detection phase. Setting threshold to 0 returns all detections without merging or thresholding.
The effectiveness of specific 'MergeThreshold' values, such as 24, 128, or 6, can vary based on image quality, object sizes as well as the pre-trained model's characteristics. Different objects and image conditions might require different thresholds to achieve the best balance between detection accuracy and false positive suppression. Since the property is tunable you can experiment with different threshold values to suit your use-case.
For more information regarding Cascade object detector refer to the below MathWorks documentation:
cascadeobjectdetector: www.mathworks.com/help/vision/ref/vision.cascadeobjectdetector-system-object.html
