基于正则表达式融合语义的农产品自动识别方法

Automatic recognition of agricultural products based on fusion semantics of regular expression

针对同一条传送带上不同类别农产品自动分拣识别率低的问题,根据应用场景下轮廓、颜色和空间相对位置等物理特征不变,以及不同农产品会产生不同振动波的特征,通过建立基于颜色、线条、位置、振动等信息的特征元素库,利用正则表达式的语义规则,对基础特征元素进行先验知识的有序组织,赋予它们描述不同农产品匹配特征的能力,从而快速构建出不同农产品的匹配模型,缩短分拣过程中农产品目标模型生成时间,同时由于融合了振动特征,使得外形相似农产品的识别准确率得到较大提升.试验结果表明:基于正则表达式融合语义特征提取的农产品识别方法能够快速、准确识别各类农产品,从而实现自动分拣的目的,在固定视角下其识别率为92.5%,平均识别时间为50.3 ms,相较传统的尺度不变特征变换(SIFT)和加速鲁莽特征(SURF)算法,本算法在固定视角下的分拣精度和分拣效率均有所提高.

To solve the problem of low rate of automatic sorting and identification for different agriculture products on the same transmission line, according to the physical identification characteristics of object contour, color , relative distance and different vibration characteristics by different products under special application scenario, the characteristic element library of colors,lines,locations and vibration was established. According to heuristic knowledge, the match feature of agriculture products was constructed by regular expression,and the matching model of agriculture products was constructed quickly.The generation time of model was shortened in image recognition process,and the efficiency of image recognition was greatly improved.The automatic sorting and recognition rate of products for different categories with similar contour was greatly improved due to the fusion of vibration characteristics.The experiment results show that the proposed method can achieve significant improvement in target recognition rate of agriculture products,and the recognition rate is 92.5% under constant perspective with average recognition time of 50.3 ms.Compared with conventional algorithms of scale invariant feature transform（SIFT） and speeded up robust feature（SURF）,the proposed method achieves significant improvement of automatic sorting and identification under constant perspective.