Fluorescent clothes wearing identification method, device and equipment based on target detection, medium and product

Abstract

The invention discloses a fluorescent clothes wearing identification method, device and equipment based on target detection, a medium and a product. The method comprises the following steps: acquiring a first field video frame, comprising at least one operation object, of an operation field; performing fluorescent clothes wearing identification on the first field video frame according to a fluorescent clothes wearing identification model to obtain a fluorescent clothes wearing result; based on the first field video frame, determining a first object display attribute of the at least one operation object in the first field video frame; and determining a result processing mode corresponding to the fluorescent clothes wearing result according to the first object display attribute corresponding to the operation object, and processing the fluorescent clothes wearing result based on the result processing mode. According to the technical scheme, on the basis of carrying out fluorescent clothes wearing identification on the operation object in the field video frame based on the deep learning model, the effectiveness of the fluorescent clothes wearing result is judged in combination with the object display attribute, so that the effect of ensuring the accuracy of the fluorescent clothes wearing result is achieved.

Description

Technical Field

[0001] The present invention relates to the field of machine vision technology, and in particular to a method, device, equipment, medium and product for identifying fluorescent clothing wearing based on target detection. Background Art

[0002] With the development of society and the expansion of industrial production, the demand for safe production is becoming increasingly stringent, and the safety of workers working at heights in settings such as factory sites is receiving increasing attention. In these scenarios, fluorescent clothing can attract visual attention and indicate pedestrian presence at night. However, due to workers' lack of safety awareness, they are prone to improper or no wearing of clothing, which poses a high risk. Therefore, the identification of fluorescent clothing during production operations is crucial.

[0003] Currently, supervisors at production sites typically use back-end video equipment to identify workers wearing fluorescent clothing. However, manual identification of workers in production scenarios consumes significant labor costs and has low recognition efficiency and accuracy, which in turn impacts production efficiency. Summary of the Invention

[0004] The present invention provides a method, device, equipment, medium and product for fluorescent clothing wearing recognition based on target detection, so as to realize the fluorescent clothing wearing recognition of working objects in on-site video frames based on a deep learning model, and also judge the validity of the fluorescent clothing wearing results output by the model in combination with the object display attributes to ensure the accuracy of the fluorescent clothing wearing results.

[0005] According to one aspect of the present invention, a method for identifying fluorescent clothing wearing based on target detection is provided, the method comprising:

[0006] Acquire a first on-site video frame of a work site, wherein the first on-site video frame includes at least one work object located at the work site and in motion;

[0007] Performing fluorescent clothing recognition on the first scene video frame according to a pre-trained fluorescent clothing recognition model to obtain a fluorescent clothing wearing result corresponding to at least one of the work subjects; wherein the fluorescent clothing recognition model is used to identify whether the work subject in the first scene video frame is wearing and correctly wearing the fluorescent clothing;

[0008] Determining, based on the first live video frame, a first object display attribute of at least one of the operation objects in the first live video frame; wherein the first object display attribute is used to characterize display clarity and / or display completeness of the operation object in the first live video frame;

[0009] For at least one of the operation objects, a result processing method corresponding to the fluorescent clothing wearing result is determined according to the first object display attribute corresponding to the operation object, and the fluorescent clothing wearing result is processed based on the result processing method.

[0010] According to another aspect of the present invention, a fluorescent clothing wearing recognition device based on target detection is provided, the device comprising:

[0011] A video frame acquisition module is configured to acquire a first on-site video frame of a work site, wherein the first on-site video frame includes at least one work object located at the work site and in motion;

[0012] a fluorescent clothing wearing recognition module, configured to perform fluorescent clothing wearing recognition on the first scene video frame based on a pre-trained fluorescent clothing wearing recognition model, and obtain a fluorescent clothing wearing result corresponding to at least one of the work subjects; wherein the fluorescent clothing wearing recognition model is configured to identify whether the work subject in the first scene video frame is wearing and correctly wearing the fluorescent clothing;

[0013] an object display attribute determination module, configured to determine, based on the first live video frame, a first object display attribute of at least one of the operation objects in the first live video frame; wherein the first object display attribute is used to characterize display clarity and / or display completeness of the operation object in the first live video frame;

[0014] The wearing result processing module is used to determine, for at least one of the operation objects, a result processing method corresponding to the fluorescent clothing wearing result according to the first object display attribute corresponding to the operation object, and process the fluorescent clothing wearing result based on the result processing method.

[0015] According to another aspect of the present invention, an electronic device is provided, comprising:

[0016] at least one processor; and

[0017] a memory communicatively connected to the at least one processor; wherein,

[0018] The memory stores a computer program that can be executed by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the fluorescent clothing wearing recognition method based on target detection as described in any embodiment of the present invention.

[0019] According to another aspect of the present invention, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to implement the fluorescent clothing wearing recognition method based on target detection as described in any embodiment of the present invention when executed.

[0020] According to another aspect of the present invention, a computer program product is provided, which includes a computer program. When the computer program is executed by a processor, it implements the fluorescent clothing wearing recognition method based on target detection described in any embodiment of the present invention.

[0021] The technical solution of the embodiment of the present invention is to obtain a first-site video frame of the work site, wherein the first-site video frame includes at least one work object located at the work site and in motion; before performing fluorescent clothing wearing recognition, motion detection is first performed on the captured on-site video data to obtain a first-site video frame containing the work object in motion, thereby reducing the call of the algorithm model. At this time, the algorithm consumption resources are saved and the performance cost is reduced; further, fluorescent clothing recognition is performed on the first-site video frame according to the pre-trained fluorescent clothing wearing recognition model to obtain a fluorescent clothing wearing result corresponding to at least one work object, thereby achieving the effect of fluorescent clothing wearing recognition of the work object in the first-site video frame based on the deep learning model, and improving the recognition accuracy and recognition efficiency of fluorescent clothing wearing recognition; further, based on the first-site video frame, it is determined that at least one work object is in the first The first object display attribute in the on-site video frame provides a basis for determining the validity of the subsequent fluorescent clothing wearing result; further, for at least one work object, the result processing method corresponding to the fluorescent clothing wearing result is determined based on the first object display attribute corresponding to the work object, and the fluorescent clothing wearing result is processed based on the result processing method, which solves the problem that manual recognition in related technologies consumes a lot of manpower costs, and the recognition efficiency and recognition accuracy are low, thereby affecting production operation efficiency. It realizes the fluorescent clothing wearing recognition of the work object in the on-site video frame based on the deep learning model, and also combines the object display attribute to determine the validity of the fluorescent clothing wearing result output by the model to ensure the accuracy of the fluorescent clothing wearing result, further improving the recognition accuracy and recognition efficiency of fluorescent clothing wearing recognition, thereby effectively ensuring the operation safety at the work site.

