A construction safety detection system based on color quantization

Abstract

The invention belongs to the technical field of construction monitoring and early warning, and specifically discloses a construction safety detection system based on color quantization, comprising: an image receiving unit for acquiring image data of a construction site taken by a camera; the camera is also connected to an image display unit; a color calibration unit calibrates the image data received by the image receiving unit through reference color information; an image division unit is used to divide a non-safe area corresponding to the construction site on a scene image corresponding to the image data; a target tracking unit is used to identify a moving object in a scene image at adjacent moments, and judge whether to generate an alarm message according to the motion state of the moving object in the non-safe area; when the alarm message is generated, a real-time early warning unit marks the alarm message on the image data displayed by the image display unit through associated information; the system has the following advantages: improved color recognition accuracy, real-time monitoring and alarm response, reduced safety accidents, and rapid handling of emergency situations.

Description

Technical Field

[0001] The present invention relates to the technical field of construction monitoring and early warning, and in particular to a building construction safety detection system based on color quantization. Background Art

[0002] Construction refers to the production activities in the implementation stage of engineering construction. It is the construction process of various buildings. It can also be said that it is the process of turning various lines on the design drawings into physical objects at designated locations. During the construction process, people's focus is always on project quality, safety (including environmental protection), progress and cost. However, the existing online risk detection system for construction quality and safety is inaccurate for wall detection; at the same time, the tilt angle is manually analyzed and judged by specialized monitoring personnel, which will not only sharply increase the labor cost of the building monitoring system, but also cause calculation deviations due to the subjectivity and inaccuracy of manual judgment. It also cannot guarantee the accuracy of the data analysis results, which may lead to the collapse of the building.

[0003] In order to solve the defects of the prior art, a search revealed that a construction quality and safety online risk detection system and method with publication number CN117073628A includes: a foundation detection module, a curtain wall detection module, a main body quality detection module, a building inclination detection module, a building door and window detection module, a data comparison module, a data evaluation module, and an output module; an intelligent control module is connected to the foundation detection module, the curtain wall detection module, the main body quality detection module, the building inclination detection module, the building door and window detection module, the data comparison module, the data evaluation module, and the output module, and is used to control each module to work normally through a PLC. The present application greatly improves the detection accuracy; each module is controlled by a PLC to detect each construction data, and the data is compared, and a building quality and safety risk assessment is made based on the compared data.

[0004] Although the above-mentioned technical solution solves the problem of detection accuracy, in order to realize detection, numerous small modules are required to be tested separately, and finally unified and summarized to achieve accurate detection. This is not only time-consuming but also the workload of preliminary data collection is also huge. At the same time, the solution detected also needs to observe the results of multiple modules to verify, which is too complicated.

[0005] Therefore, a construction safety detection system based on color quantization is proposed to solve the above-mentioned problems. Summary of the invention

[0006] The present invention aims to provide a construction safety detection system based on color quantization to solve or improve at least one of the above-mentioned technical problems.

[0007] In view of this, a first aspect of the present invention is to provide a construction safety detection system based on color quantization.

[0008] The first aspect of the present invention provides a construction safety detection system based on color quantization, comprising: an image receiving unit, connected to a camera located at a construction site; the image receiving unit is used to obtain image data of the construction site taken by the camera at each moment; the camera is also connected to an image display unit to synchronously display the image data received by the image receiving unit; a color calibration unit, which is internally provided with a plurality of reference color information collected through markers in the construction site; the color calibration unit calibrates the image data received by the image receiving unit through at least one of the reference color information; an image division unit, which is used to divide a non-safe area corresponding to the construction site on a scene image corresponding to the image data; a target tracking unit, which is used to identify moving objects in the scene images at adjacent moments, and determine whether to generate alarm information according to the motion state of the moving objects in the non-safe area; a real-time warning unit, which is used to establish association information between the image data displayed by the image display unit and the scene image; when the alarm information is generated, the real-time warning unit marks the alarm information on the image data displayed by the image display unit through the association information.

[0009] In any of the above technical solutions, the color calibration unit includes: an identification module, used to identify the color value corresponding to the reference color information in the image data received by the image receiving unit; a calibration module, used to calculate the difference between the color value and the corresponding reference color information, and calibrate all color values ​​in the image data according to the difference value.

[0010] In any of the above technical solutions, the calibration module performs linear transformation on all the color values ​​according to the difference value, and removes noise from the color values ​​after the linear transformation.

[0011] In any of the above technical solutions, the linear transformation is calculated by the following formula: ; wherein said is a 3x3 color correction matrix; is an offset assigned by the difference value; is the RGB color of the image data before calibration; It is the RGB color of the image data after calibration.

[0012] In any of the above technical solutions, the noise removal is calculated by the following formula: ; wherein said is the value at position after mean filtering. The pixel value of is the horizontal axis, is the vertical axis; is the normalization factor of the mean filter, and n is the size of the filter window; For the original image at position The pixel value of For location The pixel values ​​in a surrounding n×n window are summed up; To indicate the offset of the row; To indicate the offset of the column; is the filter window radius, and .

[0013] In any of the above technical solutions, the image segmentation unit detects and segments the non-safe area through the image data processed by linear transformation, and the pixel points corresponding to the non-safe area are removed from the noise through the calibration module.

[0014] In any of the above technical solutions, the markers include the enclosure walls and construction equipment and identification vests of construction workers in the construction site; the unsafe area is formed by the enclosure walls; and the reference color information for calculating the difference value is collected through the enclosure walls.

[0015] In any of the above technical solutions, the target tracking unit obtains a set of moving objects in the construction site through a recognition algorithm. ; where each moving object is represented by , is the spatial position of the moving object, is the speed of the moving object, is the acceleration of the moving object.

[0016] In any of the above technical solutions, the alarm information is generated by judging the following rules: judging whether the color value of the moving object in the scene image corresponds to the reference color information, and generating the alarm information if it does not correspond; when the color value of the moving object in the scene image corresponds to the reference color information, judging whether the motion state at the current moment is greater than a preset threshold, and generating the alarm information if it is greater.

