An early warning method and device for a commercial concrete truck water injection detection system

Abstract

The present invention provides an early warning method and device for a commercial concrete truck water injection detection system, comprising: positioning and arranging a camera device on the commercial concrete truck; collecting video stream data in real time based on the positioning and arrangement results and transmitting it to an on-board edge device for image processing; calibrating the water injection area based on the processed image data according to the current operating environment; identifying and analyzing water injection behavior in the water injection area, and performing an alarm operation when water injection behavior occurs; and automatically activating an upload mechanism to upload data when the alarm operation is triggered. This effectively improves the accuracy of water injection behavior identification and automatically activates the upload mechanism to upload data when the alarm operation is triggered, effectively ensuring data traceability, thereby effectively implementing data management and improving the effectiveness and accuracy of commercial concrete truck water injection detection.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to an early warning method and device for a water injection detection system of a commercial concrete truck. Background Art

[0002] At present, commercial concrete trucks are responsible for the key task of transporting commercial concrete to construction sites, and are an indispensable part of construction. As an important building material, the quality and use of concrete must strictly comply with relevant standards. These standards clearly stipulate that it is prohibited to add water privately during the transportation and pouring of concrete to ensure that the concrete mix ratio is accurate, thereby protecting the strength, durability and overall quality of the concrete. However, in actual operations, due to the influence of various factors, some non-standard operating behaviors are difficult to eliminate and difficult to effectively supervise. For example, workers privately add water to facilitate pumping or improve fluidity. This behavior will change the original mix ratio of concrete, thereby causing quality problems and affecting the safety of commercial concrete. Therefore, how to effectively supervise and prevent such non-standard behaviors has become an urgent problem to be solved.

[0003] Based on the above needs, the invention patent with patent number CN115830503A, "A method and system for distinguishing illegal water addition behavior of concrete tankers", discloses a method based on fixed camera installation, building an object list according to the detection results of multiple targets, and using three different behavior judgment modules to confirm whether water is added. Although these methods have achieved certain results, there are still some shortcomings in the actual process: First, the environmental adaptability is insufficient. The camera installation position and angle in the existing solution are fixed, resulting in certain key areas that cannot be effectively monitored. It is difficult to adapt to different vehicle models and complex construction scenes, and is easily affected by light and occlusion. Once the vehicle leaves the fixed construction site, the water injection behavior of the commercial concrete truck cannot be effectively monitored; secondly, based on multi-target detection, there are some problems in the implementation of the three different water injection behavior discrimination methods. The first method is to use a single-frame image to judge the water injection behavior based on the overlap between multiple target objects, without considering the false detection situation; the second method is to record the license plate features of the commercial concrete truck, and detect the same vehicle twice within 10 minutes to judge the water injection behavior, without considering the false recognition of the license plate; the third method is to use a behavior recognition model to determine whether there is water injection behavior. When processing large amounts of video data, the behavior recognition model not only has high computational complexity, but also consumes a lot of memory, which brings significant high resource usage problems. Summary of the Invention

[0004] The present invention provides an early warning method and device for a commercial concrete truck water injection detection system, which are used to achieve real-time monitoring of the water injection area of ​​the commercial concrete truck in different scenarios by selecting a suitable vehicle-mounted camera and an installation position. The method effectively improves the accuracy of water injection behavior identification by performing water injection behavior identification and analysis in the water injection area and performing an alarm operation when water injection behavior exists. When the alarm operation is triggered, an upload mechanism is automatically started to upload data, which can effectively ensure the traceability of the data, thereby effectively realizing data management and improving the effectiveness and accuracy of commercial concrete truck water injection detection.

[0005] The present invention provides an early warning method for a commercial concrete truck water injection detection system, comprising:

[0006] Step 1: Position and arrange the camera device on the commercial concrete truck;

[0007] Step 2: Collect video stream data in real time based on the positioning and placement results and transmit it to the vehicle-mounted edge device for image processing;

[0008] Step 3: Use the processed image data to calibrate the water injection area according to the current working environment;

[0009] Step 4: Identify and analyze water injection behavior in the water injection area, and perform alarm operations when water injection behavior occurs;

[0010] Step 5: When the alarm operation is triggered, the upload mechanism is automatically started to upload data.

[0011] Preferably, an early warning method for a water injection detection system of a commercial concrete truck, in step 1, positioning and arranging a camera device on the commercial concrete truck, includes:

[0012] Obtaining structural characteristics of the commercial concrete truck and determining a water injection area of ​​the commercial concrete truck based on the structural characteristics;

[0013] Determine the overall position distribution of the water injection area on the commercial concrete truck, and determine the effective monitoring points when collecting images of the water injection area based on the overall position distribution and structural characteristics;

[0014] Pre-collect the shooting angle and shooting range of each effective monitoring point based on the camera device, and determine the effective recording range and position of the water injection area of ​​the commercial concrete truck under each effective monitoring point in the collected image based on the pre-collection results;

[0015] Based on the effective recording range and position, the effective monitoring points are screened according to the preset arrangement quantity, and the camera device is positioned and arranged based on the screening result.

[0016] Preferably, an early warning method for a commercial concrete truck water injection detection system positions and arranges a camera device based on a screening result, comprising:

[0017] Allocate a communication route for the positioned camera device in the vehicle edge device, and determine the bandwidth requirement of the camera device for the communication route based on the service function of the camera device;

[0018] Adapt the communication parameters of the communication route based on bandwidth requirements, and conduct collaborative communication tests between the camera device and the vehicle-mounted edge device after successful communication parameter adaptation;

[0019] After the cooperative communication test is passed, it is determined that personalized working parameters of the camera device are adapted based on the image acquisition requirements, and the camera device is authorized to take effect after the adaptation is successful.

[0020] Preferably, an early warning method for a commercial concrete truck water injection detection system, in step 4, performing water injection behavior identification and analysis in the water injection area, and performing an alarm operation when water injection behavior occurs includes:

[0021] Get the video frame data of each timestamp in the video stream data;

[0022] Build a detection model, process the detection model in the vehicle edge device based on weight parameters and activation values, and deploy the detection model in the vehicle edge device based on the processing results;

[0023] Based on the on-board edge device, the time-stamped video frame data is sequentially transmitted to the detection model to identify the water injection personnel;

[0024] When the detection model detects the presence of water injection personnel in the water injection area, an alarm operation is performed.

