Image-Based Particle Size Analysis in Mineral Processing

Image-based particle size analysis has emerged as a pivotal technique for quantifying the size distribution of mineral particles in comminution, separation and transport circuits. By capturing high-resolution images of ore particles on conveyor belts or blasted piles, digital image processing pipelines segment individual fragments and classify their dimensions without the need for traditional sieving. Key steps include image pre-processing to mitigate noise, adaptive thresholding to generate binary masks, morphological operations and contour detection to delineate particles, and advanced machine-learning models to refine segmentation and categorisation. These approaches enable continuous, non-intrusive monitoring of particle-size distributions, facilitating real-time optimisation of crushing, grinding and separation processes. Globally, such innovations contribute to enhanced throughput, reduced energy consumption and improved safety in mining operations by providing timely insights into equipment performance and product quality.

Research from Nature Portfolio

Recent studies have presented an adaptive watershed segmentation algorithm tailored to the challenges of blasted rock imagery. By integrating a Phansalkar binarisation for enhanced contrast, contour-solidity criteria to mark seed points and distance-transformation filtering, the method effectively resolves issues of adhesion, stacking and blurred edges in heterogeneous rock piles. Validation on both limestone and granite piles demonstrated segmentation accuracies exceeding 95% for particles above defined area thresholds, with area-distribution curves closely matching manual measurements. This development offers a robust route to rapid field deployment for precise particle-size monitoring in open-pit and quarry environments.

Image-Based Particle Size Analysis in Mineral Processing publication trend

The graph below shows the total number of articles in image-based particle size analysis in mineral processing across all publications each year (not limited to Nature Index journals).

Technical terms

Segmentation: The process of partitioning an image into regions corresponding to individual particles or features.

Binarisation: Conversion of a greyscale image into a binary mask by applying a threshold to distinguish objects from the background.

Watershed algorithm: A morphological technique that treats image intensities as topographical surfaces to separate adjacent objects by flooding basins from seed points.

U-Net: A convolutional neural network architecture featuring an encoder–decoder structure with skip connections for precise image segmentation.

Phansalkar binarisation: A local adaptive thresholding method that computes thresholds based on mean and standard deviation in neighbourhood windows, enhancing low-contrast regions.

Intersection over Union (IoU): A metric assessing segmentation accuracy by dividing the overlap area of predicted and ground-truth masks by their union area.

References

1. Rock Fragmentation Classification Applying Machine Learning Approaches. Engineering (2023).
2. Ore image segmentation method using U-Net and Res_Unet convolutional networks. RSC Advances (2020).
3. A method of blasted rock image segmentation based on improved watershed algorithm. Scientific Reports (2022).
4. An Improved Boundary-Aware U-Net for Ore Image Semantic Segmentation. Sensors (2021).