为了更加有效地提取枪口火焰边缘,设计了一种基于融合三次B样条小波变换和Canny算法的边缘检测算法。首先,取三次B样条小波函数对得到的火焰图像进行卷积运算;其次,针对Canny算法存在的问题进行改进,将梯度模板改用Scharr算子在原有方向的基础上增加两个方向,通过引入图像复杂度改进的Otsu算法选取高低阈值;最后,通过一定的融合规则获取高低频分量,进行小波逆变换,重构图像获取最终边缘图像。试验结果表明,该融合优化算法可以准确定位边缘,有效抑制噪声,火焰边缘检测结果完整清晰,客观评价优于单一算法结果。

In order to extract muzzle flame edge information more efficiently, an edge detection algorithm combining cubic B-spline wavelet transform and Canny algorithm was designed. Firstly, the cubic B-spline wavelet function is convolved with the flame image. Secondly, according to the problems existing in Canny algorithm, the gradient template was changed to Scharr operator to add two directions on the basis of the original direction, and the Otsu algorithm with improved image complexity was introduced to select high and low thresholds. Finally, high and low frequency components were obtained by a certain fusion rule, and the final margin image can be obtained by inverse wavelet transform. The experimental results show that the fusion algorithm can accurately locate edge contour, effectively suppress noise very well, and the flame edge detection results are complete and clear. Objective evaluation was better than those results by using single algorithm.