<p>Research based on computer vision in the construction industry has focused on detecting the presence of objects, people and equipment on construction sites to increase safety and productivity. The construction sector has identified ‘fall from height’ (FFH) as the leading cause of fatalities. Barricades are installed to safeguard employees against FFH. The missing barricades are commonly detected by manual inspection, which is labor-intensive. Computer vision techniques are also used but they concentrate on pattern recognition, which limits their capacity to detect the missing barricade accurately and makes them time-consuming. Additionally, these methods are not robust to challenges such as dynamic environment changes and object occlusion. To overcome these issues, we proposed an automatic Missing 3D object detection model (M3D-ODM) using a 3D Sequential-Siamese-U-Net (3D SSU-Net) model for accurate detection of missing barricades. Initially, to establish a relationship between the imaging plane and its projected images, a vanishing point-based camera calibration method is used. The structure of 3D volume (3DV) is constructed to obtain 3D features in regular space based on plane sweep volume and geometric volume information. A bounding box annotation method is used to detect barricades with distance as well as depth information, and finally, the missing barricade is detected using 3D SSU-Net. The performance metrics of the proposed method is evaluated and compared with the other existing approaches. The results show that the proposed model is superior to existing methods, achieving a precision of 97.83%, recall of 98.51%, F1-score of 98.01%, Dice coefficient (DC) of 98.22%, and detection accuracy (DA) of 97.92%, clearly demonstrating its robustness and accuracy in real-world construction site environments.</p>