Concrete sampling behavior detection system based on deep learning
Through deep learning technology, the information of concrete transport vehicles and bucket trucks is automatically identified and recorded, and the missed detection and misjudgment problems caused by manual operations are solved, the accuracy and traceability of concrete sampling inspection are improved, and intelligent management is realized. Reduces costs and risks.
Patent Information
- Application Number
- CN202411170180.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-26
- Publication Date
- 2024-11-22
- Estimated Expiration
- 2044-08-26
AI Technical Summary
The existing concrete sampling methods rely on manual operations, and there are problems such as missed inspection and record errors, resulting in construction delays and increased costs, and lack of automatic detection technology.
Using a concrete sampling behavior detection system based on deep learning, the image acquisition module, target detection module and behavior recognition module are used to automatically identify and record the information of concrete transport vehicles and bucket vehicles, calculate the level and depth spacing, and judge the residence time to achieve sampling Automatic detection and recording of behavior.
It improves detection accuracy and data traceability, reduces manual intervention, reduces costs and misjudgment risks, realizes intelligent management, and improves concrete quality monitoring and construction safety.
Smart Images

Figure CN118691802B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of deep learning technology, and in particular to a concrete sampling behavior detection system based on deep learning. Background Art
[0002] Concrete quality is an extremely important indicator in construction projects, which directly affects the safety and service life of buildings. During the transportation process from the concrete manufacturing plant to the construction site, the quality of concrete is affected by many factors such as raw materials, mix ratio, transportation time, etc. Therefore, every concrete transport truck arriving at the construction site needs to be sampled and tested to ensure that the concrete quality meets the design specifications and construction requirements.
[0003] Existing concrete sampling methods usually rely on manual operations, which are not only time-consuming and labor-intensive, but also prone to problems such as missed detection and recording errors. Once concrete is missed or sample records are incomplete, it may be impossible to trace the source of the concrete after construction. If the concrete is found to be unqualified, the constructed parts need to be demolished and redone according to safety regulations, causing delays in construction and increased costs. In addition, manual recording methods also increase labor costs and there is a risk of human negligence or error. The prior art lacks a technical solution for automatically detecting concrete sampling behavior. Summary of the invention
[0004] In order to solve the above problems, the present invention proposes a concrete sampling behavior detection system based on deep learning. The system can automatically identify and record relevant information of concrete transport vehicles and sampling processes, identify, track and count concrete transport vehicles and bucket trucks through deep learning algorithms, thereby realizing automatic detection and recording of sampling behaviors, reducing human intervention and errors, and improving the efficiency and accuracy of the sampling process.
[0005] In order to achieve the above-mentioned invention object, the technical solution provided by the present invention includes:
[0006] The concrete sampling behavior detection system based on deep learning includes:
[0007] An image acquisition module for acquiring real-time images including concrete transport vehicles and bucket trucks;
[0008] A target detection module for detecting the concrete transport vehicle and the bucket truck in the current image by using a target detection algorithm and calculating the horizontal and vertical distances between the concrete transport vehicle and the bucket truck, and for determining whether the horizontal and vertical distances respectively meet preset distance thresholds;
[0009] A behavior recognition module for tracking and identifying the bucket truck and determining whether the bucket truck stays within the image range for a time exceeding a preset time threshold;
[0010] A reporting module for reporting the image information and time information of this concrete sampling behavior when the bucket truck stays within the image range for a time exceeding a preset time threshold.
[0011] Preferably, the image information of the concrete sampling behavior includes license plate information of the concrete transport vehicle.
[0012] Preferably, the image information of the concrete sampling behavior also includes image frames acquired at preset time intervals.
[0013] Preferably, the method for the target detection module to calculate the horizontal distance between the transport vehicle and the bucket truck includes:
[0014] The target detection algorithm is used to obtain the detection frames in the detection results of the transport vehicle and the bucket truck in the current image, and the pixel distance between the two detection frames in the horizontal direction is calculated as the horizontal spacing between the transport vehicle and the bucket truck.
[0015] Preferably, the method for the target detection module to calculate the depth spacing between the transport vehicle and the bucket truck includes:
[0016] A monocular depth estimation algorithm is used to obtain the detection results of the transport vehicle and the bucket truck in the current image, and the distance between the two detection results and the image acquisition device is calculated, thereby calculating the depth spacing between the transport vehicle and the bucket truck.