[0022] It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present invention, nor is it intended to limit the scope of the present invention. Other features of the present invention will become readily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0024] Figure 1 1 is a flow chart of a method for identifying fluorescent clothing wearing based on target detection according to an embodiment of the present invention;

[0025] Figure 2 1 is a flow chart of a method for identifying fluorescent clothing wearing based on target detection according to an embodiment of the present invention;

[0026] Figure 3 1 is a flow chart of a method for identifying fluorescent clothing wearing based on target detection according to an embodiment of the present invention;

[0027] Figure 4 1 is a schematic structural diagram of a fluorescent clothing wearing recognition device based on target detection according to an embodiment of the present invention;

[0028] Figure 5 The present invention is a schematic diagram of the structure of an electronic device for implementing the fluorescent clothing wearing recognition method based on target detection according to an embodiment of the present invention. DETAILED DESCRIPTION

[0029] In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of the present invention.

[0030] It should be noted that the terms "first", "second", etc. in the description and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the numbers used in this way can be interchanged where appropriate so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0031] Figure 1 This is a flow chart of a fluorescent clothing wearing recognition method based on target detection provided by an embodiment of the present invention. This embodiment is applicable to the case of performing fluorescent clothing wearing recognition on working objects located at a work site. The method can be executed by a fluorescent clothing wearing recognition device based on target detection. The fluorescent clothing wearing recognition device based on target detection can be implemented in the form of hardware and / or software. The fluorescent clothing wearing recognition device based on target detection can be configured in a terminal and / or server. Figure 1 As shown, the method includes:

[0032] S110. Acquire a first on-site video frame of the work site, wherein the first on-site video frame includes at least one work object located at the work site and in motion.

[0033] The work site may be an area where work is to be performed or is currently being performed. This area is where workers are required to properly wear fluorescent clothing to ensure safe and efficient work. Optionally, the work site includes road construction sites, sanitation work sites, outdoor power work sites, railway maintenance sites, construction sites, and the like. The first scene video frame may be video data captured by a camera device located at the work site, and this video data includes moving objects located at the work site. It should be noted that during video detection of the work site using the camera device, there may not be any moving objects most of the time. Continuing to perform the subsequent fluorescent clothing wear recognition process would result in unnecessary waste of resources. Therefore, moving object detection can be performed on the scene video data captured by the camera device. Furthermore, when a moving object is detected in a video frame, the video frame can be captured and used as the first scene video frame. In this case, the first scene video frame includes at least one moving object located at the work site. The object is a person working at or at the work site. Being in motion means that the position of the work object in the work site changes multiple times within a certain period of time.

[0034] In a specific implementation, a camera device can be installed at the work site to enable real-time monitoring of the situation at the work site. Furthermore, the camera device can be used to capture the work site and obtain live video data. To reduce the number of subsequent process calls, conserve the resources required for algorithm execution, and lower overall performance costs, motion detection can be performed on the obtained live video data. If a moving target is detected in any video frame, that video frame can be captured and used as the first live video frame. Furthermore, fluorescent clothing can be used to identify working subjects in motion in the first live video frame.

[0035] It should be noted that there are multiple ways to perform motion detection on live video data. Optionally, motion detection can be performed on live video data based on a background subtraction algorithm, that is, by establishing a background model and comparing the current live video frame with the background model to detect the moving target; motion detection can also be performed by the frame difference method, that is, continuously obtaining two or more frames of live video frames, and detecting the moving target by calculating the difference between the corresponding pixels between adjacent frames.

[0036] S120: Perform fluorescent clothing wearing recognition on the first scene video frame according to the pre-trained fluorescent clothing wearing recognition model to obtain a fluorescent clothing wearing result corresponding to at least one work object.

[0037] The fluorescent clothing recognition model is used to identify whether the work subject in the first-scene video frame is wearing and correctly wearing the fluorescent clothing. In other words, the fluorescent clothing recognition model can determine, based on the first-scene video frame, whether the work subject in the first-scene video frame is wearing the fluorescent clothing, and if so, whether the clothing is correctly worn. The fluorescent clothing recognition model can be trained using a deep learning model based on a set of sample images including work subjects wearing fluorescent clothing correctly, incorrectly, and without fluorescent clothing. The fluorescent clothing recognition model can be a deep learning model with any model architecture capable of object detection. In this embodiment, to improve detection accuracy, effectively reduce false alarm rates, and deliver superior performance in complex environments and varying lighting conditions, the fluorescent clothing recognition model can be a deep learning model including the YOLO-V8 model. The YOLO-V8 model has a deeper and optimized convolutional neural network structure, capable of capturing a wider variety of features from complex environmental images. Regardless of the work scene, this model can accurately extract unique features related to fluorescent clothing, avoiding misjudgments caused by background interference. Furthermore, the YOLO-V8 model can quickly adjust its learning of fluorescent clothing features based on target changes in consecutive video frames, ensuring continued accurate detection in dynamic scenes. Furthermore, in complex environments where fluorescent clothing targets may exist at varying distances, the YOLO-V8 model can detect and track fluorescent clothing at varying resolutions, accurately identifying everything from small targets at long distances to large targets at close range.

[0038] The fluorescent clothing wearing result can be used to indicate whether the work subject is wearing the fluorescent clothing correctly. The fluorescent clothing wearing result can include correctly wearing the fluorescent clothing, incorrectly wearing the fluorescent clothing, or not wearing the fluorescent clothing at all. Correctly wearing the fluorescent clothing means wearing the fluorescent clothing in accordance with the work site's operating specifications.

[0039] In a specific implementation, after obtaining the first on-site video frame, a pre-trained fluorescent clothing recognition model can be retrieved. Furthermore, the first on-site video frame can be input into the fluorescent clothing recognition model, and target detection and fluorescent clothing recognition can be performed on the work subject in the first on-site video frame based on the fluorescent clothing recognition model to determine whether the work subject in the first on-site video frame is wearing fluorescent clothing and whether the work subject is wearing it correctly. Furthermore, a fluorescent clothing wearing result corresponding to at least one work subject can be output.

[0040] S130: Determine a first object display attribute of at least one operation object in the first scene video frame based on the first scene video frame.

