A method and device for counting passenger flow based on tof camera

Abstract

The invention discloses a TOF camera-based passenger flow counting method and a device. The method comprises the following steps of 1, acquiring a depth image; 2, screening a detection region througha height threshold value to obtain a mask image; 3, performing morphological treatment on the mask image; 4, acquiring a detection candidate region of a passenger head in a color image according to the mask image; 5, positioning the passenger head in the candidate region by adopting a depth neural network model; 6, tracking by KCF through combining Kalman filtering; 7, counting the passenger flow.According to the invention, not only the problem that the pattern recognition function cannot be realized by adopting a depth image is solved, but also invalid regions in the image are filtered through the depth image. Therefore, the detection speed and the detection accuracy are accelerated. In conclusion, the purpose of quickly and accurately counting the passenger flow can be achieved.

Description

technical field [0001] The invention relates to the field of computer vision, in particular to a method and device for counting passenger flow based on a TOF camera. Background technique [0002] Traditional bus passenger flow counting methods, including infrared devices and pressure sensors, have large statistical errors and have been gradually abandoned. The direction of the current bus passenger flow counting method is developing towards the field of image processing. The bus passenger flow statistics are based on ordinary color or black and white images, generally combined with a deep neural network. Such complex scenes have the disadvantages of high computational complexity and failure of pattern recognition will result in over-counting and missing counting; bus passenger flow statistics based on depth images has the advantage of low computational complexity. Simple scenes have high accuracy. The disadvantage is that because the depth image loses too much texture infor...

AI Technical Summary

Problems solved by technology

The direction of the current bus passenger flow counting method is developing towards the field of image processing. The bus passenger flow statistics are based on ordinary color or black and white images, generally combined with a deep neural network. Such complex scenes have the disadvantages of high computational complexity and failure of pattern recognition will result in over-counting and missing counting; bus passenger flow statistics based on depth images has the advantage of low computational complexity. Simple scenes have high accuracy. The disadvantage is that because the depth image loses too much texture information, pattern recognition cannot be used directly on the depth image. In complex scenes, it is easy to over-count and miss-count, and some depth cameras detect depth based on infrared reflection. Yes, objects with smooth surfaces (such as human hair or mobile phones) will produce specular reflection of infrared, and the distance cannot be measured, thus affecting the counting results

Images