The early displacement monitoring of excavation pits is currently a primary method to prevent pit collapse accidents. However, conventional pit displacement monitoring mainly relies on regular manual measurement or automated professional instrument measurement, which results in high cost, complex operation and deployment, non-real-time measurement, and non-intuitive data collection. Consequently, this paper proposes a method for the visual measurement of three-dimension (3D) displacement of pits based on a two-stage Perspective-n-Point (PNP) algorithm. Initially, markers are installed on the reference points with known 3D spatial coordinates and on each displacement measurement point of the excavation pit. All markers are scanned sequentially by panning the camera and zooming the lens, ensuring at least two adjacent markers enter the frame and are captured in each shot. Secondly, the improved Histogram of Oriented Gradients (HOG) combined with Support Vector Machine (SVM) rapid algorithm is utilized to perform quick detection and coarse positioning of each marker in the aforementioned photos. Thirdly, the sub-pixel algorithm and image morphology are combined. Based on the coarse positioning, the pixel coordinate values of the markers in the images are calculated. Lastly, the two-stage PNP algorithm is applied in a cyclical fashion: by using the known 3D spatial coordinates and its pixel coordinates of the previous marker to solve for the current camera’s 3D spatial coordinates and pose, the 3D spatial coordinates of the next marker can be derived using the current camera’s 3D spatial coordinates and pose and the pixel coordinates of the next marker. This cycle continues until the 3D spatial coordinates of all markers are fully resolved, thereby achieving accurate measurement of the 3D displacement of the pit. Experimental results show that the accuracy of the proposed method is nearly as high as the measurement accuracy of manual total stations, meeting the requirements for daily 3D displacement measurement of excavation pits. Meanwhile, this method offers the advantages of low cost, simple deployment and operation, and a high degree of automation.

In the process of automatic container terminal loading and unloading, the three-dimensional attitude of the container affects the security of loading and unloading operations, so the three-dimensional attitude positioning of the container is very important. In this paper, a visual non-contact measurement method is used to realize the real-time orientation of the three-dimensional attitude of the container. First, the container corner is coarsely positioned by a small-scale deep learning network. Secondly, the precise position of the container keyhole is obtained by the secondary positioning of the container corner through the traditional image processing algorithm, and the container posture is measured in three dimensions by combining the physical motion model of the container during loading and unloading. After testing, unlike previous measurement methods, the measurement accuracy of this method met the requirements of automatic loading and unloading of container terminals, and the measurement time met the requirements of real-time measurement.

With large-scale, integrated, intelligence for ports, many ports begin to use intelligent detection systems to make their operations more efficient. The container truck recognition and positioning system is also beginning to apply into container quayside to assist the joint operations between quay cranes and container trucks. However, the traditional vehicle detection by using motion region detection cannot recognize the type of moving object, and the traditional pattern recognition method cannot meet the requirements in real-time operation. In order to solve these problems, an algorithm fused by regional clustering and two-stage SVM classifier is proposed in this paper. The method consists of two phases, which are independently executed in two camera systems on quay cranes. In the first stage, a fast motion regional clustering algorithm is used to detect moving image patches as the truck candidate sub-windows. In the second stage, the container trucks will be recognized in these sub-windows by an optimized two-stage SVM classifier. Compared with existing traditional algorithm, experimental results in container terminal show that the fusion algorithm with regional clustering and two-stage SVM has higher efficiency and better truck recognition performance.

This paper analyzes the problems of existing container positioning methods and proposed a vision-based container position measuring system to provide precise parameters for container lifting operations. This system uses camera to get container information, then detects container corners by the method that combined with convolutional neural network and traditional image processing algorithm. This system is used to provide specific parameters associated with container lifting operation. In the first of detection, it uses the modified SSD (Single Shot MultiBox Detector) neural network to detect the coarse position of container corners in the image, second stage detection uses the usage of rectangle fitting to detect the precise position of corner holes in the coarse position. In the last step the offset distance and deflection angle were calculated by precise corner position. The experiment shows the detection rate of the proposed system reach 94%. The positioning errors between 14.3 and 19.6 mm for a frame rate of 10 fps are obtained.

The visual sensor network (VSN) is an important part of the automated container terminal. VSN present also a of problems such as the redundant number of visual sensors and limited computing resources. Different vision sensors are reported in literature when about computing resources (CR) reduced number of publications provides information about required CR as part of visual sensor network. In this context the paper propose solution that transforms the visual sensor network terminal task scheduling process into Markov Decision Process. Thus, a visual sensor network terminal task scheduling algorithm based on Deep-Q Learning is considered. In this algorithm, an innovative return value function is proposed to achieve better algorithm convergence. To verify the effectiveness of the model, several experiments were carried out under different conditions. The result shows that the recognition rate is improved by using the proposed algorithm. Based on considered method the number of visual sensors can be reduced, that conducts to a rational use of limited computing resources. At the same time cost reduction is also provided that is an important requirement of port operation optimization.