[0025] Preferably, an early warning method for a commercial concrete truck water injection detection system constructs a detection model, including:

[0026] Capture the video stream data of commercial concrete trucks operating in construction sites in multiple environments, and create a training dataset for personnel in commercial concrete truck water injection scenarios based on the video stream data.

[0027] Label the personnel targets in the commercial concrete truck water injection scene personnel training dataset, where the annotation content is the position information and category of the visible area of ​​the personnel targets;

[0028] The commercial concrete truck water injection scene personnel training data set is input into the backbone network to extract image features, and the image features are input into the target algorithm model detection head to obtain the detection target;

[0029] According to the detection target and the marked personnel target, the target loss function is called to optimize the parameters of the backbone network and the target algorithm model detection head to obtain the final detection model.

[0030] Preferably, an early warning method of a commercial concrete truck water injection detection system, when identifying water injection personnel, includes:

[0031] When the detection model detects the presence of a target person, it marks the location information of the target person and determines the target frame of all the target persons;

[0032] Get the marked water injection area;

[0033] Calculate the target area ratio based on the target frame and water injection area;

[0034] ;

[0035] in, Indicates the target area ratio; Indicates the The coordinate data of the target frame; Indicates the Coordinate data of the marked area; Represents the intersection area of ​​the target frame and the water injection area; Indicates the The width of the target box; Indicates the height of the i-th target box;

[0036] Obtain a preset benchmark threshold, and compare the target area ratio with the preset benchmark threshold to determine whether the target person is a water injection person;

[0037] When the target area ratio is greater than the preset benchmark threshold, the target person is determined to be a water injection person;

[0038] Otherwise, the target person is determined to be an irrelevant person and is filtered out from the video data.

[0039] Preferably, an early warning method of a commercial concrete truck water injection detection system performs an alarm operation when a detection model detects the presence of a water injection person in the water injection area, including:

[0040] When the detection model detects that there is a water injection person in the water injection area, the current frame video image in which the water injection person is detected is read and the current frame video image is used as the starting recording frame;

[0041] Obtaining a first preset time interval, and performing frame-by-frame statistics on the detection results of the video data within the first preset time interval according to the starting recording frame;

[0042] Determine the total number of first target appearances of the water injection personnel in each video frame within a first preset time interval according to the statistical results, and trigger a first alarm operation when the total number of first target appearances is equal to or greater than a first preset threshold;

[0043] When the first alarm operation is triggered, a second preset time interval is obtained, wherein the second preset time interval includes a plurality of first preset time intervals;

[0044] The total number of second targets that trigger the first alarm operation is recorded within the second preset time interval according to the starting recording frame, and when the total number of second targets is equal to or greater than the second preset threshold, it is determined that there is water injection behavior, and the second alarm operation is performed according to the determination result, and the determination result is fed back to the monitoring terminal.

[0045] Preferably, in an early warning method for a commercial concrete truck water injection detection system, in step 5, when an alarm operation is triggered, an upload mechanism is automatically started to upload data, including:

[0046] When the alarm operation is triggered, the data to be uploaded is read, wherein the data to be uploaded includes: saving the video clip of the water injection behavior in the video stream data; recording the time point of the water injection behavior and using the time point as the timestamp of the video clip; locating the real-time geographic location of the commercial concrete truck, determining the geographic information, and at the same time, determining the license plate number of the commercial concrete truck;

[0047] Start the upload mechanism to upload the data to be uploaded to the management terminal.

[0048] Preferably, an early warning method for a commercial concrete truck water injection detection system, in step 2, real-time video stream data is collected according to the positioning arrangement result and transmitted to the vehicle-mounted edge device for image processing, including:

[0049] Controlling the camera device to collect video stream data in real time based on the positioning arrangement result, and adding a data source to the video stream data based on the physical location of the camera device;

[0050] The video stream data with the added data source is transmitted to the vehicle edge device, and the video stream data is framed based on the vehicle edge device to obtain an image frame sequence;

[0051] At the same time, the video stream data is time-traced, and the timestamp of each image frame is obtained based on the time-tracing result;

[0052] Add the timestamp to the corresponding image frame to obtain the target image frame sequence;

[0053] Based on the time development order, the pixels of each image frame in the target image frame sequence are traversed in turn to obtain the pixel features of each image frame, and the pixel features are analyzed in multiple dimensions based on the preset processing standards to obtain the preprocessing parameters of each image frame in each dimension;

[0054] Each image frame is preprocessed in turn based on the preprocessing parameters, wherein the preprocessing includes illumination equalization, defogging enhancement, and color correction optimization.

[0055] The present invention provides an early warning device for a water injection detection system of a commercial concrete truck, comprising:

[0056] Device positioning and arrangement module, used to position and arrange the camera device on the commercial concrete truck;

[0057] Image processing module, used to collect video stream data in real time based on positioning layout results and transmit it to the vehicle edge device for image processing;

[0058] The calibration module is used to calibrate the water injection area based on the processed image data according to the current working environment;

[0059] An alarm module is used to identify and analyze water injection behavior in the water injection area and to issue an alarm when water injection behavior occurs;

[0060] The data upload module is used to automatically start the upload mechanism to upload data when the alarm operation is triggered.

[0061] Compared with the prior art, the present invention has the following beneficial effects:

[0062] By selecting appropriate on-board cameras and installation locations, real-time monitoring of the water injection area of ​​commercial concrete trucks in different scenarios can be achieved. By identifying and analyzing water injection behavior in the water injection area and performing alarm operations when water injection behavior occurs, the accuracy of water injection behavior identification can be effectively improved. When the alarm operation is triggered, the upload mechanism is automatically started to upload data, which can effectively ensure the traceability of the data, thereby effectively realizing data management and improving the effectiveness and accuracy of commercial concrete truck water injection detection.

[0063] Other features and advantages of the present invention will be described in the following description, and in part will become apparent from the description, or will be understood by practicing the present invention. The purpose and other advantages of the present invention can be achieved and obtained through the structures specifically pointed out in this application document.

