论文部分内容阅读
为了实现土壤压实度检测,建立了土壤压实度的激光图像测量系统。首先采集土壤激光图像,并采用4邻域平均法对其平滑去噪;其次,采用Canny算法提取出激光图像中的激光光斑;然后选择含水量、激光光斑半径、吸收系数和散射系数作为分类器的输入特征参数;最后,利用反向传播(BP)神经网络预测压实度。实验结果表明,BP神经网络经过11次学习后,达到测量精确度的要求;与环刀法实际测量值相比较,平均绝对和相对误差在2%左右。因此,本文测量系统的检测精确度满足土壤压实度的检测要求。
In order to achieve soil compaction detection, a soil compaction laser image measurement system was established. Firstly, the soil laser image was collected and smoothed and denoised by 4-neighborhood averaging method. Secondly, Canny algorithm was used to extract the laser spot in the laser image. Then the water content, laser spot radius, absorption coefficient and scattering coefficient were selected as classifiers Finally, the compaction degree is predicted by back propagation (BP) neural network. The experimental results show that the BP neural network meets the requirements of measurement accuracy after 11 times of learning. Compared with the actual measurement of the ring knife method, the average absolute and relative errors are about 2%. Therefore, the detection accuracy of the measurement system in this paper meets the testing requirements of soil compaction.