[0017] In any of the above technical solutions, the motion state is calculated by the following formula: ; wherein said is the observed motion state; is the observation matrix; is the observation noise; The motion state at time t is represented by the state vector at time t+1, and ; is the state transfer matrix; is the control input matrix; is the control input; For moving objects At the moment The state vector of Including position and speed; is the control input vector, and includes Information of external control input.

[0018] Compared with the prior art, the present invention has the following beneficial effects:

[0019] By collecting reference color information from markers on the construction site and using this information to calibrate the image data, the accuracy of color recognition is significantly improved. This accuracy is critical for correctly identifying and monitoring personnel and equipment on the construction site, ensuring that safety management measures are targeted and effective.

[0020] The object tracking unit can monitor moving objects in a continuous moment and generate real-time alerts based on their dynamic behavior in unsafe areas. This real-time monitoring and alert generation can respond to potential safety threats in a timely manner, thus preventing accidents.

[0021] The addition of the image segmentation unit enables the system to effectively segment and manage the non-safe areas of the construction site. This not only reduces the safety accidents that may be caused by mistakenly entering high-risk areas, but also enhances the accuracy and efficiency of area management.

[0022] The real-time warning unit associates the displayed image data with the real-time monitored scene image so that once a potential risk is detected, the alarm information can be immediately marked on the image display unit, significantly improving the speed and efficiency of handling emergencies.

[0023] Additional aspects and advantages of embodiments according to the present invention will become apparent in the following description or may be learned through practice of embodiments according to the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] The above and / or additional aspects and advantages of the present invention will become apparent and easily understood from the description of the embodiments in conjunction with the following drawings, in which:

[0025] Figure 1 It is a logic block diagram of the present invention. DETAILED DESCRIPTION

[0026] In order to more clearly understand the above-mentioned purpose, features and advantages of the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that the embodiments of the present application and the features in the embodiments can be combined with each other without conflict.

[0027] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Therefore, the protection scope of the present invention is not limited to the specific embodiments disclosed below.

[0028] See also Figure 1 , a construction safety detection system based on color quantization according to some embodiments of the present invention is described below.

[0029] The embodiment of the first aspect of the present invention provides a construction safety detection system based on color quantization. In some embodiments of the present invention, such as Figure 1 As shown, the building construction safety detection system includes:

[0030] The image receiving unit is connected to a camera located at the construction site; the image receiving unit is used to obtain image data of the construction site taken by the camera at each moment; the camera is also connected to an image display unit to synchronously display the image data received by the image receiving unit; real-time playback through the image display unit can restore the original appearance of the construction site to the greatest extent, and is not affected by subsequent image processing and detection steps, so as to achieve independent playback.

[0031] The color calibration unit is internally provided with a plurality of reference color information collected through markers in the construction site; the color calibration unit calibrates the image data received by the image receiving unit through at least one reference color information; the easily distinguishable color on the marker is used as the reference color information, so as to realize the pre-color calibration, and it appears at a high frequency in the construction site, so as to improve the color recognition accuracy under environmental factors such as light and haze environment.

[0032] The image division unit is used to divide the non-safe area corresponding to the construction site on the scene image corresponding to the image data; dividing the sub-safe area can demarcate the places where people or vehicles are not wanted in the construction site, ensuring the smooth progress of the construction and the safety of people and ordinary vehicles outside the area.

[0033] The target tracking unit is used to identify moving objects in scene images at adjacent moments and determine whether to generate alarm information based on the movement state of the moving objects in the unsafe area.

[0034] The real-time warning unit is used to establish association information between the image data displayed by the image display unit and the scene image; when the alarm information is generated, the real-time warning unit marks the alarm information on the image data displayed by the image display unit through the association information. The association information is the identity between the processed image data and the image data responsible for playback, so that the same image data responsible for playback and display can be found after the alarm information appears in the processed image data, and the alarm information, such as the prompt pattern, is displayed on the image display unit together with the number of the image played and displayed.

[0035] The present invention provides a construction safety detection system based on color quantization, in which an image receiving unit receives image data from a camera at a construction site. The camera is installed at a key position at the construction site, and can capture a continuous image stream at the site, and transmit the data to the image receiving unit in real time. This unit is responsible for receiving data from one or more cameras to ensure the integrity and real-time nature of the image data; the camera is the data acquisition device of the system, and is responsible for capturing dynamic images of the construction site. These cameras are usually equipped with high resolution and wide viewing angle functions to ensure that a wide area is covered and detailed image information is captured, providing sufficient data support for subsequent image processing; the image display unit connected to the image receiving unit is responsible for synchronously displaying the received image data. This unit can be a display screen of a monitoring center or other types of visual output devices, which allows security monitoring personnel to view the status of the construction site in real time. In this way, the image display unit can restore and display the actual situation of the construction site to the greatest extent, and will not affect the independence and clarity of real-time playback even when image processing and analysis are performed.

[0036] The system continuously collects images of the construction site through cameras and monitors potential safety hazards in real time, such as workers who are not wearing helmets and equipment that is not operated according to regulations. This real-time monitoring helps on-site managers respond to potential dangers quickly and intervene in time; after the image receiving unit obtains the image data, the system analyzes the image through color quantization technology. Color quantization involves simplifying the colors in the image into several predefined colors, which are used to mark different types of safety equipment or areas. For example, helmets, safety nets, etc. can be identified by specific colors, which makes the image analysis process simpler and more intuitive; the system design ensures that the image display unit can independently play the real-time video stream without being affected by the image analysis and processing modules. This means that even when performing complex image processing operations, the monitoring personnel can still see the unmodified real-time image, ensuring the continuity and real-time nature of monitoring; the system can automatically analyze the processed image data, identify violations or dangerous behaviors, and generate safety reports. In addition, the system can be configured to automatically trigger an alarm when a serious safety violation is detected, notifying the management personnel in real time.

