论文部分内容阅读
提出了一种复杂地面背景下的红外车辆检测算法。首先,提出一种新的自适应分段线性灰度拉伸方法来增强当图像整体亮度偏低时的目标信息。其次,利用拉伸后图像的显著性图生成目标潜在的兴趣区。再次,利用平均梯度法在兴趣区内进行目标的边缘再分割,完成目标精确分割检测。最后,利用车辆的红外融合特征计算目标置信度,对目标进行评估和确认。实验结果表明:对实际拍摄的红外图像进行检测的算法可有效地检测出地面车辆目标。
An infrared vehicle detection algorithm based on complex terrain is proposed. First, a new adaptive piecewise linear gray-scale stretching method is proposed to enhance the target information when the overall image brightness is low. Secondly, using the saliency map of the stretched image to generate the target potential area of interest. Thirdly, using the average gradient method to segment the edge of the target in the region of interest to complete the target accurate segmentation detection. Finally, the vehicle’s infrared fusion characteristics are used to calculate the confidence of the target and the target is evaluated and confirmed. The experimental results show that the algorithm of detecting the actual infrared image can effectively detect the target of the ground vehicle.