Direction Analysis
One of the most commonly used methods for analyzing the direction of target movement is the optical flow method, which approximates the movement of the target through the optical flow in two adjacent frames of images. The optical flow method is more suitable for estimating the target movement trend in a short period of time (such as several adjacent frames), and the optical flow method is very sensitive to image noise, as shown in the figure below, which is the calculation result of the two optical flows of the video in the experiment.
Due to the complex and noisy environment in the video, and the irregular motion of the target in two adjacent frames of images, the effect of optical flow analysis is not ideal.
The motion analysis method based on the motion history graph is more suitable for estimating the motion trend of the target in a long period of time. This method accumulates the movement of the target over a period of time, and estimates the direction of the target's movement by calculating the gradient in the time domain.
The algorithm in the experiment is as follows:
Step1: Calculate the Sobel edge map of the detection area (the Sobel operator is simple and the calculation speed is fast). Since only the speed of the target's uplink and downlink is considered, only the gradient of the image in the vertical direction is calculated. The kernel is as follows:
4 Correlation and statistics
Because the movement duration of the pedestrian head target in the detection area is short, the displacement is small, and the occlusion is less, so using the tracking gate to associate the target has a better effect in the experiment.
The feature description of each head model is: {x, y, dx, dy, width, height, score, scoreperfrm , scoremax, frame, startfrm }, explained as follows:
------------- ----------------------
https://blog.51cto.com/u_2847568/4919182