Landfill Impervious Layer Damage in Operation Based on Machine Vision

Abstract

Aiming at the problem that the leakage area of high-density polyethylene membrane cannot be longitudinally located and contour identified under landfill conditions. First, according to the vibration difference characteristics of the landfill medium layer, a cross-correlation processing model is proposed using the calibration signal and the actual collected working signal. Combined with the complex wavelet domain transform, the time-frequency information of the prominent components is extracted. When the main frequency of the vibration signal is concentrated in 1,000–2,000 Hz and the correlation value is greater than 0.8, the longitudinal positioning of the target layer is completed. After removing the membrane medium in the leakage area, the leakage area image is collected, the data set is constructed, and the image preprocessing is carried out. Then the support vector machine classification and recognition model is constructed based on binary tree theory, and an adaptive particle swarm optimization algorithm is proposed to optimize the parameters of the model. When the penalty factor C = 8 and the kernel parameter g = 4.74, the model is optimal. Finally, the experimental results show that the overall recognition accuracy of the optimized classification model is 98.89%, of which the complete image recognition accuracy is 99.58%, the block damage recognition accuracy is 98.33%, and the seam damage recognition accuracy is 98.75%, which is significantly improved compared with that before optimization.