Human Head Detection Based on Coefficient of Roundness and Ratio of Continuous Curvature
Yanwu Hong, Anton Louise de Ocampo · 2024
This article presents a new approach for human head detection that can well balance the detection accuracy and real-time performance based on calculating the coefficient of roundness (COR) and ratio of continuous curvature (ROCC). Firstly with the Xor operation between background image and foreground image, we can get the edge of the foreground image. Then through the opening operation of morphology and removing the minimum area under 350, the connected regions of the foreground image can be gained. According to the coordinates of boundary points of each region, we can calculate the COR and ROCC of each region to determine the possibility of being a human head. Set the threshold values of COR and ROCC as 0.4 and 9 to test the accuracy of the method proposed in this paper. The experiment shows that the technique can reduce the interference regions greatly. The final results show that the accuracy of detecting the human head with the method in this paper can be 92% and the average time taken to calculate on a regular laptop is only about 0.35s. In conclusion, the method proposed in this paper can be used to detect human heads in real time to estimate the crowd distance.