[0017] Preferably, the method by which the behavior recognition module determines whether the bucket truck stays within the image range for a time period exceeding a preset time threshold comprises:
[0018] When the horizontal and vertical distances between the transport vehicle and the bucket truck in the current image both meet the preset distance threshold, the current time is recorded as the first timestamp;
[0019] Tracking and identifying the bucket truck, when the bucket truck leaves the image range, recording the current time as a second timestamp;
[0020] According to the time difference between the second timestamp and the first timestamp, it is determined whether it exceeds a preset time threshold.
[0021] Beneficial Effects
[0022] 1. Improve detection accuracy: The present invention can accurately identify concrete transport trucks and bucket trucks, and accurately measure the horizontal and vertical distances between them through the target detection algorithm based on deep learning and the monocular depth estimation algorithm, effectively eliminating the interference of non-sampling bucket trucks and passing bucket trucks in the camera's field of view, and ensuring the accuracy of sampling behavior detection.
[0023] 2. Enhanced data traceability: The system can automatically record the image information, license plate information and time information of each concrete sampling behavior, ensuring that the source and sampling process of each concrete sample are traceable, providing reliable data support for the tracing and handling of subsequent quality issues.
[0024] 3. Reduce manual intervention: By automating the detection and recording of sampling behaviors, the workload of manual recording and detection is reduced, the labor cost is reduced, and the risk of human negligence and recording errors is reduced, thereby improving the efficiency and reliability of the sampling process.
[0025] 4. Realize intelligent management: The concrete sampling and detection system based on deep learning can be integrated with other intelligent management systems to build a comprehensive intelligent construction management platform, realize the whole process monitoring and management of concrete quality, and further improve the quality and safety level of construction projects. BRIEF DESCRIPTION OF THE DRAWINGS
[0026] Figure 1 A structural schematic diagram of a concrete sampling behavior detection system based on deep learning provided by the present invention;
[0027] Figure 2 A schematic diagram of a method for calculating the horizontal spacing between a concrete transport vehicle and a bucket truck provided in a preferred embodiment of the present invention;
[0028] Figure 3 It is a schematic diagram of a method for calculating the depth spacing between a concrete transport vehicle and a bucket truck provided in a preferred embodiment of the present invention. DETAILED DESCRIPTION
[0029] In order to make the purpose, technical solution and advantages of the present invention clearer, the present invention is further described below in conjunction with the accompanying drawings. In the description of the present invention, it should be understood that the terms "upper", "lower", "front", "back", "left", "right", "top", "bottom", "inside", "outside" and the like indicate directions or positional relationships based on the directions or positional relationships shown in the accompanying drawings, which are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific direction, be constructed and operated in a specific direction, and therefore cannot be understood as limiting the present invention.
[0030] like Figure 1 As shown, this embodiment provides a concrete sampling behavior detection system based on deep learning, including:
[0031] An image acquisition module for acquiring real-time images including concrete transport vehicles and bucket trucks. The image acquisition module includes an image acquisition device (such as a camera, a video camera, etc.) arranged in a target area for real-time or timed acquisition. The image acquisition module is arranged in a target detection area and continuously monitors the area to acquire clear images including concrete transport vehicles and bucket trucks, ensuring the continuity and real-time nature of the images for subsequent image processing and analysis. The bucket truck refers to a small vehicle used to transport or load materials. They usually have a tilted or dumpable bucket for convenient loading and unloading of materials. In the concrete sampling process, the bucket truck is usually used to remove concrete samples from the transport vehicle and transport them to the sampling and testing site. In the present invention, the presence of the bucket truck is used as an identifier of the sampling behavior. By detecting and tracking the bucket truck and the concrete transport vehicle, automatic monitoring and recording of the concrete sampling behavior can be achieved.
[0032] A target detection module is used to detect concrete transport vehicles and bucket trucks in the current image through a target detection algorithm, and calculate the horizontal and depth distances between the two. The target detection module is also used to determine whether the horizontal and depth distances meet the preset distance thresholds respectively. Those skilled in the art will know that since concrete sampling requires the bucket truck to maintain a certain distance from the rear of the concrete transport truck, it is possible to determine whether a sampling operation is being performed by determining whether the distance between the concrete transport truck and the bucket truck in the horizontal and depth directions meets the preset conditions. Specifically, when it is detected that the horizontal distance and depth distance of the bucket truck at the rear of the concrete transport truck are both within the preset range, it can be determined that the bucket truck is performing a concrete sampling operation, thereby eliminating possible interference caused by other non-sampling bucket trucks within the image range.