[0041] The first object display attribute is used to characterize the display clarity and / or display completeness of the work object in the first-scene video frame. Display clarity can be associated with the display brightness of the work object in the first-scene video frame. Optionally, display clarity is represented based on an overexposure attribute and / or an overdarkness attribute. Display completeness can be understood as the completeness of the object display area of the work object in the first-scene video. Display completeness can be represented based on an occlusion attribute. Optionally, the first object display attribute includes at least one of an occlusion attribute, an overexposure attribute, and an overdarkness attribute. The occlusion attribute is used to characterize whether the work object is obscured in the first-scene video frame. The occlusion attribute includes an obscured state or an unobstructed state. The overexposure attribute is used to characterize whether the brightness of the object display area of the work object in the first-scene video frame is too high. The overexposure attribute includes an overexposed state or a non-overexposed state. The overdarkness attribute is used to characterize whether the brightness of the object display area of the work object in the first-scene video frame is too low. The overdarkness attribute includes an overdarkness state or a non-overdarkness state.

[0042] In this embodiment, after obtaining a fluorescent clothing wearing result corresponding to at least one work object, since the fluorescent clothing wearing result is obtained based on a single frame of first-scene video, it may be invalid due to the work object's low display integrity and / or low display clarity in the first-scene video frame. To determine the validity of the fluorescent clothing wearing result, the object display attributes of the at least one work object in the first-scene video frame can be determined. Furthermore, the validity of the fluorescent clothing wearing result can be determined based on the object display attributes, and the fluorescent clothing wearing result can be processed based on the validity result. This can thereby reduce the occurrence of false alarms.

[0043] Different first object display attributes correspond to different determination methods. The following will describe the determination method of each first object display attribute separately.

[0044] Optionally, the first object display attribute includes an occlusion attribute; based on the first scene video frame, determining the first object display attribute of at least one work object in the first scene video frame includes: processing the first scene video frame according to a pre-trained occlusion detection model to obtain the occlusion attribute of at least one work object in the first scene video frame.

[0045] The occlusion detection model is at least used to determine whether at least one work object in the first scene video frame is occluded. The occlusion detection model can also determine the degree of occlusion of at least one work object in the first scene video frame, and the occlusion degree includes complete occlusion or partial occlusion. The occlusion detection model can be obtained by training a deep learning model based on a sample image set containing occluded objects and non-occluded objects. The occlusion detection model can be a deep learning model containing any model structure. Optionally, the occlusion detection model can be a deep learning model containing a mask region convolutional neural network, a fast region convolutional neural network, or a YOLO series model.

[0046] In a specific implementation, the first scene video frame can be input into a pre-trained occlusion detection model, and occlusion detection can be performed on the work object in the first scene video frame based on the occlusion detection model. By performing mask segmentation on the work object in the first scene video frame, the ratio of the area of the occluded region in the object segmentation mask to the total object segmentation mask can be determined. Furthermore, this ratio can be compared with a preset ratio threshold; if the ratio is greater than the preset ratio threshold, the occlusion attribute of the work object is determined to be in an occluded state; if the ratio is less than the preset ratio threshold, the occlusion attribute of the work object is determined to be in an unoccluded state.

[0047] It should be noted that in addition to using the occlusion detection model to determine the occlusion properties, other methods can also be used to determine the occlusion properties. For example, a contour detection algorithm can be used to extract the object contour of the work object in the first scene video frame, and then the occlusion properties of the work object can be determined through the object contour.

[0048] Optionally, the first object attribute includes an overexposure attribute; the overexposure attribute includes an overexposure state or a non-overexposure state; based on the first on-site video frame, determining the first object display attribute of at least one operating object in the first on-site video frame, including: performing color space transformation on the first on-site video frame to obtain a second on-site video frame, and performing object segmentation on the first on-site video frame to obtain an object segmentation mask map; wherein the object segmentation mask map includes an object segmentation area corresponding to at least one operating object; for at least one operating object, determining the regional luminance information corresponding to the object segmentation area according to the second on-site video frame and the object segmentation mask map, and determining the average luminance information and the luminance standard deviation based on the regional luminance information; when the average luminance information is greater than the first overexposure threshold and the luminance standard deviation is less than the second overexposure threshold, determining the overexposure attribute of the operating object to be an overexposure state; when the average luminance information is not greater than the first overexposure threshold or the luminance standard deviation is not less than the second overexposure threshold, determining the overexposure attribute of the operating object to be a non-overexposure state.

[0049] Among them, color space conversion can be a method of image processing, that is, converting an image from its current color space to another color space for representation. It is understood that color can usually be described by three independent attributes. Using different variables to represent an independent attribute constitutes a spatial coordinate, and the space corresponding to the spatial coordinate is the color space. Optionally, the color space includes RGB color space, YUV color space, HSV color space, and HSV color space. In this embodiment, the first scene video frame is a video frame extracted from the live video data, which is an RGB image composed of pixel values of the three channels R, G, and B. The RGB image represents hue, brightness, and saturation together, making it difficult to perform brightness analysis on the first scene video frame. Therefore, the first scene video frame can be subjected to color space conversion to convert the first scene video frame from an RGB image to an image in another color space. Exemplarily, the first scene video frame can be converted from the RGB color space to the HSV color space to obtain a second scene video frame represented by pixel values of the three channels H, S, and V.

[0050] Among them, the HSV color space is a way to represent pixel points in the RGB color space in an inverted cone. HSV stands for Hue, Saturation, and Value, and can also be called HSB (B stands for Brightness). Hue is the basic attribute of color, that is, the name of the color, such as red, blue, or yellow; saturation refers to the purity of the color. The higher the saturation, the purer the color, and the lower the saturation, the gradually graying of the color. Its value range can be a value of 0-100%. Lightness refers to the brightness of the color, and its value range can be a value of 0-max. The HSV color space can be described by a cone space model. At the vertex of the cone, V = 0, H and S are undefined, representing black. At the center of the top surface of the cone, V = max, S = 0, H is undefined, representing white.

[0051] The object segmentation mask image may be an image representing the object contours of each work object in the first-scene video frame. The object segmentation mask image includes at least one object segmentation region corresponding to the work object. The object segmentation mask image may be a binary image represented by two pixel values. Exemplarily, when performing object segmentation on the first-scene video frame, the pixel values within the object region where the work object is located in the first-scene video frame may be adjusted to a first value, and the pixel values of other image regions excluding the work object may be adjusted to a second value. Furthermore, the image with adjusted pixel values may be used as the object segmentation mask image.

[0052] The regional luminance information includes the luminance value of each pixel in the object segmentation region. The luminance value is the V value of each pixel in the second scene video frame. The average luminance information can be the average luminance value obtained by averaging the regional luminance information. The luminance standard deviation can be the standard deviation obtained by calculating the standard deviation of the regional luminance information. The first overexposure threshold can be any value, optionally, 0.8. The second overexposure threshold can be any value, optionally, 0.1.