[0064] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0066] Figure 1 This is a flow chart of an early warning method for a commercial concrete truck water injection detection system according to an embodiment of the present invention;

[0067] Figure 2 This is a logic diagram of an early warning method of a commercial concrete truck water injection detection system according to an embodiment of the present invention;

[0068] Figure 3 A schematic diagram of the installation position of the camera device in an embodiment of the present invention;

[0069] Figure 4 This is a schematic diagram of the water injection area in an embodiment of the present invention;

[0070] Figure 5 This is a schematic diagram of alarm statistics in an embodiment of the present invention;

[0071] Figure 6 This is a structural diagram of an early warning device of a commercial concrete truck water injection detection system in an embodiment of the present invention. DETAILED DESCRIPTION

[0072] The preferred embodiments of the present invention are described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0073] Example 1:

[0074] This embodiment provides an early warning method for a commercial concrete truck water injection detection system. Figure 1-Figure 2 As shown, including:

[0075] Step 1: Position and arrange the camera device on the commercial concrete truck;

[0076] Select a high-resolution camera and install it in a position that can fully cover the water injection area of ​​the commercial concrete truck. Adjust it according to different vehicle models and working environments to provide a solid foundation for the next step of real-time image acquisition and preprocessing, and water injection area calibration.

[0077] Step 2: Collect video stream data in real time based on the positioning and placement results and transmit it to the vehicle-mounted edge device for image processing;

[0078] Real-time image acquisition and preprocessing are key steps in the commercial concrete truck water injection detection system. The camera installed in step 1 obtains real-time vehicle monitoring data, and the video stream captured by the camera is transmitted stably and with low latency to the low-power on-board edge device. The on-board edge device is used to preprocess each frame of the image to ensure the quality of the input data.

[0079] Step 3: Use the processed image data to calibrate the water injection area according to the current working environment;

[0080] After the camera is installed, the water injection detection system will pre-define and calibrate the water injection areas on the left and right sides of the commercial concrete truck. After obtaining real-time input data through step 2, the calibration of the water injection area is remotely adjusted using the cloud platform according to the current working environment. It also supports dynamic setting of polygon vertices to improve the accuracy of water injection behavior recognition. The camera is installed as follows: Figure 3 shown.

[0081] Step 4: Identify and analyze water injection behavior in the water injection area, and perform alarm operations when water injection behavior occurs;

[0082] After obtaining the real-time video stream through step 2, water injection behavior recognition analysis is performed to determine whether there is water injection behavior. Water injection behavior recognition first detects the water injection personnel from a single frame image through the target detection algorithm (an improved YOLOv5 detection algorithm), and then counts the detection results of multiple frames in a short period of time. After confirming that there is water injection behavior in a short period of time, an alarm is triggered. Finally, the cumulative statistics of multiple short-term alarm results are performed over a long period of time, and the final water injection behavior recognition result is output. The water injection area annotation example is shown in the figure below. Figure 4 shown.

[0083] Step 5: When the alarm operation is triggered, the upload mechanism is automatically started to upload data.

[0084] Once water injection is confirmed, the early warning mechanism is immediately triggered, and the background upload mechanism is automatically activated. This not only uploads video footage of the water injection, but also records key data such as time, geographic information, and license plate numbers, achieving a comprehensive record of uploaded data. This information is integrated into a structured data packet for efficient transmission and storage. The data is uploaded to the cloud platform for classification and archiving, facilitating subsequent retrieval and review. Administrators can use the cloud platform to view historical records, export reports, and conduct data analysis, providing deeper support and optimization for monitoring and management. After processing a time interval of T seconds, the system waits for the next frame to arrive, then returns to the image acquisition and preprocessing steps and continues the above process. The entire process forms a closed-loop system, ensuring continuous and efficient monitoring and early warning capabilities, and maintaining the system's real-time response characteristics.

[0085] The beneficial effects of the above technical solution are: by selecting appropriate on-board cameras and installation locations, real-time monitoring of the water injection area of ​​commercial concrete trucks in different scenarios can be achieved; by identifying and analyzing water injection behavior in the water injection area and performing alarm operations when water injection behavior occurs, the accuracy of water injection behavior identification can be effectively improved; when the alarm operation is triggered, the upload mechanism is automatically started to upload data, which can effectively ensure the traceability of the data, thereby effectively realizing data management and improving the effectiveness and accuracy of water injection detection of commercial concrete trucks.

[0086] Example 2:

[0087] Based on Example 1, this embodiment provides an early warning method for a water injection detection system of a commercial concrete truck. In step 1, positioning and arranging a camera device on the commercial concrete truck includes:

[0088] Obtaining structural characteristics of the commercial concrete truck and determining a water injection area of ​​the commercial concrete truck based on the structural characteristics;

[0089] Determine the overall position distribution of the water injection area on the commercial concrete truck, and determine the effective monitoring points for image acquisition of the water injection area based on the overall position distribution and structural characteristics;

[0090] Pre-collect the shooting angle and shooting range of each effective monitoring point based on the camera device, and determine the effective recording range and position of the water injection area of ​​the commercial concrete truck under each effective monitoring point in the collected image based on the pre-collection results;

[0091] Based on the effective recording range and position, the effective monitoring points are screened according to the preset arrangement quantity, and the camera device is positioned and arranged based on the screening result.

[0092] In this embodiment, the structural features refer to the distribution of structural components of the commercial concrete truck, so that the water injection area of ​​the commercial concrete truck can be determined based on the structural features.

[0093] In this embodiment, the effective monitoring point refers to a position point when capturing images of the water injection area, that is, a point where comprehensive and effective image capture of the water injection area can be performed, and the effective monitoring point is not unique.

[0094] In this embodiment, pre-collection refers to determining the shooting angle and shooting range of each effective monitoring point by a camera device.

[0095] In this embodiment, screening the effective monitoring points based on the effective recording range and position according to the preset arrangement number means screening out the points with the best effective recording range and position from the effective monitoring points, that is, screening out the points that can comprehensively collect water injection areas and whose collection angles meet the requirements.

[0096] The beneficial effect of the above technical solution is: by analyzing the structural characteristics of the commercial concrete truck, the water injection area can be effectively determined, and then the effective monitoring points can be determined according to the water injection area, and the determined effective monitoring points can be screened, ensuring the effective deployment of the camera device and providing a guarantee for the detection of the water injection area.

