Estimating construction waste truck payload volume using monocular vision

Quantifying truck-loaded materials is a problem in many industrial operations. In construction and demolition waste (CDW) management, inspectors at disposal facilities are often required to measure the amount of different waste components loaded by incoming trucks to determine admissibility. Due to the bulky and mixed nature of construction materials, accurate quantification of specific waste categories without sacrificing operability in the field is a challenge. This study proposes a CDW volume estimation algorithm based on monocular vision which can automatically quantify from a single image the amount of specific material components, e.g., rock, gravel, and wood, in waste mixtures. The algorithm achieves a relative error of 0.065 in calculating truck bucket dimensions, and a relative error of 0.169 in estimating material-level construction waste volume. It takes 3.3 s in average to process one image. In applying the algorithm to analyze 2,914 waste truckloads received by an off-site sorting facility in Hong Kong, we observe that the facility entrance received around 800.0 m(3) CDW per day of which about 10.8 m(3) were rejected. Since non-inert wood/cardboard accounts for the highest proportion among all material types, this may imply that many waste dumps accepted by the facility may have been in violation of the admissibility criteria. The study contributes to the knowledge body by providing a novel, non-destructive approach to quantifying CDW via monocular vision. It can be extended to address the general problem of truck payload quantification in scenarios such as road construction, warehouse inventory management, and logistics and supply chain management.

Automated summary

Quantifying truck-loaded materials is challenging in construction and demolition waste management. A novel CDW volume estimation algorithm based on monocular vision is proposed in this study, achieving accurate quantification of specific material components from a single image and showing promising results in practical applications.