[0037] The color calibration unit is internally provided with multiple reference color information, which is collected from common markers in the construction site. The markers can be common safety signs, tools or equipment, which appear frequently in the construction site and are easy to identify in color; the color calibration unit uses at least one reference color information to calibrate the image data received by the image receiving unit. This calibration ensures that the captured color information can be as close to the actual color as possible under different ambient lighting conditions, such as direct sunlight, shadow or haze environment.

[0038] When selecting the reference colors, special consideration is given to those colors that appear frequently and are visually easy to identify in the construction site. For example, the bright yellow or orange of the hard hat, the red and white of the safety warning signs. These colors are preset in the color calibration unit as the reference for color recognition and calibration; the lighting conditions at the construction site are variable, such as the change in the intensity of natural light and weather conditions (such as haze) that affect the color authenticity of the image captured by the camera. The color calibration unit automatically adjusts the color parameters to match the preset color by comparing the color in the real-time image captured by the camera with the preset reference color. This calibration can be performed dynamically to ensure that accurate color data can be obtained even in strong sunlight or insufficient light; the data during the calibration process is fed back to the image receiving unit and the image display unit in real time, so that the real-time video monitoring can not only display accurate color information, but also help security monitoring personnel to more accurately identify potential safety risks; the technology of the color calibration unit can be applied in the image processing software to process the image stream incoming from the camera. This processing is not limited to color calibration, but can also be extended to other aspects of the image, such as contrast adjustment and image sharpening, to further improve image quality.

[0039] The image segmentation unit is responsible for automatically identifying and segmenting the non-safe areas corresponding to the construction site in the received scene image. These areas are usually places where people and vehicles are not expected or allowed to appear in the construction site, such as deep excavation areas, heavy machinery operation areas, temporary storage areas for hazardous materials, etc. Once the non-safe areas are identified, the system will clearly mark these areas on the image, using different colors or patterns to distinguish them from ordinary areas. This visual identification allows monitoring personnel to quickly identify potential dangerous areas.

[0040] The image segmentation unit uses advanced image analysis techniques, such as edge detection, region growing, image segmentation algorithms, etc., to identify different areas in the image. These techniques can automatically distinguish unsafe areas based on features such as color, texture, and shape; in more complex embodiments, the image segmentation unit may use machine learning or deep learning models to improve the accuracy and efficiency of the segmentation. By training these models to recognize specific danger signs and environmental features in various construction scenarios, the system can automatically adapt to different working environments and conditions; after marking the unsafe areas, the system can provide real-time feedback to the monitoring system at the construction site. If the system detects that a person or vehicle enters these areas, it can automatically trigger an alarm or notify the construction site management personnel to take corresponding safety measures; the image segmentation unit can dynamically adjust the division of unsafe areas according to the construction progress and site changes. This is achieved by continuously analyzing the newly received image data and updating the definition and location of the unsafe areas.

[0041] The target tracking unit uses image processing algorithms to detect and identify moving objects, such as people, vehicles, machinery, etc. from continuous scene images. The recognition process involves analyzing the differences between adjacent image frames to determine which objects are moving; once a moving object is identified, the system will track its path in the video stream. This is usually achieved through target detection and tracking algorithms, such as using optical flow, background subtraction, or deep learning-based object recognition technology to maintain continuous tracking of the target object; the target tracking unit monitors the motion state of the tracked object and compares it with the position of pre-defined non-safe areas. If any object is detected entering or moving in these areas, the system will determine whether an alarm needs to be generated based on the set rules.

[0042] The system continuously receives image frames from the camera. Each frame is analyzed to determine and identify moving objects in the image; the identified objects are continuously tracked using methods such as Kalman filters, particle filters, or more modern deep learning methods (such as CNN-based tracking algorithms). These algorithms can help the system understand the direction and speed of movement of the object; the target tracking unit compares the location of the tracked object with the location of the known unsafe area. This comparison is performed in real time to ensure an immediate response; if the system analysis indicates that the object has entered an unsafe area, the target tracking unit will trigger an alarm. These alarms can be visual signals, sound alarms, or notifications sent immediately to on-site managers; all events, including object identification, tracking, and any alarm events, are recorded for post-analysis, system adjustments, and safety performance improvements.

[0043] One of the main functions of the real-time warning unit is to establish an association between the image data being processed and the image data played by the image display unit. This means that the system is able to track any image frame in the processing process to its corresponding frame in the monitoring display; when the system's image analysis unit (such as the target tracking or image segmentation unit) detects a potential safety issue and generates an alarm, the real-time warning unit will mark these alarm information in real time on the corresponding image data displayed by the image display unit. This includes intuitively displaying warning signs, patterns or text descriptions on the image.

[0044] The real-time warning unit uses internal algorithms to maintain synchronization between processed image data and displayed image data. This usually involves methods such as timestamp alignment and frame sequence numbers to ensure that each processed image frame accurately corresponds to the same image frame in the display unit; once a situation that requires a warning is detected, the warning unit will add an alarm mark to the corresponding image frame. This may be done through visual elements such as overlay layers, highlighted areas, or dynamically inserted icons. The labeling process needs to be quick and accurate so as not to delay the transmission of the alarm; the marked image will be immediately transmitted to the image display unit for on-site monitoring personnel or remote management personnel to view. This real-time feedback mechanism ensures that monitoring personnel can receive alarms and respond to potential risks in the shortest time; the warning unit is also responsible for maintaining data consistency and integrity, ensuring that the addition of alarm information does not interfere with the quality and information content of the original image data.

[0045] In summary, the real-time monitoring capability and independence are enhanced. Through the independent playback function, the monitoring image is ensured not to be disturbed by subsequent processing, so as to achieve accurate feedback of the on-site situation; it provides real-time, non-deformed scene image display, so that on-site managers can immediately understand the site conditions, improve the response speed and processing accuracy. The color recognition accuracy under changing environmental conditions is improved, especially under different lighting and weather conditions; through pre-color calibration, the system can more accurately identify the safety signs and warning signs on the construction site, thereby improving the effectiveness of safety monitoring. Automatically identify and divide non-safe areas, reduce dependence on manual monitoring, reduce labor costs and errors; automated area division reduces the workload of personnel and improves the efficiency and safety of risk area management. Automatically identify and track moving objects, especially activities in non-safe areas, thereby reducing dependence on specialized monitoring personnel; the real-time tracking and alarm system can quickly respond to potential risks, promptly notify managers to take measures, and reduce the possibility of accidents. By directly marking the alarm information on the real-time playback image, the problem of data consistency and instant feedback is solved; ensuring that the alarm information can be displayed synchronously with the corresponding image, the visibility and immediacy of the alarm information are improved, making the response faster and more accurate.

