基于实例分割的装车质量检测与调整方法研究

Research on Loading Quality Detection and Adjustment Method Based on Instance Segmentation

针对当前散料装车系统不能准确、完整、及时地获取料堆轮廓,料位检测易受到噪声干扰,导致无法保障装车作业质量等问题,创新性地提出了以实例分割技术为基础的装车质量检测与调整方法。基于Yolact实例分割术的货车箱体和煤堆检测,训练后的模型能在烟尘和强烈的光影环境下准确检测出煤堆和车厢边沿区域。基于煤堆和车厢边沿区域的装车撒料检测和基于煤堆和车厢侧边沿内侧垂直边缘的料位判断,检测得到的煤堆与车厢内壁接触线能够直接反映装车程度。基于实时料位信息的给料自动调整方法,通过已经装好的列车车厢数据进行自调整,适用于不同的车厢尺寸。试验结果表明,该方法不需要对车速进行设定就能实现比较理想的装车效果,对实现散料自动装车具有实际应用价值。

In response to the current problems that the bulk loading system cannot obtain the outline of the pile accurately,completely,and timely,and the material level detection is easily disturbed by noise,which leads to the failure to guarantee the quality of loading operation,a loading quality detection and adjustment method based on the instance segmentation technique is innovatively proposed.Based on Yolact instance segmentation for wagon box and coal pile detection,the trained model can accurately detect the coal pile and wagon edge area under the soot and strong light environment.Based on the coal pile and wagon edge area loading spill detection and the material level judgment based on the inner vertical edge of the coal pile and wagon side edge,the detected coal pile and wagon inner wall contact line can directly reflect the loading degree.The automatic adjustment method of feeding based on real-time material level information is self-adjusting by the data of already loaded train carriages and is applicable to different carriage sizes.The test shows that the method can achieve a more ideal loading effect without setting the car speed and has practical application value for realizing automatic loading of bulk materials.