复杂场景下铁路货车锁紧板偏转故障检测

摘要

Catering to the urgent need for safe operation of the railway and based on the research background of the deflection detection of locking plates in the trouble of moving freight car detection system (TFDS),the pos-ture fault of locking plates is discriminated in the automatic recognition mode of non -fault targets.In order to detect whether locking plate posture is located within the normal range,it was proposed a novel deflection angle estimation method.Based on the outline of the locking plate constituted by nearly straight line segment group,it was suggested to descript the object by straight segment features and form a sparse expression of features.The o-rientation encoding was also used in straight line segment features.The deflection angle estimation was achieved by the directed Chamfer match and only a template was needed the whole process.Experiments show that the al-gorithm can effectively discriminate the object that the deflection angle located within the normal range and has good robustness and accuracy.%结合铁路安全运营的迫切需求，以铁路货车运行故障图像动态检测系统（TFDS）中锁紧板偏转故障检测为研究背景，在无故障目标自动识别模式下对锁紧板姿态进行故障判别。针对锁紧板姿态是否位于正常范围内，提出一种新颖的偏转角度估计方法。算法依据锁紧板轮廓由近直线段组构成，提出采用直线段特征描述物体，形成特征稀疏表达，并对直线段特征采用方向编码方式，在有向 Chamfer 匹配下实现偏转角度估计，整个过程仅需一个模板。实验证明该算法能有效识别出偏转角度位于正常范围内的物体，并具有较好的鲁棒性和准确性。