[0046] Specifically, the step of the image receiving unit receiving the image data includes:

[0047] S1.1. Sensor selection and layout: First, select a color imaging sensor: a low-light CMOS sensor or a full-spectrum CMOS sensor, covering the areas and targets that need to be monitored according to the specific layout and safety requirements of the construction site.

[0048] S1.2, Image acquisition and processing: The color imaging module collects image data of the construction site in real time through sensors. The data includes the real-time situation of the construction area: personnel, equipment and potential safety hazards. The collected image data is pre-processed: denoising and color correction.

[0049] S1.3. Data transmission and reception: The collected image data is transmitted to the back-end processing system through the data transmission component.

[0050] S1.4. Real-time monitoring and feedback: After being processed and analyzed, the image data is fed back to the image display module in real time for management personnel to observe and analyze.

[0051] From the above, it can be seen that by selecting appropriate sensors such as low-light CMOS sensors or full-spectrum CMOS sensors, color image data of the construction site can be collected in real time. This real-time performance ensures that the monitoring system can obtain the real-time situation of the construction area in a timely manner, including personnel, equipment and potential safety hazard areas, thereby providing comprehensive monitoring coverage. The collected image data undergoes preprocessing steps such as denoising and color correction to make the image quality higher and the information clearer. This preprocessing not only improves the visual quality of the image, but also reduces noise interference in subsequent processing and analysis, ensuring that managers can accurately understand and evaluate the conditions of the construction site. The image receiving unit is responsible for transmitting the collected image data to the back-end processing system through the data transmission component. This efficient data transmission ensures the advantages of the monitoring system in real-time and feedback speed, and managers can quickly obtain processed image data for real-time monitoring and analysis.

[0052] Specifically, the step of displaying the image data by the image display unit includes:

[0053] S2.1. Image data reception: The image display module first receives the image data transmitted from the color imaging module.

[0054] S2.2, Image decoding and processing: The received image data is decoded and converted into visual image content.

[0055] S2.3, Image display device: The image display module uses a liquid crystal display (LCD), LED display or computer monitor.

[0056] S2.4, Image display control: control the image display mode and operation interface, including image zoom in, zoom out, full screen display, and multi-screen switching functions.

[0057] S2.5, Image quality management: According to the information provided by the color calibration unit, adjust the parameters of the display device to make the display effect as consistent as possible with the actual scene.

[0058] As can be seen from the above, the image display unit receives the transmitted image data from the camera, so that the system can obtain and display the real-time image of the construction site in a timely manner, providing the latest information for monitoring and analysis.

[0059] Furthermore, the image decoding expression in step S2.2 is:

[0060] ;

[0061] in, is the decoded image data; It is a decoding function whose function is to convert input data into a specific output. The specific implementation of the decoding function depends on the type of data being processed and the target application; for example, in communication systems, decoding may involve extracting information from received signals; in image compression, decoding may involve recovering compressed image data; Indicates at location The original input data refers to the original value of a pixel in the image or the data before encoding.

[0062] In any of the above embodiments, the color calibration unit includes:

[0063] The recognition module is used to identify the color value corresponding to the reference color information in the image data received by the image receiving unit; the color value acquired in real time and the preset calibrated reference color information can be determined by a range value, and within the range value, it means that the color value after shooting is the same as the reference color information.

[0064] The calibration module is used to calculate the difference between the color value and the corresponding reference color information, and calibrate all color values ​​in the image data according to the difference. Since there is a deviation from the actual reference color information after shooting, transcoding and transmission, all color values ​​are calibrated to make detection and recognition more accurate without worrying about the actual playback effect.

[0065] In this embodiment, the recognition module automatically recognizes the color value corresponding to the preset reference color information in the image data obtained by the image receiving unit. These reference colors are usually extracted from common markers (such as safety helmets, warning signs, etc.) at the construction site; by setting a color tolerance range, the recognition module compares the color value captured in real time with the reference color. If the real-time color value is within the predetermined range, the system considers that the color matches the reference color, that is, the color recognition is successful.

[0066] At the beginning of system deployment, representative colors are selected as reference colors and recorded in the system. These color values ​​are selected according to the specific needs of the construction site and the colors of the iconic objects; for each reference color, the system sets a color range value (or color tolerance). This range defines the flexibility of color matching to adapt to different lighting conditions and possible color variations; when the image receiving unit obtains new image data, the recognition module analyzes the colors in the image through the color recognition algorithm; color recognition technology (such as color histogram comparison, RGB color space analysis, etc.) is used to detect each color in the image and compare it with the preset reference color; if the detected color value falls within the range value of the reference color, the color match is considered successful. Conversely, if the color value exceeds the range value, the color match is considered to have failed, and the system settings may need to be adjusted or the color may need to be recalibrated; the results of successful matching can be directly used for subsequent safety monitoring analysis, such as calibrating the color identification of the safety area; if color matching fails frequently, the system can prompt to adjust the ambient light or readjust the color calibration parameters.

[0067] The calibration module first needs to calculate the difference between the actual captured color value and the preset reference color information. This difference value reflects the degree of deviation between the two and is the first step in the calibration process; based on the calculated difference value, the calibration module adjusts all color values ​​in the image data. This adjustment is global, meaning that the color of the entire image will be corrected accordingly based on this difference value; by adjusting the color value, the module ensures that the color in the image is closer to reality, thereby improving accuracy in the subsequent detection and recognition process and avoiding misjudgments caused by color deviation.

