多特征融合的交通标识视认性评测方法

针对目前交通标识视认性评测方法误差过大的缺点，提出了一种高精度交通标识视认性评测方法。该方法根据道路环境中不同因素对交通标识视认性的影响，计算了标志牌图像中的颜色特征、亮度特征、复杂度特征，并考虑标志牌背景的影响，计算了标志牌和背景之间的颜色对比特征、亮度对比特征和复杂度对比特征，综合考虑以上6种特征及特征相互之间的影响，利用自顶向下(top-down)和自底向上(bottom-up)视觉模型建立多特征融合的交通标识视认性评测模型。通过模型实现了对识别出的标志牌进行视认性评测，反馈模型推测的视认性值。通过实验对多特征融合的视认性评测模型的有效性及模型精度进行了评测。结果表明，该模型能够高精度推测标志牌视认性值，且达标率在89%以上。

Evaluation method of traffic signs recognition based on multi-feature fusion

This paper proposes a high precision method for evaluating the visibility of traffic signs to reduce the large deviation of the result caused by the current methods. Due to the influence of different factors in the road environment on the visibility of traffic signs, the color features, brightness characteristics, and complexity features in the signboard images are determined, and the influence of the background of the signboard is taken into consideration to obtain the color contrast feature, brightness contrast features and complexity contrast feature, comprehensively consider the above 6 characteristics and the influence of the features on each other, then using top-down and bottom-up visual models to establish multi-feature fusion traffic sign visibility evaluation models. The visibility of the identified signboard is evaluated by the model, and the visibility value estimated by the model is fed back. Finally, the effectiveness and model accuracy of the multi-feature fusion visual evaluation model were evaluated by experiments. And the final experiment shows that the model can predict the visibility of the signage with high accuracy, and the compliance rate is above 89%.