[0053] In a specific implementation, a color space transformation can be performed on the first scene video frame to convert it from the RGB color space to the HSV color space, thereby obtaining a second scene video frame. Furthermore, object segmentation can be performed on the first scene video frame to obtain an object segmentation mask. Furthermore, for at least one work object, the second scene video frame and the object segmentation mask can be processed by pixel matching to obtain regional luminance information corresponding to the object segmentation region in the second scene video frame. This regional luminance information includes the luminance value of each pixel within the object segmentation region. Furthermore, an average operation is performed on the regional luminance information, and the resulting average value is used as the average luminance information. Furthermore, a standard deviation operation is performed on the regional luminance information, and the resulting standard deviation is used as the luminance standard deviation. Furthermore, the average luminance information can be compared with a first overexposure threshold, and the luminance standard deviation can be compared with a second overexposure threshold. If the average luminance information is greater than the first overexposure threshold and the luminance standard deviation is less than the second overexposure threshold, it can be determined that the brightness of the work object in the first scene video frame is too high, and the corresponding overexposure attribute is determined to be an overexposure state. When the average luminance information is not greater than the first overexposure threshold or the luminance standard deviation is not less than the second overexposure threshold, it can be determined that the brightness of the work object in the first on-site video frame is normal, and its corresponding overexposure attribute is determined to be a non-overexposure state.

[0054] Optionally, the first object attribute includes an overly dark attribute; the overly dark attribute includes an overly dark state or a not-overly dark state; based on the first on-site video frame, determining the first object display attribute of at least one work object in the first on-site video frame, including: for at least one work object, when the average brightness information corresponding to the work object is less than a first overly dark threshold, determining that the overly dark attribute of the work object is an overly dark state; when the average brightness information corresponding to the work object is not less than the first overly dark threshold, determining that the overly dark attribute of the work object is a not-overly dark state.

[0055] The first too-dark threshold may be any value, and may optionally be 0.2 or the like.

[0056] In a specific implementation, after obtaining average brightness information corresponding to at least one work object through color space transformation, object segmentation, and brightness value averaging, the average brightness information can be compared with a first excessively dark threshold. Furthermore, if the average brightness information is less than the first excessively dark threshold, the brightness of the work object in the first scene video frame can be determined to be low, and its corresponding excessively dark attribute can be determined to be in an excessively dark state. If the average brightness information is not less than the first excessively dark threshold, the brightness of the work object in the first scene video frame can be determined to be normal, and its corresponding excessively dark attribute can be determined to be in a normal state.

[0057] S140. For at least one operation object, determine a result processing method corresponding to the fluorescent clothing wearing result according to the first object display attribute corresponding to the operation object, and process the fluorescent clothing wearing result based on the result processing method.

[0058] Among them, the result processing method can be a method for processing the fluorescent clothing wearing result. Optionally, the result processing method includes direct adoption or re-identification. In this embodiment, the first object display attribute corresponding to the operation object can affect the validity of the fluorescent clothing wearing result corresponding to the operation object, and then affect the result processing method for processing the fluorescent clothing wearing result. Therefore, in order to avoid false alarms, for at least one operation object, the validity of the fluorescent clothing wearing result corresponding to the operation object can be judged based on the first object display attribute corresponding to the operation object. Further, the result processing method corresponding to the fluorescent clothing wearing result can be determined based on the validity judgment result. Then, the fluorescent clothing wearing result can be processed based on the result processing method.

[0059] In a specific implementation, for at least one work object, if the first object display attribute corresponding to the work object includes an unobstructed state, a non-overexposed state, and a non-overdark state, it can indicate that the work object has high display clarity and display integrity in the first scene video frame. In this case, the fluorescent clothing wearing result corresponding to the work object can be determined to be valid, and the corresponding result processing method can be determined to be directly adopted. Furthermore, if the fluorescent clothing wearing result includes that the fluorescent clothing is not worn or is not worn correctly, a warning message can be generated and displayed. If the fluorescent clothing wearing result includes that the fluorescent clothing is worn correctly, no processing is performed and the fluorescent clothing wearing recognition continues.

[0060] When the first object attribute corresponding to the work object includes at least one of an obscured state, an overexposed state, and an overly dark state, it can be indicated that the display clarity and / or display completeness of the work object in the first scene video frame is low. In this case, it can be determined that the fluorescent clothing wearing result corresponding to the work object is invalid, and the corresponding result processing method is determined to be re-identification. Furthermore, the updated fluorescent clothing wearing result can be determined based on the re-identification method corresponding to the attribute state included in the first object attribute. In this case, the updated fluorescent clothing wearing result is valid. Furthermore, when the updated fluorescent clothing wearing result includes not wearing the fluorescent clothing or not wearing the fluorescent clothing correctly, a warning message can be generated and displayed. When the updated fluorescent clothing wearing result includes correctly wearing the fluorescent clothing, no processing is performed and the fluorescent clothing wearing recognition continues.

[0061] The technical solution of the embodiment of the present invention is to obtain a first-site video frame of the work site, wherein the first-site video frame includes at least one work object located at the work site and in motion; before performing fluorescent clothing wearing recognition, motion detection is first performed on the captured on-site video data to obtain a first-site video frame containing the work object in motion, thereby reducing the call of the algorithm model. At this time, the algorithm consumption resources are saved and the performance cost is reduced; further, fluorescent clothing recognition is performed on the first-site video frame according to the pre-trained fluorescent clothing wearing recognition model to obtain a fluorescent clothing wearing result corresponding to at least one work object, thereby achieving the effect of fluorescent clothing wearing recognition of the work object in the first-site video frame based on the deep learning model, and improving the recognition accuracy and recognition efficiency of fluorescent clothing wearing recognition; further, based on the first-site video frame, it is determined that at least one work object is in the first The first object display attribute in the on-site video frame provides a basis for determining the validity of the subsequent fluorescent clothing wearing result; further, for at least one work object, the result processing method corresponding to the fluorescent clothing wearing result is determined based on the first object display attribute corresponding to the work object, and the fluorescent clothing wearing result is processed based on the result processing method, which solves the problem that manual recognition in related technologies consumes a lot of manpower costs, and the recognition efficiency and recognition accuracy are low, thereby affecting production operation efficiency. It realizes the fluorescent clothing wearing recognition of the work object in the on-site video frame based on the deep learning model, and also combines the object display attribute to determine the validity of the fluorescent clothing wearing result output by the model to ensure the accuracy of the fluorescent clothing wearing result, further improving the recognition accuracy and recognition efficiency of fluorescent clothing wearing recognition, thereby effectively ensuring the operation safety at the work site.

