Top 3 People Counting Technologies

Structured light People Counter
Advantages
  • The scheme is mature, and the baseline of the camera can be relatively small, which is convenient for miniaturization.
  • Low resource consumption, single frame IR image can calculate the depth map, low power consumption.
  •  Active light source can also be used at night.
  • In a certain range, high precision, high resolution, resolution up to 1280 x 1024, frame rate up to 60 FPS
Disadvantages
  • It is easy to be disturbed by the environment and sunlight, and the outdoor experience is poor.
  • With the increase of detection distance, the accuracy will become worse.
Binocular People Counter
Advantages
  • Low hardware requirements and low cost. Ordinary CMOS camera is enough.
  • Stereo imaging, high precision.
  • Both indoor and outdoor. Not affected by strong light or weak light.
  • It can be installed in the position of 180-300cm.
  • Detection angle up to 130 °。
Disadvantages
  • There are strict requirements for the installation height, and different lenses may be needed to match for different heights. It is not suitable for monotonous and texturless scenes. Binocular vision image matching based on visual features, no features will lead to matching failure.
  • High computational complexity. This method is a pure vision method, which requires high algorithm and large amount of calculation.
  • The line limits the measurement range. The measurement range is proportional to the baseline (the distance between two cameras), which makes it impossible to miniaturize.
TOF People Counter
Advantages
  • The detection distance is long. In the case of enough laser energy, it can reach tens of meters.
  • The detection accuracy is high.
  • The interference of ambient light is relatively small.
Disadvantages
  • High requirements for equipment, especially time measurement module. So it costs twice as much as binocular
  • It consumes a lot of resources. In this scheme, multiple sampling and integration are needed to detect phase offset, and the computation is large.
  • The center detection precision is high and the edge precision is low.
  • Limited to resource consumption and filtering, frame rate and resolution can not be high.
  • The detection distance should be more than 1m
  • The detection angle is less than 60 °