[0097] Example 3:

[0098] Based on Example 2, this embodiment provides an early warning method for a commercial concrete truck water injection detection system, which positions and arranges the camera device based on the screening results, including:

[0099] Allocate a communication route for the positioned camera device in the vehicle edge device, and determine the bandwidth requirement of the camera device for the communication route based on the service function of the camera device;

[0100] Adapt the communication parameters of the communication route based on bandwidth requirements, and conduct collaborative communication tests between the camera device and the vehicle-mounted edge device after successful communication parameter adaptation;

[0101] After the cooperative communication test is passed, it is determined that personalized working parameters of the camera device are adapted based on the image acquisition requirements, and the camera device is authorized to take effect after the adaptation is successful.

[0102] In this embodiment, the service function refers to the type and requirement of the service performed by the camera device, so as to facilitate determination of the bandwidth requirement of the communication route according to the service function.

[0103] In this embodiment, the collaborative communication test refers to testing the communication between the camera device and the vehicle-mounted edge device, that is, ensuring that effective communication can be carried out between the two.

[0104] In this embodiment, personalized working parameter adaptation refers to configuring the resolution of the camera device according to camera requirements or image acquisition requirements.

[0105] The beneficial effects of the above technical solution are: by adapting the communication parameters and personalized working parameters of the camera device, the working reliability of the camera device is ensured, and the comprehensiveness and effectiveness of image acquisition of the water injection area are improved.

[0106] Example 4:

[0107] Based on Example 1, this embodiment provides an early warning method for a commercial concrete truck water injection detection system. In step 4, water injection behavior identification and analysis are performed in the water injection area. When water injection behavior occurs, an alarm operation is performed, including:

[0108] Get the video frame data of each timestamp in the video stream data;

[0109] Build a detection model, process the detection model in the vehicle edge device based on weight parameters and activation values, and deploy the detection model in the vehicle edge device based on the processing results;

[0110] Based on the on-board edge device, the time-stamped video frame data is sequentially transmitted to the detection model to identify the water injection personnel;

[0111] When the detection model detects the presence of water injection personnel in the water injection area, an alarm operation is performed.

[0112] In this embodiment, a detection model is constructed, including: capturing operation video stream data of commercial concrete trucks in construction sites of multiple environmental categories, and creating a personnel training data set for commercial concrete truck water injection scenarios based on the operation video stream data; marking personnel targets in the commercial concrete truck water injection scenario personnel training data set, wherein the marking content is the visible area location information and category of the personnel targets; inputting the commercial concrete truck water injection scenario personnel training data set into the backbone network to extract image features, and inputting the image features into the target algorithm model detection head to obtain the detection target; calling the target loss function according to the detection target and the marked personnel targets to optimize the parameters of the backbone network and the target algorithm model detection head to obtain the final detection model.

[0113] In this embodiment, the annotation content is the location information of the visible area of ​​the person target and the category, wherein the location information of the visible area of ​​the person target is the coordinate information of the minimum bounding rectangle of the visible area of ​​the person target. The category is the target label category, which is named watering_personnel.

[0114] Based on the above: After using the improved YOLOv5 algorithm to analyze the input image, the real water injection target of the current video frame is identified through the calibration area, and the water injection personnel information is output. Before using the improved YOLOv5 algorithm, it is necessary to collect real car scene samples and train the improved YOLOv5 algorithm model. After the training is completed, the model is deployed on the edge device using the asymmetric quantization method. The image data of step 3.2 is sent to the edge device, and the image data is analyzed and detected using the quantized model to preliminarily obtain the location information of all detected personnel, and further filter the targets in the non-calibrated area. The specific steps are as follows:

[0115] Collecting data sets: Since there are large differences between the personnel data on commercial concrete trucks and the pedestrian data in public data sets, and there is a phenomenon that personnel are partially obscured by the concrete body, it is necessary to collect data in this scenario for model fine-tuning. During the collection stage, real construction sites were selected as data collection locations, covering extreme conditions such as day and night, rain and fog, and dust, to ensure the diversity and representativeness of the data set. The on-board camera was used to capture the video stream during the operation of the commercial concrete truck in a specific manner, covering at least 336 hours of construction site operations, and to create a personnel detection dataset for the commercial concrete truck water injection scene. In the collected video stream, all the personnel targets that appeared were manually annotated, and a total of 45,100 targets were annotated, of which about 90% of the personnel were obscured to varying degrees. "Each annotation records the location information and category of the visible area of ​​the personnel target.

[0116] Model training: The training samples are sent to the ShuffleNetV2 backbone network to extract image feature information. The feature information is then sent to the improved YOLOv5 algorithm model detection head to obtain the detection target. Finally, the improved Focal-EIoU loss function is used to optimize the model parameters to obtain the final training model. IoU is a basic indicator for measuring the degree of overlap between the predicted box and the true box. It represents the ratio of the intersection and union between the predicted box and the true box. EIoU and Focal-EIoU are both extended forms of IoU. The formula of the Focal-EIoU loss function is as follows:

[0117]

[0118]

[0119]

[0120] in represents the loss function of Focal-EIoU, It is a hyperparameter used to control the curvature of the curve. represents the loss function of EIoU; and Represent the coordinates of the detection box and the real box respectively; represents the Euclidean distance, Represent the center points of the predicted box and the real box respectively, represents the diagonal distance of the minimum enclosing rectangle, Represents the width of the predicted box and the real box respectively, Indicates the width of the minimum enclosing rectangle, Represents the height of the predicted box and the real box respectively, Indicates the height of the minimum enclosing rectangle.

[0121] In this embodiment, when deploying on edge devices, it includes: obtaining the original training model, and performing asymmetric quantization on the weight parameters and activation values ​​when deploying on the edge device, so as to further improve computing efficiency and reduce resource usage while maintaining high precision. Specifically, the asymmetric quantization method allows for more flexible mapping of floating-point values ​​to integer ranges during the quantization process, thereby effectively reducing quantization errors and retaining the accuracy of the original model to the maximum extent; retaining 12-bit integer operations in the forward reasoning stage to replace the traditional 8-bit quantization scheme, avoiding the problem of a significant decrease in accuracy caused by 8-bit quantization. The 12-bit quantization strategy provides higher numerical representation accuracy, controls the accuracy loss after model quantization within a very small range, and ensures the high accuracy of the model in practical applications. While minimizing the accuracy loss after model quantization, this solution can still meet real-time requirements, providing cost-effective algorithm support for commercial concrete truck water injection detection.

