Hongtao Zhang

Haiyang Peng, Yi Zhan, Benkang Wang et al.

In High-definition (HD) maps, lane elements constitute the majority of components and demand stringent localization requirements to ensure safe vehicle navigation. Vision lane detection with LiDAR position assignment is a prevalent method to acquire initial lanes for HD maps. However, due to incorrect vision detection and coarse camera-LiDAR calibration, initial lanes may deviate from their true positions within an uncertain range. To mitigate the need for manual lane correction, we propose a patch-wise lane correction network (PLCNet) to automatically correct the positions of initial lane points in local LiDAR images that are transformed from point clouds. PLCNet first extracts multi-scale image features and crops patch (ROI) features centered at each initial lane point. By applying ROIAlign, the fix-sized ROI features are flattened into 1D features. Then, a 1D lane attention module is devised to compute instance-level lane features with adaptive weights. Finally, lane correction offsets are inferred by a multi-layer perceptron and used to correct the initial lane positions. Considering practical applications, our automatic method supports merging local corrected lanes into global corrected lanes. Through extensive experiments on a self-built dataset, we demonstrate that PLCNet achieves fast and effective initial lane correction.

Omar Alfarisi, Aikifa Raza, Hongtao Zhang et al.

Automated image processing algorithms can improve the quality, efficiency, and consistency of classifying the morphology of heterogeneous carbonate rock and can deal with a massive amount of data and images seamlessly. Geoscientists face difficulties in setting the direction of the optimum method for determining petrophysical properties from rock images, Micro-Computed Tomography (uCT), or Magnetic Resonance Imaging (MRI). Most of the successful work is from the homogeneous rocks focusing on 2D images with less focus on 3D and requiring numerical simulation. Currently, image analysis methods converge to three approaches: image processing, artificial intelligence, and combined image processing with artificial intelligence. In this work, we propose two methods to determine the porosity from 3D uCT and MRI images: an image processing method with Image Resolution Optimized Gaussian Algorithm (IROGA); advanced image recognition method enabled by Machine Learning Difference of Gaussian Random Forest (MLDGRF). We have built reference 3D micro models and collected images for calibration of IROGA and MLDGRF methods. To evaluate the predictive capability of these calibrated approaches, we ran them on 3D uCT and MRI images of natural heterogeneous carbonate rock. We measured the porosity and lithology of the carbonate rock using three and two industry-standard ways, respectively, as reference values. Notably, IROGA and MLDGRF have produced porosity results with an accuracy of 96.2% and 97.1% on the training set and 91.7% and 94.4% on blind test validation, respectively, in comparison with the three experimental measurements. We measured limestone and pyrite reference values using two methods, X-ray powder diffraction, and grain density measurements. MLDGRF has produced lithology (limestone and Pyrite) volumes with 97.7% accuracy.