[0062] Figure 2 This is a flowchart of a method for identifying fluorescent clothing wear based on target detection, provided by an embodiment of the present invention. Based on the aforementioned embodiment, when the first object display attribute corresponding to the work object includes an unobstructed state, a non-overexposed state, and a non-dark state, the result processing method corresponding to the fluorescent clothing wearing result is determined to be direct adoption. Furthermore, when the fluorescent clothing wearing result includes not wearing fluorescent clothing or not wearing fluorescent clothing correctly, a fluorescent clothing wearing warning message is generated and displayed. For specific implementation methods, please refer to the technical solution of this embodiment. Technical terms that are the same or similar to those in the aforementioned embodiments are not repeated here.

[0063] like Figure 2 As shown, the method includes:

[0064] S210: Acquire a first on-site video frame of the work site, wherein the first on-site video frame includes at least one work object located at the work site and in motion.

[0065] S220: Perform fluorescent clothing wearing recognition on the first scene video frame according to the pre-trained fluorescent clothing wearing recognition model to obtain a fluorescent clothing wearing result corresponding to at least one work object.

[0066] S230: Determine a first object display attribute of at least one operation object in the first scene video frame based on the first scene video frame.

[0067] S240. For at least one work object, when the first object display attribute corresponding to the work object includes an unobstructed state, a non-overexposed state, and a non-overdark state, determine that the result processing method corresponding to the fluorescent clothing wearing result is direct adoption, and when the fluorescent clothing wearing result includes not wearing the fluorescent clothing or not wearing the fluorescent clothing correctly, generate fluorescent clothing wearing warning information and display the fluorescent clothing wearing warning information.

[0068] Among them, direct adoption means directly using the fluorescent clothing wearing result as the final wearing recognition result. The fluorescent clothing wearing warning information can be used to warn that there are workers at the work site who are not wearing fluorescent clothing or are not wearing fluorescent clothing correctly. The fluorescent clothing wearing warning information includes multiple pieces of information associated with the worker being warned. Optionally, the fluorescent clothing wearing warning information includes the worker's location information at the work site, fluorescent clothing warning prompt text (such as "The person is not wearing the fluorescent clothing correctly"), and warning prompt sound information.

[0069] In a specific implementation, for at least one work object, when the first object display attribute corresponding to the work object includes an unobstructed state, a non-overexposed state, and a non-overdark state, it can be determined that the result processing method of the fluorescent clothing wearing result corresponding to the work object is direct adoption. Furthermore, the fluorescent clothing wearing result can be detected. When the fluorescent clothing wearing result includes not wearing fluorescent clothing or not wearing fluorescent clothing correctly, fluorescent clothing wearing warning information can be generated and the generated fluorescent clothing wearing warning information can be displayed. The fluorescent clothing wearing warning information can be displayed on the display interface of any terminal device associated with the work site. Optionally, the fluorescent clothing wearing warning information can be displayed on the display interface of the background system terminal device; or, the fluorescent clothing wearing warning information can also be displayed on the display interface of the terminal device belonging to the relevant staff, etc.

[0070] The technical solution of the embodiment of the present invention determines, for at least one work object, that the result processing method corresponding to the fluorescent clothing wearing result is direct adoption when the first object display attribute corresponding to the work object includes an unobstructed state, a non-overexposed state, and a non-overdark state; and generates fluorescent clothing wearing warning information when the fluorescent clothing wearing result includes not wearing fluorescent clothing or not wearing fluorescent clothing correctly, and displays the fluorescent clothing wearing warning information, thereby achieving the effect of accurately and effectively warning the work objects who are not wearing fluorescent clothing or not wearing fluorescent clothing correctly in the work site based on the collected on-site video frames, thereby enhancing the safety factor of the work site, and determining the processing method of the fluorescent clothing wearing result through the object display attribute, thereby effectively reducing the probability of false warning events, and significantly improving the robustness and safety of the fluorescent clothing wearing recognition system.

[0071] Figure 3 This is a flowchart of a method for identifying fluorescent clothing wearers based on target detection, provided by an embodiment of the present invention. Based on the aforementioned embodiment, when the first object display attribute corresponding to the work object includes at least one of an obscured state, an overexposed state, and an overly dark state, the result processing method corresponding to the fluorescent clothing wear result is determined to be re-identification, and the fluorescent clothing wear result is processed based on the result processing method. For detailed implementation details, please refer to the technical solution of this embodiment. Technical terms that are identical or similar to those in the aforementioned embodiments are not repeated here.

[0072] like Figure 3 As shown, the method includes:

[0073] S310: Acquire a first on-site video frame of the work site, wherein the first on-site video frame includes at least one work object located at the work site and in motion.

[0074] S320: Perform fluorescent clothing wearing recognition on the first scene video frame according to the pre-trained fluorescent clothing wearing recognition model to obtain a fluorescent clothing wearing result corresponding to at least one work object.

[0075] S330: Determine a first object display attribute of at least one operation object in the first scene video frame based on the first scene video frame.

[0076] S340. For at least one work object, when the first object display attribute corresponding to the work object includes at least one of an obscured state, an overexposed state, and an overly dark state, determine that the result processing method corresponding to the fluorescent clothing wearing result is re-identification, and process the fluorescent clothing wearing result based on the result processing method.

[0077] Among them, re-identification can be understood as re-executing the fluorescent clothing wearing identification process.

[0078] It should be noted that, when the first object display attribute corresponding to the job object includes different attribute information, the corresponding re-identification methods are also different, which will be explained separately below.

[0079] Optionally, the fluorescent clothing wearing result is processed based on the result processing method, including: when the display attribute of the first object includes an occluded state, obtaining a third on-site video frame containing the work object, re-identifying the third on-site video frame, and obtaining an updated fluorescent clothing wearing result.

[0080] In this embodiment, when the first object display attribute includes an obscured state, it indicates that the work object is in an obscured state in the first on-site video frame. Because the work object is in motion in the work site, in this case, the work object can be re-identified by re-acquiring other on-site video frames containing the work object.

[0081] The third on-site video frame may be a on-site video frame containing the work object that is re-extracted from the on-site video data, and the timestamp of the on-site video frame is close to the timestamp of the first on-site video frame. The third on-site video frame may be a on-site video frame acquired by the camera device that acquires the first on-site video frame. In this case, the on-site video frame may be a on-site video frame before the timestamp of the first on-site video frame, or a on-site video frame after the timestamp of the first on-site video frame; or, the third on-site video frame may be acquired by other camera devices that shoot the work site, that is, a on-site video frame obtained after shooting the work object from other shooting angles. The third on-site video frame may be consistent with the timestamp of the first on-site video frame, or may be a on-site video frame before the timestamp of the first on-site video frame, or may be a on-site video frame after the timestamp of the first on-site video frame.