[0122] In this embodiment, the timestamp video frame data refers to the frame data obtained by splitting the video stream data according to the time sequence, and the frame data corresponds to the time sequence one by one.

[0123] In this embodiment, the activation value refers to the configuration parameters required when the detection model is activated.

[0124] The beneficial effect of the above technical solution is: by splitting the video stream data into time-stamped video frame data, and inputting each time-stamped video frame data into the constructed detection model to identify the water injection personnel, it is possible to accurately and effectively identify whether there are water injection personnel in the water injection area, and perform corresponding alarm operations when there are water injection personnel, thereby improving the accuracy and reliability of commercial concrete truck water injection detection.

[0125] Example 5:

[0126] Based on Example 4, this embodiment provides an early warning method for a commercial concrete truck water injection detection system, which includes:

[0127] When the detection model detects the presence of a target person, it marks the location information of the target person and determines the target frame of all the target persons;

[0128] Get the marked water injection area;

[0129] Calculate the target area ratio based on the target frame and water injection area;

[0130] ;

[0131] in, Indicates the target area ratio; Indicates the The coordinate data of the target frame; Indicates the Coordinate data of the marked area; Represents the intersection area of ​​the target frame and the water injection area; Indicates the The width of the target box; Indicates the height of the i-th target box;

[0132] Obtain a preset benchmark threshold, and compare the target area ratio with the preset benchmark threshold to determine whether the target person is a water injection person;

[0133] When the target area ratio is greater than the preset benchmark threshold, the target person is determined to be a water injection person;

[0134] Otherwise, the target person is determined to be an irrelevant person and is filtered out from the video data.

[0135] In this embodiment, water injection personnel filtering is a method that uses improved target detection to obtain detection results and then uses a threshold judgment based on the overlap area ratio calculated from the calibration area to filter out irrelevant personnel. Camera images often capture water injection personnel as well as other pedestrians on the road, necessitating filtering out irrelevant personnel.

[0136] In this embodiment, the target frame refers to a marking frame obtained after marking the target person.

[0137] In this embodiment, the preset reference threshold is set in advance and is used as a basis for measuring whether the target area ratio meets the requirements.

[0138] The beneficial effect of the above technical solution is: by marking the target personnel in the water injection area, and judging whether the target personnel are water injection personnel based on the marking results, and when the target personnel are irrelevant personnel, the irrelevant personnel are filtered out, thereby ensuring the accuracy and reliability of water injection detection of commercial concrete trucks.

[0139] Example 6:

[0140] Based on Example 4, this embodiment provides an early warning method for a commercial concrete truck water injection detection system. When the detection model detects the presence of a water injection person in the water injection area, an alarm operation is performed, including:

[0141] When the detection model detects that there is a water injection person in the water injection area, the current frame video image in which the water injection person is detected is read and the current frame video image is used as the starting recording frame;

[0142] Obtaining a first preset time interval, and performing frame-by-frame statistics on the detection results of the video data within the first preset time interval according to the starting recording frame;

[0143] Determine the total number of first target appearances of the water injection personnel in each video frame within a first preset time interval according to the statistical results, and trigger a first alarm operation when the total number of first target appearances is equal to or greater than a first preset threshold;

[0144] When the first alarm operation is triggered, a second preset time interval is obtained, wherein the second preset time interval includes a plurality of first preset time intervals;

[0145] The total number of second targets that trigger the first alarm operation is recorded within the second preset time interval according to the starting recording frame, and when the total number of second targets is equal to or greater than the second preset threshold, it is determined that there is water injection behavior, and the second alarm operation is performed according to the determination result, and the determination result is fed back to the monitoring terminal.

[0146] In this embodiment, short-time (i.e., first preset time interval) multi-frame detection statistics are a method of performing multi-frame statistics based on the step-by-step single-frame detection results to obtain a stable result output. When the number of frames containing water injection personnel detected reaches a set threshold, it is considered that water injection behavior exists this time. In a short time (e.g., 1 second), the system inputs 6 frames of data into the system at equal intervals in sequence, and counts the number of frames containing water injection personnel detected. When the number of detected frames reaches or exceeds 4 frames (i.e., the first preset threshold), it is considered that water injection behavior exists in a short period of time, and an alarm is triggered. Long-time (i.e., second preset time interval) cumulative alarm statistics are performed over a longer period of time, and the number of consecutive short-time alarms is cumulatively counted. When the set threshold is reached or exceeded, the system determines that water injection behavior is detected and issues a final water injection behavior alarm. For example, if the long time is set to 5 seconds, according to the cumulative statistics of the number of alarms per 1 second of the short time, when the long detection time reaches 5 seconds (i.e., the second preset threshold), when the number of alarms reaches or exceeds 4 times, the system will finally confirm the existence of water injection behavior and output a reliable water injection detection result. The specific alarm statistics are as follows: Figure 5 shown.

[0147] In this embodiment, the first preset time interval is set in advance.

[0148] In this embodiment, the first target total number of times refers to the total number of times the water injection personnel appear in all video frames within the first preset time zone.

[0149] In this embodiment, the first preset threshold is set in advance and is a reference for determining whether a first alarm is required.

[0150] In this embodiment, the second preset time interval is set in advance.

[0151] In this embodiment, the second preset threshold is set in advance and is used to measure whether the water injection personnel have engaged in water injection behavior.

[0152] The beneficial effect of the above technical solution is: by determining the number of times the detected water injection personnel appear in the water injection area, a first alarm operation is performed when the number of appearances is greater than a preset threshold, and simultaneously a judgment is made as to whether the water injection personnel have any water injection behavior, and then a second alarm is performed when there is water injection behavior, thereby improving the rigor and reliability of water injection detection of commercial concrete trucks.