[0033] It should be understood that the present invention can use the target detection algorithms commonly used in the art to detect and identify concrete transport vehicles and bucket trucks, such as Faster R-CNN, YOLO series algorithms, SSD, RetinaNet, etc. Using these algorithms, concrete transport vehicles and bucket trucks in the image can be accurately identified and located, so as to calculate their distance and perform subsequent processing.
[0034] The horizontal distance between the concrete transport vehicle and the bucket truck can be measured by binocular vision or other commonly used distance measurement methods (such as ultrasonic distance measurement, infrared distance measurement), etc. In some preferred embodiments, Figure 2As shown, an example of calculating the horizontal distance based on the target detection algorithm is given, specifically including: using the target detection algorithm to obtain the detection frame in the detection results of the transport vehicle and the bucket car in the current image, and calculating the pixel distance between the two detection frames in the horizontal direction as the horizontal distance between the transport vehicle and the bucket car. In the figure, d is the horizontal distance between the transport vehicle and the bucket car. By directly calculating the horizontal distance through the target detection algorithm, additional hardware settings and complex distance measurement equipment are avoided, and the system structure and hardware configuration are simplified. No additional distance measurement equipment (such as ultrasonic sensors, infrared sensors, etc.) is required, which reduces hardware costs and maintenance costs, and has high economic benefits. By directly obtaining the horizontal distance through the existing target detection algorithm, the system calculation amount and data processing time are reduced, and the overall operation efficiency and real-time performance of the system are improved.
[0035] For the depth distance between the concrete transport vehicle and the bucket truck, binocular vision or other commonly used distance measurement methods (such as ultrasonic distance measurement, infrared distance measurement) can also be used. In some preferred embodiments, Figure 3 As shown, an example of calculating the depth spacing based on the target detection algorithm is given, specifically including: using a monocular depth estimation algorithm to obtain the detection results of the transport vehicle and the bucket truck in the current image, and calculating the distance between the two detection results and the image acquisition device, and calculating the depth spacing of the transport vehicle and the bucket truck accordingly, where L in the figure is the depth spacing of the transport vehicle and the bucket truck. The specific implementation method is described below:
[0036] First, the target is detected on the current image, and the upper left, upper right, lower right, and lower left corner coordinates of the rectangular area where the target (i.e., concrete truck and bucket truck) is located are obtained. Based on these four coordinates, the coordinates of the center point of the rectangular area are calculated as the representative of the target location. Then, the monocular depth estimation is performed on the entire current image to obtain the distance from these center points to the camera as the distance between the target and the image acquisition device.
[0037] A behavior recognition module is used to track and identify the bucket truck and determine whether the bucket truck stays within the image range for a time that exceeds a preset time threshold. There are many methods for tracking and identifying bucket trucks. Since this part of the content is not the focus of the present invention, the present invention does not further limit this. Those skilled in the art can choose a suitable implementation method in the prior art according to actual needs. As mentioned above, the target detection module eliminates the interference of non-sampling bucket trucks by calculating the distance between the concrete transport truck and the bucket truck. However, due to the complexity of vehicles and personnel on the construction site, there may be interference from bucket trucks that meet the conditions but are just passing by. To this end, the behavior recognition module eliminates the interference of passing bucket trucks by examining the bucket truck's stay time within the image range. Specifically, only when the bucket truck stays within the image range for a time that exceeds the preset threshold, it is determined that it is performing a sampling operation, thereby further improving the accuracy and reliability of the detection.
[0038] In some preferred embodiments, when examining the time that the bucket truck stays within the image range, the interference of passing bucket trucks or other irrelevant bucket trucks that stay within the image range is further eliminated, and the following method for determining whether the time that the bucket truck stays within the image range exceeds a preset time threshold is provided, specifically including:
[0039] When the horizontal and depth distances between the transport vehicle and the bucket truck in the current image meet the preset distance threshold, the current time is recorded as the first timestamp; the bucket truck is tracked and identified, and when the bucket truck leaves the image range, the current time is recorded as the second timestamp; based on the time difference between the second timestamp and the first timestamp, it is determined whether it exceeds the preset time threshold.