[0082] In a specific implementation, for at least one work object, if the first object display attribute corresponding to the work object includes an obscured state, a third live video frame containing the work object can be obtained. Furthermore, fluorescent clothing wearing recognition can be performed on the third live video frame based on a fluorescent clothing wearing recognition model to obtain a new fluorescent clothing wearing result. Subsequently, based on the third video frame, the second object display attribute corresponding to the work object is determined. Furthermore, if the second object display attribute includes an unobstructed state, a non-overexposed state, and a non-dark state, the new fluorescent clothing wearing result is used as the updated fluorescent clothing wearing result. If the updated fluorescent clothing wearing result includes the fluorescent clothing not being worn or not being worn correctly, a fluorescent clothing wearing warning message is generated and displayed.

[0083] Optionally, the fluorescent clothing wearing result is processed based on the result processing method, including: when the first object display attribute includes an overexposed state and / or an overly dark state, adjusting the shooting parameters of the camera device, and collecting a fourth scene video frame through the adjusted camera device, re-identifying the fourth scene video frame, and obtaining an updated fluorescent clothing wearing result.

[0084] In this embodiment, if the first object display attribute includes an overexposed state and / or an overly dark state, it may indicate that the overall image clarity of the first scene video frame is low. In this case, the shooting parameters of the camera device can be adjusted and the scene video frame can be recaptured; or, the first scene video frame can be enhanced to obtain a new scene video frame. Then, the new scene video frame can be re-recognized to obtain an updated fluorescent clothing wearing result.

[0085] The camera device may be the camera device used to capture the first on-site video frame. Shooting parameters refer to various values and options that can be adjusted and set when shooting with the camera device, and these parameters directly affect the shooting effect. Optionally, shooting parameters include aperture, shutter speed, sensitivity, and white balance. The fourth on-site video frame may be a live video frame obtained by re-shooting the work site using the adjusted camera device, and the live video frame includes at least one work object located at the work site and in motion.

[0086] In a specific implementation, when the first object display attribute includes an overexposed state and / or an overly dark state, the shooting parameters of the camera device can be adjusted, and the adjusted camera device can be used to shoot the work site to obtain a fourth scene video frame. Furthermore, the fourth scene video frame can be used to identify the wearing of fluorescent clothing according to the fluorescent clothing wearing recognition model to obtain a new fluorescent clothing wearing result. Thereafter, based on the fourth video frame, the third object display attribute corresponding to the work object is determined. Furthermore, when the third object display attribute includes an unobstructed state, a non-overexposed state, and a non-overly dark state, the new fluorescent clothing wearing result is used as the updated fluorescent clothing wearing result. When the updated fluorescent clothing wearing result includes not wearing fluorescent clothing or not wearing fluorescent clothing correctly, a fluorescent clothing wearing warning message is generated and displayed.

[0087] In this embodiment, re-identification can also be performed by performing image enhancement on the first scene video frame. In a specific implementation, if the first object display attributes include an overexposed state and / or an overly dark state, the first scene video frame is image enhanced to obtain an enhanced first scene video frame. Based on the enhanced first scene video frame, the object display attributes and the fluorescent clothing wearing result of the work object in the enhanced first scene video frame are determined. Furthermore, if the obtained object display attributes include an unobstructed state, a non-overexposed state, and a non-overly dark state, the obtained fluorescent clothing wearing result is used as the updated first fluorescent clothing wearing result. If the first fluorescent clothing wearing result includes not wearing fluorescent clothing or not wearing fluorescent clothing correctly, and if the updated fluorescent clothing wearing result includes not wearing fluorescent clothing or not wearing fluorescent clothing correctly, a fluorescent clothing wearing warning message is generated and displayed. Image enhancement includes histogram equalization and / or a retinal adaptive algorithm.

[0088] The technical solution of the embodiment of the present invention determines that the result processing method corresponding to the fluorescent clothing wearing result is re-identification for at least one work object when the first object display attribute corresponding to the work object includes at least one of the blocked state, overexposed state and too dark state, and processes the fluorescent clothing wearing result based on the result processing method, thereby achieving the effect of accurately and effectively warning the work objects who are not wearing fluorescent clothing or are not wearing fluorescent clothing correctly in the work site based on the collected on-site video frames, thereby enhancing the safety factor of the work site, and, by determining the processing method of the fluorescent clothing wearing result through the object display attribute, effectively reduces the probability of false warning events, and significantly improves the robustness and safety of the fluorescent clothing wearing recognition system.

[0089] Figure 4 FIG is a schematic diagram of a fluorescent clothing wearing recognition device based on target detection provided by an embodiment of the present invention. Figure 4 As shown, the device includes: a video frame acquisition module 410, a fluorescent clothing wearing recognition module 420, an object display attribute determination module 430 and a wearing result processing module 440.

[0090] Among them, the video frame acquisition module 410 is used to obtain a first-site video frame of the work site, wherein the first-site video frame includes at least one work object located at the work site and in motion; the fluorescent clothing wearing recognition module 420 is used to perform fluorescent clothing wearing recognition on the first-site video frame according to a pre-trained fluorescent clothing wearing recognition model to obtain a fluorescent clothing wearing result corresponding to at least one work object; wherein the fluorescent clothing wearing recognition model is used to identify whether the work object in the first-site video frame is wearing and wearing the fluorescent clothing correctly; the object display attribute determination module 430 is used to determine the first object display attribute of at least one work object in the first-site video frame based on the first-site video frame; wherein the first object display attribute is used to characterize the display clarity and / or display completeness of the work object in the first-site video frame; the wearing result processing module 440 is used to determine, for at least one work object, a result processing method corresponding to the fluorescent clothing wearing result based on the first object display attribute corresponding to the work object, and process the fluorescent clothing wearing result based on the result processing method.

