A Method of Mining Truck Loading Volume Detection Based on Deep Learning and Image Recognition

Area of Science:

• Mining Engineering
• Computer Vision
• Artificial Intelligence

Background:

• Accurate measurement of mining truck loading volume is crucial for operational efficiency in open-pit mining.
• Existing methods often suffer from low accuracy and high operational costs.
• The scarcity of labeled mine data presents a significant challenge for developing robust detection models.

Purpose of the Study:

• To develop and validate a novel deep learning-based model for precise detection of mining truck loading volumes.
• To address the limitations of current methods by leveraging image recognition and deep learning techniques.
• To overcome the challenge of limited labeled data by utilizing a VGG16 network and least squares algorithm.

Main Methods:

• Image preprocessing was applied to a dataset of 6000 laboratory images.
• The VGG16 network model was employed for pre-classification of ore images.
• Loading volume was calculated using classification results and the least squares algorithm.
• Model performance was evaluated using Root Mean Square Error (RMSE) and Mean Absolute Error (MAE).

Main Results:

• The model achieved a high prediction accuracy with an average absolute error of 17.85 cm³ on laboratory data.
• Validation with 400 real-world open-pit mine images showed a reduced average absolute error of 2.53 m³.
• The model demonstrated good generality and effective applicability to actual mining operations.

Conclusions:

• The proposed deep learning model offers a highly accurate and cost-effective solution for mining truck loading volume detection.
• The method successfully addresses the issue of limited labeled data, enhancing model development.
• The model's proven generality makes it suitable for practical implementation in open-pit mining environments.