[0040] This embodiment considers using "when the horizontal and vertical distances between the transport vehicle and the bucket truck in the current image both meet the preset distance threshold" as the starting time, rather than the time when the bucket truck enters the image as the starting time, mainly for the following reasons:
[0041] 1. Improve detection accuracy:
[0042] Only when the horizontal and vertical distances between the concrete transport vehicle and the bucket truck simultaneously meet the preset distance thresholds, the bucket truck is considered to be possibly performing sampling operations. This can effectively exclude irrelevant vehicles and other interference factors and improve the accuracy of sampling detection.
[0043] 2. Reduce misjudgment:
[0044] If the time when the bucket truck enters the image is used as the start time, it may lead to misjudgment of non-sampling bucket trucks (such as bucket trucks passing by). By setting the distance threshold condition, only bucket trucks that are really close to the concrete transport truck and meet the sampling behavior characteristics are identified and tracked, reducing the probability of misjudgment.
[0045] 3. Adapt to complex environment:
[0046] At construction sites, vehicles and personnel are active, and it is easy to be disturbed if we only rely on the bucket truck entering the image to judge the sampling behavior. By combining the dual conditions of distance threshold and dwell time, it can better adapt to complex environments and ensure the stability and reliability of the detection system.
[0047] A reporting module for reporting the image information and time information of the concrete sampling behavior when the bucket truck stays within the image range for a time period exceeding a preset time threshold. Obviously, the reported image information should be able to reflect the object and process of the sampling behavior. Therefore, in some preferred embodiments, the image information includes the license plate information of the concrete transport vehicle. Obviously, the license plate information of the concrete transport vehicle can be realized by an image recognition algorithm, and the recognition step can be performed in any step of the entire sampling behavior detection method. On the other hand, the image information also includes image frames acquired at preset time intervals. Through these image frames, the entire process of the sampling behavior can be recorded and reflected in detail to ensure the integrity and traceability of the data.
[0048] It should be understood that if the judgment result of any of the above modules is no, the entire detection process ends and a new round of identification starts again. This ensures that each detection of concrete sampling behavior starts from the initial image acquisition, ensuring the integrity and accuracy of the process.
[0049] The above shows and describes the basic principles, main features and advantages of the present invention. It should be understood by those skilled in the art that the present invention is not limited to the above embodiments, and the above embodiments and descriptions are only for explaining the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention to be protected. The scope of protection of the present invention is defined by the attached claims and their equivalents.
Claims
1. A concrete sampling behavior detection system based on deep learning, characterized in that: include: An image acquisition module for acquiring real-time images including concrete transport vehicles and bucket trucks; A target detection module for detecting the concrete transport vehicle and the bucket truck in the current image by using a target detection algorithm and calculating the horizontal and vertical distances between the concrete transport vehicle and the bucket truck, and for determining whether the horizontal and vertical distances respectively meet preset distance thresholds; A behavior recognition module for tracking and identifying the bucket truck and determining whether the bucket truck stays within the image range for a time exceeding a preset time threshold; A reporting module for reporting the image information and time information of this concrete sampling behavior when the bucket truck stays within the image range for a time exceeding a preset time threshold; The method for calculating the horizontal distance between the transport vehicle and the bucket truck by the target detection module includes: Using a target detection algorithm to obtain the detection frames in the detection results of the transport vehicle and the bucket truck in the current image, and calculating the pixel distance between the two detection frames in the horizontal direction as the horizontal spacing between the transport vehicle and the bucket truck; The method for the target detection module to calculate the depth spacing between the transport vehicle and the bucket truck includes: A monocular depth estimation algorithm is used to obtain the detection results of the transport vehicle and the bucket truck in the current image, and the distance between the two detection results and the image acquisition device is calculated, thereby calculating the depth spacing between the transport vehicle and the bucket truck.
2. The deep learning-based concrete sampling behavior detection system according to claim 1, characterized in that: The image information of the concrete sampling behavior includes license plate information of the concrete transport vehicle.
3. The deep learning-based concrete sampling behavior detection system according to claim 1 or 2, characterized in that: The image information of the concrete sampling behavior also includes image frames acquired at preset time intervals.
4. The deep learning-based concrete sampling behavior detection system according to claim 1, characterized in that: The method for the behavior recognition module to determine whether the bucket truck stays within the image range for a time period exceeding a preset time threshold includes: When the horizontal and vertical distances between the transport vehicle and the bucket truck in the current image both meet the preset distance threshold, the current time is recorded as the first timestamp; Tracking and identifying the bucket truck, when the bucket truck leaves the image range, recording the current time as a second timestamp; According to the time difference between the second timestamp and the first timestamp, it is determined whether it exceeds a preset time threshold.