[0091] The technical solution of the embodiment of the present invention is to obtain a first-site video frame of the work site, wherein the first-site video frame includes at least one work object located at the work site and in motion; before performing fluorescent clothing wearing recognition, motion detection is first performed on the captured on-site video data to obtain a first-site video frame containing the work object in motion, thereby reducing the call of the algorithm model. At this time, the algorithm consumption resources are saved and the performance cost is reduced; further, fluorescent clothing recognition is performed on the first-site video frame according to the pre-trained fluorescent clothing wearing recognition model to obtain a fluorescent clothing wearing result corresponding to at least one work object, thereby achieving the effect of fluorescent clothing wearing recognition of the work object in the first-site video frame based on the deep learning model, and improving the recognition accuracy and recognition efficiency of fluorescent clothing wearing recognition; further, based on the first-site video frame, it is determined that at least one work object is in the first The first object display attribute in the on-site video frame provides a basis for determining the validity of the subsequent fluorescent clothing wearing result; further, for at least one work object, the result processing method corresponding to the fluorescent clothing wearing result is determined based on the first object display attribute corresponding to the work object, and the fluorescent clothing wearing result is processed based on the result processing method, which solves the problem that manual recognition in related technologies consumes a lot of manpower costs, and the recognition efficiency and recognition accuracy are low, thereby affecting production operation efficiency. It realizes the fluorescent clothing wearing recognition of the work object in the on-site video frame based on the deep learning model, and also combines the object display attribute to determine the validity of the fluorescent clothing wearing result output by the model to ensure the accuracy of the fluorescent clothing wearing result, further improving the recognition accuracy and recognition efficiency of fluorescent clothing wearing recognition, thereby effectively ensuring the operation safety at the work site.

[0092] Optionally, the first object display attribute includes an occlusion attribute; the object display attribute determination module 430 includes an occlusion attribute determination unit. The occlusion attribute determination unit is configured to process the first scene video frame based on a pre-trained occlusion detection model to obtain an occlusion attribute of at least one work object in the first scene video frame; the occlusion detection model is configured to at least determine whether the at least one work object in the first scene video frame is occluded; the occlusion detection model is trained using a deep learning model based on a sample image set containing occluded objects and non-occluded objects; and the occlusion attribute includes an occluded state or an unoccluded state.

[0093] Optionally, the first object attribute includes an overexposure attribute; the overexposure attribute includes an overexposure state or a non-overexposure state; the object display attribute determination module 430 includes: a segmentation mask image determination unit, a regional brightness information determination unit, an overexposure state determination unit and a non-overexposure state determination unit. Among them, the segmentation mask map determination unit is used to perform color space transformation on the first scene video frame to obtain the second scene video frame, and perform object segmentation on the first scene video frame to obtain the object segmentation mask map; wherein the object segmentation mask map includes an object segmentation area corresponding to at least one work object; the regional brightness information determination unit is used to determine the regional brightness information corresponding to the object segmentation area for at least one work object according to the second scene video frame and the object segmentation mask map, and determine the average brightness information and the brightness standard deviation based on the regional brightness information; wherein the regional brightness information includes the brightness value of each pixel in the object segmentation area; the overexposure state determination unit is used to determine that the overexposure attribute of the work object is an overexposure state when the average brightness information is greater than the first overexposure threshold and the brightness standard deviation is less than the second overexposure threshold; the non-overexposure state determination unit is used to determine that the overexposure attribute of the work object is a non-overexposure state when the average brightness information is not greater than the first overexposure threshold or the brightness standard deviation is not less than the second overexposure threshold.

[0094] Optionally, the first object attribute includes an overly dark attribute; the overly dark attribute includes an overly dark state or a non-overly dark state; and the object display attribute determination module 430 includes: an overly dark state determination unit and a non-overly dark state determination unit. The overly dark state determination unit is configured to, for at least one work object, determine that the overly dark attribute of the work object is an overly dark state if average brightness information corresponding to the work object is less than a first overly dark threshold; and the non-overly dark state determination unit is configured to determine that the overly dark attribute of the work object is a non-overly dark state if average brightness information corresponding to the work object is not less than the first overly dark threshold.

[0095] Optionally, the wearing result processing module 440 includes: a first result processing method determination unit and a second result processing method determination unit. The first result processing method determination unit is configured to determine that the result processing method corresponding to the fluorescent clothing wearing result is direct adoption when the first object display attribute corresponding to the operation object includes an unobstructed state, a non-overexposed state, and a non-overdark state; and the second result processing method determination unit is configured to determine that the result processing method corresponding to the fluorescent clothing wearing result is re-identification when the first object display attribute corresponding to the operation object includes at least one of an obstructed state, an overexposed state, and an overdark state.

[0096] Optionally, the result processing method is to directly adopt the result. The wearing result processing module 440 includes: a warning information generation unit. The warning information generation unit is used to generate a fluorescent clothing wearing warning message when the fluorescent clothing wearing result includes that the fluorescent clothing is not worn or is not worn correctly, and display the fluorescent clothing wearing warning message.

[0097] Optionally, the result processing method is re-identification; the wearing result processing module 440 includes: a re-identification unit; the re-identification unit is used to execute one or more of the following: when the first object display attribute includes an obscured state, obtaining a third on-site video frame including the work object, re-identifying the third on-site video frame, and obtaining an updated fluorescent clothing wearing result; when the first object display attribute includes an overexposed state and / or an over-dark state, adjusting the shooting parameters of the camera device, and collecting a fourth on-site video frame through the adjusted camera device, re-identifying the fourth on-site video frame, and obtaining an updated fluorescent clothing wearing result.

[0098] The fluorescent clothing wearing recognition device based on target detection provided in an embodiment of the present invention can execute the fluorescent clothing wearing recognition method based on target detection provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the execution method.

[0099] Figure 5 A schematic diagram of the structure of an electronic device 10 that can be used to implement an embodiment of the present invention is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices (such as helmets, glasses, watches, etc.) and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present invention described and / or claimed herein.

[0100] like Figure 5As shown, the electronic device 10 includes at least one processor 11 and a memory, such as a read-only memory (ROM) 12, a random access memory (RAM) 13, etc., which is communicatively connected to the at least one processor 11. The memory stores a computer program that can be executed by the at least one processor. The processor 11 can perform various appropriate actions and processes according to the computer program stored in the read-only memory (ROM) 12 or the computer program loaded from the storage unit 18 into the random access memory (RAM) 13. Various programs and data required for the operation of the electronic device 10 can also be stored in the RAM 13. The processor 11, ROM 12, and RAM 13 are connected to each other via a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.

[0101] Multiple components in the electronic device 10 are connected to the I / O interface 15, including an input unit 16, such as a keyboard, a mouse, etc.; an output unit 17, such as various types of displays, speakers, etc.; a storage unit 18, such as a magnetic disk, an optical disk, etc.; and a communication unit 19, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 19 allows the electronic device 10 to exchange information / data with other devices via a computer network such as the Internet and / or various telecommunication networks.

[0102] The processor 11 can be any general-purpose and / or specialized processing component with processing and computing capabilities. Some examples of the processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various specialized artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any appropriate processor, controller, microcontroller, etc. The processor 11 executes the various methods and processes described above, such as the fluorescent clothing wear recognition method based on target detection.