[0153] Example 7:

[0154] Based on Example 1, this embodiment provides an early warning method for a commercial concrete truck water injection detection system. In step 5, when the alarm operation is triggered, the upload mechanism is automatically started to upload data, including:

[0155] When the alarm operation is triggered, the data to be uploaded is read, wherein the data to be uploaded includes: saving the video clip of the water injection behavior in the video stream data; recording the time point of the water injection behavior and using the time point as the timestamp of the video clip; locating the real-time geographic location of the commercial concrete truck, determining the geographic information, and at the same time, determining the license plate number of the commercial concrete truck;

[0156] Start the upload mechanism to upload the data to be uploaded to the management terminal.

[0157] In this embodiment, an early warning mechanism is triggered immediately upon confirmation of water injection, and a background upload mechanism is automatically started. This not only uploads video footage of the water injection, but also records key data such as time, geographic information, and license plate number, achieving comprehensive uploaded data records, specifically including the following:

[0158] (1) Preserving video footage of water injection behavior not only provides intuitive visual evidence but also provides rich material for subsequent detailed review and analysis;

[0159] (2) Record the exact time of the event and ensure that the timestamp of each video clip is accurate;

[0160] (3) Storing geographic information, obtaining and marking the real-time geographic location information of commercial concrete trucks through GPS or other positioning technologies, so as to facilitate tracking and management of vehicle behavior in different locations;

[0161] (4) Identify the license plate number of the commercial concrete truck to facilitate subsequent management and responsibility tracing.

[0162] The beneficial effect of the above technical solution is: ensuring that the data to be uploaded corresponding to the alarm operation is accurately and effectively uploaded to the management terminal, thereby facilitating the tracing of the alarm information and ensuring the rigor of the water injection detection of commercial concrete trucks.

[0163] Example 8:

[0164] Based on Example 1, this embodiment provides an early warning method for a commercial concrete truck water injection detection system. In step 2, video stream data is collected in real time based on the positioning layout results and transmitted to the vehicle-mounted edge device for image processing, including:

[0165] Controlling the camera device to collect video stream data in real time based on the positioning arrangement result, and adding a data source to the video stream data based on the physical location of the camera device;

[0166] The video stream data with the added data source is transmitted to the vehicle edge device, and the video stream data is framed based on the vehicle edge device to obtain an image frame sequence;

[0167] At the same time, the video stream data is time-traced, and the timestamp of each image frame is obtained based on the time-tracing result;

[0168] Add the timestamp to the corresponding image frame to obtain the target image frame sequence;

[0169] Based on the time development order, the pixels of each image frame in the target image frame sequence are traversed in turn to obtain the pixel features of each image frame, and the pixel features are analyzed in multiple dimensions based on the preset processing standards to obtain the preprocessing parameters of each image frame in each dimension;

[0170] Each image frame is preprocessed in turn based on the preprocessing parameters, wherein the preprocessing includes illumination equalization, defogging enhancement, and color correction optimization.

[0171] In this embodiment, adding a data source to the video stream data based on the physical location of the camera device refers to performing terminal differentiation on the video stream data, that is, being able to determine the camera devices corresponding to different video stream data.

[0172] In this embodiment, the target image frame sequence refers to a result obtained by adding the specific time corresponding to each image frame to the corresponding image frame.

[0173] In this embodiment, the pixel feature refers to the pixel value and corresponding resolution of each image frame.

[0174] In this embodiment, the preset processing standard is set in advance, including the parsing dimension of the pixel features and the corresponding parsing standard.

[0175] In this embodiment, the pre-processing parameters refer to specific parameters that need to be processed in each dimension of each image frame, such as the adjustment amount of image brightness and the adjustment amount of resolution.

[0176] The beneficial effect of the above technical solution is: by transmitting the collected video stream data to the on-board edge device, and analyzing and processing the video stream data through the on-board edge device, the preprocessing parameters of each image frame in each dimension can be accurately and effectively determined, thereby achieving reliable preprocessing of each image frame, providing convenience and basis for water injection detection of commercial concrete trucks.

[0177] Example 9:

[0178] Based on Example 1, this embodiment provides an early warning method for a commercial concrete truck water injection detection system. In step 3, the processed image data is calibrated according to the current working environment to determine the water injection area, including:

[0179] Obtain historical image data and analyze the historical image data to obtain a set of commercial concrete truck models and a set of operating environments;

[0180] Determine the corresponding benchmark water injection area from the preset instruction manual based on the commercial concrete truck model. At the same time, determine the user's behavioral characteristics when water injection occurs based on the commercial concrete truck water injection detection requirements, and determine the user's activity range within the benchmark water injection area based on the behavioral characteristics;

[0181] Performing a first correction on the reference water injection area based on the active range, and determining a first water injection area in the image data based on the first correction result;

[0182] At the same time, the operating environment is analyzed to obtain the interference factors on the first water injection area under different operating environments, and the interference range of the interference factors on the first water injection area is determined;

[0183] Performing a second correction on the first water injection area based on the interference range, and obtaining a second water injection area based on the second correction;

[0184] determining relative positions between different objects in the second water injection area, and determining an effective monitoring range in the second water injection area based on the relative positions;

[0185] Determine the dynamic configuration strategy of polygon vertices in different operating environments based on the effective monitoring range;

[0186] Based on the dynamic configuration strategy of polygon vertices, the water injection area in the processed image data is locked according to the current working environment, and calibration is performed based on the locked result.

[0187] In this embodiment, the preset instruction manual is known in advance and is used to record the water injection areas corresponding to commercial concrete trucks of different models.

[0188] In this embodiment, the water injection detection requirements for commercial concrete trucks are known in advance.

[0189] In this embodiment, the behavioral feature refers to the specific behavioral action of the user when the user is filling water.

[0190] In this embodiment, interference factors refer to factors that affect the determination of the first water injection area under different operating environments, such as obstacles.

[0191] In this embodiment, the polygon vertex dynamic configuration strategy refers to a strategy or scheme that considers the scope when determining the water injection area under different operating environments. For example, when there are obstacles interfering with the water injection area, the monitoring range of the water injection area can be adjusted according to the shape of the obstacle.