[0068] The calibration module analyzes the difference between the reference color and the actual captured color, which usually involves the conversion and comparison of color space, such as the difference calculation of RGB or HSV color space; uses color adjustment algorithms, such as linear mapping or more complex color mapping techniques, to adjust each color in the image according to the difference value. This may involve adjusting the color's brightness, contrast, and saturation properties; the calibration adjustment is applied to each pixel of the image data to ensure that the color representation of the entire image is corrected. This global calibration helps eliminate color deviations introduced during shooting, transcoding, and transmission; the calibration module can provide real-time feedback on the calibration effect, allowing the system operator or automatic calibration algorithm to adjust the calibration parameters according to the actual output to optimize the calibration effect.

[0069] Specifically, the colorimetric expression in the color calibration unit is:

[0070] ;

[0071] in is the set of colors in the color space, Indicates color category The corresponding normalized color value.

[0072] In any of the above embodiments, the calibration module performs linear transformation on all color values ​​using the difference values, and removes noise from the color values ​​after the linear transformation.

[0073] In this embodiment, the color values ​​in the image are adjusted to make them closer to the colors of the actual objects, especially the color deviation caused by equipment characteristics and environmental factors during shooting and transmission; by calculating the difference between the actual color and the reference color, the linear transformation formula is applied to adjust the color value of each pixel. This transformation usually involves adjusting the RGB value, which can be expressed as C′=a×C+b, where C is the original color value, C′ is the adjusted color value, and a and b are coefficients determined experimentally to achieve the best color match. Linear transformation may amplify the noise in the image, especially when the color value is close to the extreme value (such as very bright or very dark). Noise removal is to improve the visual quality of the image and the accuracy of subsequent processing; various image denoising techniques are used, such as average filtering, median filtering, Gaussian filtering, etc. These methods can effectively reduce random noise in the image without significantly affecting the details of the image.

[0074] The system first analyzes the real-time image captured by the camera, compares it with the preset reference color, and calculates the necessary color adjustment parameters (a and b). These parameters are applied to each pixel of the image data, and the original color value is adjusted by a predetermined linear formula to reduce the color deviation introduced by equipment and environmental factors. After color adjustment, the image may show enhanced noise or uneven color. At this time, the calibration module will apply a denoising algorithm to process the image; select a suitable denoising algorithm based on the characteristics of the image, for example, use an average filter in areas with uniform color, and use a median filter or Gaussian filter in areas with rich details. These filters can effectively reduce noise without excessively blurring image details. Based on the quality and color accuracy of the processed image, the system can feedback and adjust the calibration parameters to optimize the processing effect of subsequent images; regular or continuous calibration and optimization can ensure that the system maintains efficient and accurate response to changing environmental conditions and equipment aging.

[0075] In any of the above embodiments, the linear transformation is calculated by the following formula:

[0076] ;

[0077] in, is a 3x3 color correction matrix; The offset assigned by the difference value; It is the RGB color of the image data before calibration; The RGB color of the image data after calibration.

[0078] In this embodiment, the system adjusts the color value of each pixel through the matrix M and the vector T. The matrix M is mainly used for color correction, which can enlarge, reduce or rotate the color space to compensate for the color deviation of the camera or the influence of ambient light; the vector T is usually used to adjust the brightness and contrast of the entire image, or to fine-tune a specific color more accurately.

[0079] Each pixel of the image is composed of three basic colors: red (R), green (G), and blue (B). The system first reads these raw color values. For each pixel in the image, its color value is entered into the formula as a vector; using matrix multiplication, the raw color vector is multiplied by the matrix M. This step adjusts the color value, which may include color enhancement, suppression, or correction of color cast. Vector T is added to the above result to further adjust the brightness and saturation of the color to achieve the desired visual effect. The calculation results represent calibrated color values, which will be used to update the display of the image or further image processing.

[0080] In any of the above embodiments, the noise removal is calculated by the following formula:

[0081] ;

[0082] in, is the value at position after mean filtering. The pixel value of is the horizontal axis, is the vertical coordinate; is the normalization factor of the mean filter, and n is the size of the filter window; For the original image at position The pixel value of For location The pixel values ​​in a surrounding n×n window are summed up; To indicate the offset of the row; To indicate the offset of the column; is the filter window radius, and .

[0083] In this embodiment, the main function of the mean filter is to replace the value of each pixel in the image by calculating the average value of the pixel values ​​in the neighborhood around each pixel. Doing so can reduce random noise in the image, because noise is usually unevenly distributed in the image, and these inconsistencies can be eliminated to a certain extent by taking the average.

[0084] For each pixel in the image, the system defines a neighborhood, which is usually a square area centered on the current pixel, the size of which is determined by the size of the filter (e.g. 3x3, 5x5, etc.); the system calculates the average value of all pixels in the neighborhood defined above. This involves adding the color values ​​(red, green, blue) of all pixels in the neighborhood and then dividing it by the total number of pixels (i.e. the number of pixels in the neighborhood); the resulting average value will replace the pixel value in the center of the neighborhood. This means that the original center pixel value will be updated to the average color value of the pixel values ​​around it; this process is repeated for each pixel in the image, except for edge pixels, because the neighborhood of edge pixels may exceed the image boundary. For edge pixels, special measures can be taken, such as ignoring them or only calculating the average value within the valid neighborhood.

[0085] In any of the above embodiments, the image segmentation unit detects and segments the non-safe area through the image data processed by linear transformation. By focusing on the detection and segmentation of the non-safe area, the processing and recognition calculation amount of the image data can be reduced, and it can adapt in real time when the actual construction site changes, such as expansion and reduction; and the pixel points corresponding to the non-safe area are denoised by the calibration module. During the noise removal operation, only the pixel points of the non-safe area on the scene image are focused on, which can further reduce the pressure of computational processing.

