Abstract: As a vital public safety application,the crowd counting system plays a crucial role in assisting decision-makers to better manage and monitor crowd density,thereby preventing safety hazards caused by excessive crowd congestion. However,existing systems has some limitations such as weak environmental adaptability,inadequate accuracy,and inability to respond in real-time. To select the optimal solution,TRIZ theory is applied to analyze the design contradictions in the development of a crowd density estimation system. The counting system was designed using the Haar Cascade classifier from OpenCV and Python language to realize real-time statistics of changes in the number of people in videos.

Key words: TRIZ theory, crowd counting system, stampede accident