[0192] The beneficial effects of the above technical solution are: by analyzing historical image data, the vehicle model set and operating environment set of commercial concrete trucks can be determined, and then different vehicle models and operating environments can be analyzed to effectively formulate dynamic polygon vertex configuration strategies for commercial concrete trucks in different operating environments. Finally, the water injection area in the processed image data of the current operating environment can be locked according to the dynamic polygon vertex configuration strategy, ensuring the accuracy and reliability of the locking of the water injection area.

[0193] Example 10:

[0194] This embodiment provides an early warning device for a commercial concrete truck water injection detection system. Figure 6 As shown, including:

[0195] Device positioning and arrangement module, used to position and arrange the camera device on the commercial concrete truck;

[0196] Image processing module, used to collect video stream data in real time based on positioning layout results and transmit it to the vehicle edge device for image processing;

[0197] The calibration module is used to calibrate the water injection area based on the processed image data according to the current working environment;

[0198] An alarm module is used to identify and analyze water injection behavior in the water injection area and to issue an alarm when water injection behavior occurs;

[0199] The data upload module is used to automatically start the upload mechanism to upload data when the alarm operation is triggered.

[0200] The beneficial effects of the above technical solution are: by selecting appropriate on-board cameras and installation locations, real-time monitoring of the water injection area of ​​commercial concrete trucks in different scenarios can be achieved; by identifying and analyzing water injection behavior in the water injection area, and performing alarm operations when water injection behavior occurs, the accuracy of water injection behavior identification can be effectively improved; when the alarm operation is triggered, the upload mechanism is automatically started to upload data, which can effectively ensure the traceability of the data, thereby effectively realizing data management and improving the effectiveness and accuracy of water injection detection of commercial concrete trucks.

[0201] Obviously, those skilled in the art may make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if such changes and modifications fall within the scope of the claims and their equivalents, the present invention is intended to include such changes and modifications.

Claims

1. An early warning method for a commercial concrete truck water injection detection system, characterized in that: include: Step 1: Position and arrange the camera device on the commercial concrete truck; Step 2: Collect video stream data in real time based on the positioning and placement results and transmit it to the vehicle-mounted edge device for image processing; Step 3: Use the processed image data to calibrate the water injection area according to the current working environment; Step 4: Identify and analyze water injection behavior in the water injection area, and perform alarm operations when water injection behavior occurs; Step 5: When the alarm operation is triggered, the upload mechanism is automatically started to upload data; In step 3, the processed image data is calibrated according to the current working environment to determine the water injection area, including: Obtain historical image data and analyze the historical image data to obtain a set of commercial concrete truck models and a set of operating environments; Determine the corresponding benchmark water injection area from the preset instruction manual based on the commercial concrete truck model. At the same time, determine the user's behavioral characteristics when water injection occurs based on the commercial concrete truck water injection detection requirements, and determine the user's activity range within the benchmark water injection area based on the behavioral characteristics; Performing a first correction on the reference water injection area based on the active range, and determining a first water injection area in the image data based on the first correction result; At the same time, the operating environment is analyzed to obtain the interference factors on the first water injection area under different operating environments, and the interference range of the interference factors on the first water injection area is determined; Performing a second correction on the first water injection area based on the interference range, and obtaining a second water injection area based on the second correction; determining relative positions between different objects in the second water injection area, and determining an effective monitoring range in the second water injection area based on the relative positions; Determine the dynamic configuration strategy of polygon vertices in different operating environments based on the effective monitoring range; Based on the dynamic configuration strategy of polygon vertices, the water injection area in the processed image data is locked according to the current working environment, and calibration is performed based on the locked results; In step 2, based on the positioning results, video stream data is collected in real time and transmitted to the vehicle edge device for image processing, including: Controlling the camera device to collect video stream data in real time based on the positioning arrangement result, and adding a data source to the video stream data based on the physical location of the camera device; The video stream data with the added data source is transmitted to the vehicle edge device, and the video stream data is framed based on the vehicle edge device to obtain an image frame sequence; At the same time, the video stream data is time-traced, and the timestamp of each image frame is obtained based on the time-tracing result; Add the timestamp to the corresponding image frame to obtain the target image frame sequence; Based on the time development order, the pixels of each image frame in the target image frame sequence are traversed in turn to obtain the pixel features of each image frame, and the pixel features are analyzed in multiple dimensions based on the preset processing standards to obtain the preprocessing parameters of each image frame in each dimension; Each image frame is preprocessed in turn based on the preprocessing parameters, wherein the preprocessing includes illumination equalization, defogging enhancement, and color correction optimization.

2. The early warning method of a commercial concrete truck water injection detection system according to claim 1, characterized in that: In step 1, the camera device is positioned and arranged on the commercial concrete truck, including: Obtaining structural characteristics of the commercial concrete truck and determining a water injection area of ​​the commercial concrete truck based on the structural characteristics; Determine the overall position distribution of the water injection area on the commercial concrete truck, and determine the effective monitoring points for image acquisition of the water injection area based on the overall position distribution and structural characteristics; Pre-collect the shooting angle and shooting range of each effective monitoring point based on the camera device, and determine the effective recording range and position of the water injection area of ​​the commercial concrete truck under each effective monitoring point in the collected image based on the pre-collection results; Based on the effective recording range and position, the effective monitoring points are screened according to the preset arrangement quantity, and the camera device is positioned and arranged based on the screening result.

3. The early warning method of a commercial concrete truck water injection detection system according to claim 2, characterized in that: Positioning and arranging the camera device based on the screening results includes: Allocate a communication route for the positioned camera device in the vehicle edge device, and determine the bandwidth requirement of the camera device for the communication route based on the service function of the camera device; Adapt the communication parameters of the communication route based on bandwidth requirements, and conduct collaborative communication tests between the camera device and the vehicle-mounted edge device after successful communication parameter adaptation; After the cooperative communication test is passed, it is determined that personalized working parameters of the camera device are adapted based on the image acquisition requirements, and the camera device is authorized to take effect after the adaptation is successful.

4. The early warning method of a commercial concrete truck water injection detection system according to claim 1, characterized in that: In step 4, water injection behavior identification and analysis is performed in the water injection area. When water injection behavior occurs, alarm operations are performed including: Get the video frame data of each timestamp in the video stream data; Build a detection model, process the detection model in the vehicle edge device based on weight parameters and activation values, and deploy the detection model in the vehicle edge device based on the processing results; Based on the on-board edge device, the time-stamped video frame data is sequentially transmitted to the detection model to identify the water injection personnel; When the detection model detects the presence of water injection personnel in the water injection area, an alarm operation is performed.