[0086] In this embodiment, the image segmentation unit uses linear transformation to process image data and enhance image features, so that the non-safe area is more prominent in the image, which is convenient for subsequent detection and segmentation. The unit focuses on identifying those areas that are specified as non-safe, such as heavy machinery operation areas, high-risk work areas, etc. These areas have been clearly marked in construction drawings or safety specifications; when actual changes occur at the construction site, such as expansion or reduction, the image segmentation unit can automatically adjust the division of non-safe areas through real-time image analysis to ensure the continued effectiveness of safety monitoring; in order to improve the accuracy of non-safe area detection, the unit will also remove noise from the image data of these areas through a calibration module. Noise removal is mainly concentrated in non-safe areas to reduce the overall computational load.

[0087] The image data first goes through a linear transformation process, which usually includes adjustments to color, brightness, and contrast to enhance key features in the image and make non-safe areas easier to identify; image segmentation algorithms such as thresholding, edge detection, or more complex machine learning methods are applied to identify and divide non-safe areas. These algorithms can automatically identify areas of specific color, texture, or shape based on preset rules and standards; the image segmentation unit dynamically adjusts the boundaries of non-safe areas by continuously analyzing real-time images from the camera. This real-time performance is critical to cope with the dynamic changes common in construction sites; after identifying non-safe areas, local noise removal techniques (such as local smoothing filtering, median filtering) are used to process the image data specifically for these areas. This can improve the recognizability of non-safe areas in the image while reducing the computational burden of processing other parts of the image.

[0088] In any of the above embodiments, the markers include enclosure walls and construction equipment in the construction site, and identification vests of construction workers.

[0089] The unsafe area is formed by enclosing a fence wall. Since the construction site is usually accompanied by fence railings, walls and markings in actual operations, enclosing a unsafe area with a fence wall can minimize the size of the unsafe area to be concerned and reduce the probability of misjudgment.

[0090] The reference color information for calculating the difference value is collected through the enclosure wall. Collecting through the enclosure wall can ensure that the corresponding actual object must appear in the picture taken by the camera, ensuring real-time calibration operation.

[0091] In this embodiment, the enclosure walls are usually used to isolate the construction area to prevent unauthorized entry and reduce the impact of construction activities on the surrounding environment. They are often painted in eye-catching colors or carry safety warning signs, which are easy to be identified by the security monitoring system; construction equipment such as cranes, excavators and other large machinery often have specific color codes, which makes them easy to distinguish and track in the monitoring system. Identifying these equipment helps to monitor their operating status and confirm whether they are in a safe operating area; construction workers usually wear high-visibility identification vests, which are usually bright colors (such as fluorescent yellow or orange) and have reflective strips. This makes it easier to identify people under surveillance cameras, especially in dim light or poor visibility.

[0092] The security monitoring system uses advanced visual processing techniques, such as color detection and shape recognition algorithms, to identify markers. The system is pre-configured with the color and shape information required to identify these markers; the video stream captured by the camera is analyzed in real time, and the position and movement of the markers are continuously tracked. For enclosure walls and construction equipment, the monitoring system checks whether they remain within the predetermined safety area; for construction personnel, the system ensures that they do not enter prohibited or dangerous areas; when the monitoring system detects that a marker (such as a person or equipment) enters an unsafe area, an alarm or notification can be automatically triggered. In addition, the system can dynamically adjust the monitoring strategy according to actual conditions, such as expanding the monitoring range or increasing the monitoring density in a specific area; all monitoring data, including the identification and location information of the markers, are recorded and stored. This data can be used for post-analysis to help improve future safety measures and construction planning.

[0093] Use enclosure walls and other physical isolation measures (railings, safety markings, etc.) to physically divide and define non-safe areas on the construction site. These areas are areas where unauthorized personnel or non-construction personnel are not allowed to enter, usually including places where heavy machinery is operated, high-risk operations or potentially hazardous substances are handled; clearly defining the boundaries of non-safe areas through physical isolation means helps to narrow the scope of monitoring and management, allowing the safety monitoring system to focus more on these high-risk areas, reduce resource waste and improve monitoring efficiency and accuracy; clear and accurate non-safe area definitions help reduce the probability of misjudgment of the safety monitoring system when identifying and responding to potential risks. Clear physical markings provide a consistent and unchanging reference, reducing the need to rely on subjective judgment.

[0094] At the beginning of construction, according to the construction safety plan, fences, safety railings or obvious markings are set up around areas that need special protection. These isolation measures not only serve as physical barriers to restrict access, but also as important visual signs for the security monitoring system; the security monitoring system captures real-time images through cameras installed on the construction site. The system's built-in image processing software can identify fences and other safety markers, and automatically distinguish between non-safe areas and ordinary areas; once the monitoring system identifies any unauthorized personnel or equipment entering the area defined as non-safe, it immediately triggers the alarm system to notify the site management to take action. In addition, the system can update the status of the area in real time to adapt to the dynamic changes of the construction site; the collected data is used to analyze the safety status of non-safe areas, evaluate the effectiveness of existing safety measures, and adjust safety strategies and physical isolation measures according to actual conditions.

[0095] Fence walls are ideal sources of reference color information because they are commonly found on construction sites and have relatively uniform and stable colors. They are usually painted in standard colors (such as orange, blue, etc.) that are easy to identify and visually prominent. Since fence walls are often within the camera's field of view on construction sites, using them as a reference for color calibration can ensure the real-time calibration process. The camera can continuously capture images of these fence walls, allowing for real-time color comparison and calibration. The color of fence walls is usually designed to resist the influence of environmental factors such as sunlight and rain, which makes their color change less under different lighting and weather conditions, providing a reliable reference color.

[0096] The camera captures image data of the enclosure wall at regular intervals. This data is used to analyze the color characteristics of the wall, such as RGB values ​​or parameters of other color models; the system stores preset reference values ​​for the color of the enclosure wall. When real-time image data is obtained from the camera, the system compares the actual captured color with the preset reference color and calculates the color difference value. This difference value reflects the deviation between the actual color and the reference color; based on the calculated difference value, the system automatically adjusts the representation of all colors in the image to ensure that the color display is as close to the actual color as possible. This includes adjusting the color gain, offset and other related parameters of the image data; the system continuously monitors the color status of the enclosure wall and environmental changes, and updates the reference value of color calibration when necessary to adapt to changes in construction site conditions.

