摘要
为能够在复杂背景下更高效识别路面裂缝,通过加入自适应因子对人工蜂群(Artificial bee colony,ABC)算法的搜索位置和概率选择进行改进,利用改进的ABC算法去优化BP神经网络的权值与阈值,建立一种改进的ABC-BP混合神经网络路面裂缝识别算法。实验结果表明,该方法在收敛速度和准确度上优于基本的ABC-BP算法和BP算法,准确率、召回率和综合评价指标都超过了95%,验证了算法的通用性与有效性。
In order to identify pavement cracks more efficiently in complex backgrounds, an improved artificial bee colony and backward propagation neural network algorithm(ABC-BP) was proposed to identify pavement cracks. The improved ABC algorithm can improve search position and probability selection of ABC algorithm by adding adaptive factors, and was therefore used to optimize the weights and thresholds of the BP neural network. Experimental results show that the proposed method exceeds 95% in accuracy, recall rate and comprehensive evaluation indicators. In addition, this method is superior to the basic ABC-BP algorithm and BP algorithm in terms of convergence speed and accuracy. Furthermore, the versatility and effectiveness of the algorithm were verified.
作者
谭卫雄
王育坚
李深圳
TAN Weixiong;WANG Yujian;LI Shenzhen(School of information,Beijing Union University,Beijing 100101,China)
出处
《铁道科学与工程学报》
CAS
CSCD
北大核心
2019年第12期2991-2998,共8页
Journal of Railway Science and Engineering
基金
国家自然科学基金资助项目(61572077,61871038)