5. The early warning method of a commercial concrete truck water injection detection system according to claim 4, characterized in that: Build a detection model, including: Capture the video stream data of commercial concrete trucks operating in construction sites in multiple environments, and create a training dataset for personnel in commercial concrete truck water injection scenarios based on the video stream data. Label the personnel targets in the commercial concrete truck water injection scene personnel training dataset, where the annotation content includes the visible area location information and category of the personnel targets; The commercial concrete truck water injection scene personnel training data set is input into the backbone network to extract image features, and the image features are input into the target algorithm model detection head to obtain the detection target; According to the detection target and the marked personnel target, the target loss function is called to optimize the parameters of the backbone network and the target algorithm model detection head to obtain the final detection model.

6. The early warning method of a commercial concrete truck water injection detection system according to claim 4, characterized in that: When identifying water injection personnel, include: When the detection model detects the presence of a target person, it marks the location information of the target person and determines the target frame of all the target persons; Get the marked water injection area; Calculate the target area ratio based on the target frame and water injection area; ； in, Indicates the target area ratio; Indicates the The coordinate data of the target frame; Indicates the Coordinate data of the marked area; Represents the intersection area of ​​the target frame and the water injection area; Indicates the The width of the target box; Indicates the height of the i-th target box; Obtain a preset benchmark threshold, and compare the target area ratio with the preset benchmark threshold to determine whether the target person is a water injection person; When the target area ratio is greater than the preset benchmark threshold, the target person is determined to be a water injection person; Otherwise, the target person is determined to be an irrelevant person and is filtered out from the video data.

7. The early warning method of a commercial concrete truck water injection detection system according to claim 4, characterized in that: When the detection model detects the presence of water injection personnel in the water injection area, an alarm operation is performed, including: When the detection model detects that there is a water injection person in the water injection area, the current frame video image in which the water injection person is detected is read and the current frame video image is used as the starting recording frame; Obtaining a first preset time interval, and performing frame-by-frame statistics on the detection results of the video data within the first preset time interval according to the starting recording frame; Determine the total number of first target appearances of the water injection personnel in each video frame within a first preset time interval according to the statistical results, and trigger a first alarm operation when the total number of first target appearances is equal to or greater than a first preset threshold; When the first alarm operation is triggered, a second preset time interval is obtained, wherein the second preset time interval includes a plurality of first preset time intervals; The total number of second targets that trigger the first alarm operation is recorded within the second preset time interval according to the starting recording frame, and when the total number of second targets is equal to or greater than the second preset threshold, it is determined that there is water injection behavior, and the second alarm operation is performed according to the determination result, and the determination result is fed back to the monitoring terminal.

8. The early warning method of a commercial concrete truck water injection detection system according to claim 1, characterized in that: In step 5, when the alarm operation is triggered, the upload mechanism is automatically started to upload data, including: When the alarm operation is triggered, the data to be uploaded is read, wherein the data to be uploaded includes: saving the video clip of the water injection behavior in the video stream data; recording the time point of the water injection behavior and using the time point as the timestamp of the video clip; locating the real-time geographic location of the commercial concrete truck, determining the geographic information, and at the same time, determining the license plate number of the commercial concrete truck; Start the upload mechanism to upload the data to be uploaded to the management terminal.

9. An early warning device for a water injection detection system of a commercial concrete truck, characterized in that: include: Device positioning and arrangement module, used to position and arrange the camera device on the commercial concrete truck; Image processing module, used to collect video stream data in real time based on positioning layout results and transmit it to the vehicle edge device for image processing; The calibration module is used to calibrate the water injection area based on the processed image data according to the current working environment; An alarm module is used to identify and analyze water injection behavior in the water injection area and to issue an alarm when water injection behavior occurs; Data upload module, used to automatically start the upload mechanism to upload data when the alarm operation is triggered; In the calibration module, the processed image data is calibrated according to the current working environment to determine the water injection area, including: Obtain historical image data and analyze the historical image data to obtain a set of commercial concrete truck models and a set of operating environments; Determine the corresponding benchmark water injection area from the preset instruction manual based on the commercial concrete truck model. At the same time, determine the user's behavioral characteristics when water injection occurs based on the commercial concrete truck water injection detection requirements, and determine the user's activity range within the benchmark water injection area based on the behavioral characteristics; Performing a first correction on the reference water injection area based on the active range, and determining a first water injection area in the image data based on the first correction result; At the same time, the operating environment is analyzed to obtain the interference factors on the first water injection area under different operating environments, and the interference range of the interference factors on the first water injection area is determined; Performing a second correction on the first water injection area based on the interference range, and obtaining a second water injection area based on the second correction; determining relative positions between different objects in the second water injection area, and determining an effective monitoring range in the second water injection area based on the relative positions; Determine the dynamic configuration strategy of polygon vertices in different operating environments based on the effective monitoring range; Based on the dynamic configuration strategy of polygon vertices, the water injection area in the processed image data is locked according to the current working environment, and calibration is performed based on the locked results; In the image processing module, video stream data is collected in real time based on the positioning layout results and transmitted to the vehicle edge device for image processing, including: Controlling the camera device to collect video stream data in real time based on the positioning arrangement result, and adding a data source to the video stream data based on the physical location of the camera device; The video stream data with the added data source is transmitted to the vehicle edge device, and the video stream data is framed based on the vehicle edge device to obtain an image frame sequence; At the same time, the video stream data is time-traced, and the timestamp of each image frame is obtained based on the time-tracing result; Add the timestamp to the corresponding image frame to obtain the target image frame sequence; Based on the time development order, the pixels of each image frame in the target image frame sequence are traversed in turn to obtain the pixel features of each image frame, and the pixel features are analyzed in multiple dimensions based on the preset processing standards to obtain the preprocessing parameters of each image frame in each dimension; Each image frame is preprocessed in turn based on the preprocessing parameters, wherein the preprocessing includes illumination equalization, defogging enhancement, and color correction optimization.

Images