[0097] In any of the above embodiments, the target tracking unit obtains a set of moving objects in the construction site through a recognition algorithm. ;

[0098] Among them, each moving object is represented as , is the spatial position of the moving object, is the speed of the moving object, is the acceleration of the moving object.

[0099] In this embodiment, the system is able to identify the current state of each moving object in the construction site, including its position (represented by coordinates x, y), speed (v), and other related dynamic attributes (a). These attributes may include the object's direction, acceleration, etc., depending on the monitoring needs and the technology implemented; the system not only identifies the initial state of each object, but also updates this state information in real time to ensure that the monitoring data reflects the latest site conditions. This is particularly important for preventing accidents and handling emergencies; all collected data is associated with the object's identification number (O), and each object has a unique identifier in the system. This makes it possible to track a specific object or view its historical movement trajectory.

[0100] Use cameras or other sensor devices to capture images and other sensor data of moving objects in the site. These devices are installed in strategic locations to cover the entire construction area; the images and sensor data are transmitted to the central processing system, where specialized software application algorithms parse the data and identify the position, speed and other dynamic parameters of each object; the system maintains a dynamically updated record for each identified object, including real-time data on its position, speed and other parameters. This information is displayed to monitoring personnel through a graphical user interface or used in automated safety control systems; based on real-time data, the safety management system can quickly respond to potential safety threats, such as automatically sending warnings, triggering shutdown measures or notifying on-site managers to take action.

[0101] In any of the above embodiments, the alarm information is generated by judging the following rules:

[0102] Determine whether the color value of the moving object in the scene image corresponds to the reference color information. If not, generate an alarm message. When the color value of the moving object in the scene image does not correspond to the reference color information, the moving object is determined to be an outsider or vehicle other than the construction work, and an alarm should be issued immediately.

[0103] When the color value of the moving object in the scene image corresponds to the reference color information, it is determined whether the current motion state is greater than the preset threshold. If it is, an alarm message is generated. When the motion state is greater than the preset threshold, it is determined that the construction personnel or equipment are moving rapidly, and a production accident may occur. At this time, an alarm should be issued immediately.

[0104] In this embodiment, the system analyzes the colors of moving objects in the scene image captured by the camera and checks whether these colors match the pre-defined reference color information. The reference colors usually correspond to the standard colors of the safety clothing of construction workers or construction vehicles; if the color of the moving object identified in the image does not match any of the reference colors, the system interprets this situation as an intrusion by an unauthorized individual - perhaps an outsider or a non-construction vehicle; in this case, the system automatically triggers an alarm to notify security personnel for further processing.

[0105] The camera monitors the construction site in real time and captures images containing moving objects. These images are transmitted to the central processing system, which contains a color recognition algorithm. The system uses color detection technology (such as analysis based on the HSV color space) to identify the color of moving objects in the image. The system has preset reference colors allowed for the construction site. These colors represent the colors of the construction workers' uniforms or construction vehicles. For the color of each moving object in the image, the system will check whether it matches any of the reference colors. If the color of a moving object does not match any of the preset reference colors, the system determines that it may be an unauthorized intruder. Once the system makes such a determination, it automatically activates the alarm system, which may include sound alarms, sending security warnings to the monitoring center, or directly notifying site security personnel. The system continuously monitors every mobile event in the scene and updates the color recognition results and alarm status in real time. Security personnel can view the alarm status and related images in real time through the monitoring interface to quickly respond to possible security threats.

[0106] The system monitors the speed and acceleration of each moving object (such as construction workers and equipment) in the scene image, and these data are used to evaluate its current dynamic state; for each identified object, the system first confirms whether its color matches the preset reference color of the construction site (i.e. whether it is a construction worker or construction equipment), and then the system evaluates whether the object's motion state exceeds the preset safety threshold. If the motion state (such as speed or acceleration) exceeds the safe range, the system determines that this may be a high-risk event; when the system detects a high-risk motion state, it automatically generates an alarm message and warns the monitoring personnel or the on-site safety management team through various channels (sound alarm, visual warning, SMS or email notification, etc.).

[0107] Cameras installed at construction sites capture video streams in real time and use image processing technology to identify the color and position of each moving object in the scene. The system uses changes between image frames to calculate the object's speed and acceleration. The color recognition module confirms whether the object belongs to construction personnel or equipment (based on color matching), and the motion analysis module evaluates whether the object's speed or acceleration exceeds a preset safety threshold. If an object's color matches the baseline color and its motion state exceeds the safety threshold, the system considers it a potential safety accident risk. This dual verification mechanism (color and motion state) helps improve the accuracy and timeliness of alarms. Once an alarm is triggered, the system issues a sound or light warning through the connected alarm system and notifies relevant safety managers through the communication system for emergency response.

[0108] In any of the above embodiments, the motion state is calculated by the following formula:

[0109] ;

[0110] in, is the observed motion state; is the observation matrix; is the observation noise; The motion state at time t is represented by the state vector at time t+1, and ; is the state transfer matrix; is the control input matrix; is the control input; For moving objects At the moment The state vector of Including position and speed; is the control input vector, and includes Information of external control input.

[0111] In this embodiment, the system can predict the state of each monitored object at the next point in time. This is done by analyzing their current state data (such as position, speed, etc.) and combining time series analysis techniques; using a set of parameterized models that combine historical data and real-time inputs to estimate the future state of the monitored objects. These parameters include but are not limited to movement speed, direction, and possible acceleration; the system updates the state of each object in real time and dynamically adjusts the monitoring strategy based on the prediction results. For example, if it is predicted that an object may enter an unsafe area, the system will issue an alarm in advance.

[0112] Cameras and other sensors collect dynamic data about objects in the scene, such as position and speed, in real time. This data is fed into a central processing system for analysis; algorithms built into the system use the currently collected data as input to calculate the future state of the object through parameterized mathematical models. These models are based on historical data and predefined physical laws, such as kinematic equations; for each object's predicted state, the system compares it with safety thresholds. If the predicted state indicates a potential safety risk (such as entering a restricted area or going too fast), the system triggers an alarm; depending on the actual situation and the accuracy of the prediction, the system may adjust the parameter model to optimize the prediction accuracy. This feedback mechanism ensures that the system improves over time and adapts to changes in the environment and operating conditions.

[0113] Specifically, the alarm steps of the real-time early warning unit include:

[0114] S3.1. Setting safety standards: Setting a set of rules based on construction safety standards. The rules are the definition of safe and unsafe behaviors: entering dangerous areas, working at heights, and corresponding color calibration information;

[0115] S3.2, color calibration information: create a color calibration reference table, and store the color calibration information corresponding to safe behaviors and unsafe behaviors and other areas or objects of concern for subsequent image processing and recognition;

[0116] S3.3, Real-time image acquisition: The real-time color imaging system is used to collect image data of the construction site in real time and transmit the data to the real-time warning module for processing;

[0117] S3.4, Target Detection: The image is processed through the target detection algorithm to identify the personnel, equipment and possible safety hazard areas in the image, and extract feature information for the next step of analysis;

[0118] S3.5, Safety standard verification: Compare the target detection results with the preset safety standards to determine whether there are any behaviors or situations that are inconsistent with the safety standards;

[0119] S3.6, color calibration comparison: compare the color information of the identified target area with the pre-set color calibration information to confirm whether there are unsafe behaviors or areas, and determine whether to trigger the warning condition;

[0120] S3.7. Real-time warning: If an abnormal situation is found that meets the warning conditions, a warning signal, alarm or reminder will be issued immediately.

[0121] Furthermore, the construction safety detection system also includes a data storage and transmission module, which stores image data is represented as a function, where and are spatial coordinates, is the time, and other system data are expressed as , including system status and log information.

[0122] The data storage expression is:

[0123] ;

[0124] The data transfer expression is:

[0125] .

[0126] In the description of the present invention, it should be understood that the terms "longitudinal", "lateral", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside" and "outside" etc., indicating orientations or positional relationships, are based on the orientations or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be understood as a limitation on the present invention.

[0127] The embodiments described above are only descriptions of the preferred modes of the present invention, and are not intended to limit the scope of the present invention. Without departing from the design spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by ordinary technicians in this field should all fall within the protection scope determined by the claims of the present invention.

Claims

1. The construction safety detection system based on color quantization is characterized by: include: An image receiving unit is connected to a camera located at the construction site; the image receiving unit is used to obtain image data of the construction site taken by the camera at each moment; the camera is also connected to an image display unit to synchronously display the image data received by the image receiving unit; A color calibration unit, which is internally provided with a plurality of reference color information collected through markers in the construction site; The color calibration unit calibrates the image data received by the image receiving unit by using at least one of the reference color information; the color calibration unit includes a recognition module and a calibration module; the recognition module is used to recognize the color value corresponding to the reference color information in the image data received by the image receiving unit; the calibration module is used to calculate the difference between the color value and the corresponding reference color information, and calibrate all the color values ​​in the image data according to the difference; the calibration module linearly transforms all the color values ​​by using the difference value, and removes noise from the color values ​​after the linear transformation; The linear transformation is calculated by the following formula: [R', G', B'] = M · [R, G, B] + T; Wherein, the M is a 3x3 color correction matrix; the T is an offset assigned by the difference value; the [R, G, B] is the RGB color of the image data before calibration; the [R', G', B'] is the RGB color of the image data after calibration; The noise removal is calculated by the following formula: Among them, the I smooth (x, y) is the pixel value at the position (x, y) after the mean filter processing, wherein x is the horizontal coordinate and y is the vertical coordinate; is the normalization factor of the mean filter, and n is the size of the filter window; I(x+i,y+j) is the pixel value of the original image at position (x+i,y+j); is an operation for summing up the pixel values ​​in an n×n window around the position (x, y); i is an offset representing a row; j is an offset representing a column; k is a filter window radius, and An image division unit, configured to divide a non-safe area corresponding to the construction site on the scene image corresponding to the image data; A target tracking unit, used to identify a moving object in the scene images at adjacent moments, and determine whether to generate an alarm message according to the motion state of the moving object in the non-safe area; A real-time warning unit is used to establish association information between the image data displayed by the image display unit and the scene image; when the alarm information is generated, the real-time warning unit marks the alarm information on the image data displayed by the image display unit through the association information.

2. The construction safety detection system according to claim 1, characterized in that: The image segmentation unit detects and segments the non-safe area through the image data processed by linear transformation, and the pixel points corresponding to the non-safe area are de-noised by the calibration module.

3. The construction safety detection system according to claim 2, characterized in that: The markers include enclosure walls and construction equipment in the construction site, and identification vests of construction workers; The non-safe area is formed by enclosing the enclosure wall; The reference color information for calculating the difference value is collected through the enclosure wall.

4. The construction safety detection system according to claim 1, characterized in that: The target tracking unit obtains a set of moving objects in the construction site through a recognition algorithm. Among them, each moving object is represented by o i =(x i ,y i ,v i ,a i ), the (x i ,y i ) is the spatial position of the moving object, and v i is the speed of the moving object, and a i is the acceleration of the moving object.

5. The construction safety detection system according to claim 4, characterized in that: The alarm information is generated by judging the following rules: Determining whether the color value of the moving object in the scene image corresponds to the reference color information, and generating the alarm information if it does not correspond; When the color value of the moving object in the scene image corresponds to the reference color information, it is determined whether the current motion state is greater than a preset threshold, and if so, the alarm information is generated.

6. The construction safety detection system according to claim 5, characterized in that: The motion state is calculated by the following formula: Wherein, z is the observed motion state; H is the observation matrix; δ is the observation noise; The motion state at time t is represented by the state vector at time t+1, and A is the state transfer matrix; B is the control input matrix; ε is the control input; For moving object o i The state vector at time t is Including position and speed; is the control input vector, and includes External control input information.

Images