[0103] In some embodiments, the fluorescent clothing wearing recognition method based on target detection can be implemented as a computer program, which is tangibly contained in a computer-readable storage medium, such as a storage unit 18. In some embodiments, part or all of the computer program can be loaded and / or installed on the electronic device 10 via the ROM 12 and / or the communication unit 19. When the computer program is loaded into the RAM 13 and executed by the processor 11, one or more steps of the fluorescent clothing wearing recognition method based on target detection described above can be performed. Alternatively, in other embodiments, the processor 11 can be configured to execute the fluorescent clothing wearing recognition method based on target detection by any other appropriate means (for example, by means of firmware).

[0104] Various embodiments of the systems and techniques described herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), system-on-chip systems (SOCs), programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments can include being implemented in one or more computer programs that are executable and / or interpreted on a programmable system that includes at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0105] Computer programs for implementing the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when the computer program is executed by the processor, the functions / operations specified in the flowcharts and / or block diagrams are implemented. The computer program may be executed entirely on the machine, partially on the machine, as a stand-alone software package, partially on the machine and partially on a remote machine, or entirely on a remote machine or server.

[0106] In the context of the present invention, computer-readable storage media can be tangible media that can contain or store a computer program for use with an instruction execution system, device or equipment or used in combination with an instruction execution system, device or equipment. Computer-readable storage media can include but are not limited to electronic, magnetic, optical, electromagnetic, infrared or semiconductor systems, devices or equipment, or any suitable combination of the foregoing. Alternatively, computer-readable storage media can be machine-readable signal media. More specific examples of machine-readable storage media can include electrical connections based on one or more lines, portable computer disks, hard disks, random access memories (RAM), read-only memories (ROM), erasable programmable read-only memories (EPROM or flash memory), optical fibers, portable compact disk read-only memories (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0107] To provide interaction with a user, the systems and techniques described herein can be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the electronic device. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0108] The systems and techniques described herein can be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user computer with a graphical user interface or web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), a target blockchain network, and the Internet.

[0109] A computing system may include clients and servers. The clients and servers are typically remote from each other and typically interact via a communication network. This client-server relationship arises through computer programs running on the respective computers, creating a client-server relationship. The server may be a cloud server, also known as a cloud computing server or cloud host. This server is a hosting product within the cloud computing service ecosystem that addresses the management difficulties and limited scalability of traditional physical hosting and VPS services.

[0110] In particular, according to an embodiment of the present invention, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present invention includes a computer program product that includes a computer program carried on a non-transitory computer-readable medium, the computer program containing program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network via the communication unit 19, or installed from the storage unit 18, or installed from the ROM 12. When the computer program is executed by the processor 11, the above-mentioned functions defined in the method of the embodiment of the present invention are performed.

[0111] It should be understood that the various forms of the processes shown above can be used to reorder, add, or delete steps. For example, the steps described in the present invention can be performed in parallel, sequentially, or in a different order, as long as the desired results of the technical solution of the present invention can be achieved. This is not limited herein.

[0112] The above specific embodiments do not limit the scope of protection of the present invention. Those skilled in the art will appreciate that various modifications, combinations, sub-combinations, and substitutions may be made based on design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention are intended to be included within the scope of protection of the present invention.

Claims

1. A fluorescent clothing wearing recognition method based on target detection, characterized in that: include: Acquire a first on-site video frame of a work site, wherein the first on-site video frame includes at least one work object located at the work site and in motion; Performing fluorescent clothing recognition on the first scene video frame according to a pre-trained fluorescent clothing recognition model to obtain a fluorescent clothing wearing result corresponding to at least one of the work subjects; wherein the fluorescent clothing recognition model is used to identify whether the work subject in the first scene video frame is wearing and correctly wearing the fluorescent clothing; Determining, based on the first live video frame, a first object display attribute of at least one of the operation objects in the first live video frame; wherein the first object display attribute is used to characterize display clarity and / or display completeness of the operation object in the first live video frame; For at least one of the operation objects, a result processing method corresponding to the fluorescent clothing wearing result is determined according to the first object display attribute corresponding to the operation object, and the fluorescent clothing wearing result is processed based on the result processing method.

2. The fluorescent clothing wearing recognition method based on target detection according to claim 1 is characterized in that: The first object display attribute includes an occlusion attribute; and determining the first object display attribute of at least one of the operation objects in the first live video frame based on the first live video frame includes: Processing the first scene video frame according to a pre-trained occlusion detection model to obtain an occlusion attribute of at least one of the work objects in the first scene video frame; In which, the occlusion detection model is at least used to determine whether at least one of the work objects in the first on-site video frame is obscured; the occlusion detection model is obtained by training a deep learning model based on a sample image set containing obscured objects and non-obscured objects; the occlusion attribute includes an obscured state or an unobscured state.

3. The fluorescent clothing wearing recognition method based on target detection according to claim 1 is characterized in that: The first object attribute includes an overexposure attribute; the overexposure attribute includes an overexposure state or a non-overexposure state; The determining, based on the first live video frame, a first object display attribute of at least one of the operation objects in the first live video frame includes: Performing color space transformation on the first on-site video frame to obtain a second on-site video frame, and performing object segmentation on the first on-site video frame to obtain an object segmentation mask map; wherein the object segmentation mask map includes at least one object segmentation region corresponding to the operation object; For at least one of the work objects, determining regional luminance information corresponding to the object segmentation region based on the second on-site video frame and the object segmentation mask image, and determining average luminance information and luminance standard deviation based on the regional luminance information; wherein the regional luminance information includes the luminance value of each pixel in the object segmentation region; When the average brightness information is greater than a first overexposure threshold and the brightness standard deviation is less than a second overexposure threshold, determining that the overexposure attribute of the operation object is an overexposure state; When the average brightness information is not greater than a first overexposure threshold or the brightness standard deviation is not less than a second overexposure threshold, it is determined that the overexposure attribute of the operation object is a non-overexposure state.

4. The fluorescent clothing wearing recognition method based on target detection according to claim 3 is characterized in that: The first object attribute includes an overly dark attribute; the overly dark attribute includes an overly dark state or a not overly dark state; The determining, based on the first live video frame, a first object display attribute of at least one of the operation objects in the first live video frame includes: For at least one of the operation objects, when the average brightness information corresponding to the operation object is less than a first too-dark threshold, determining that the too-dark attribute of the operation object is in an too-dark state; When the average brightness information corresponding to the operation object is not less than a first too-dark threshold, it is determined that the too-dark attribute of the operation object is in a not-too-dark state.

5. The fluorescent clothing wearing recognition method based on target detection according to claim 1 is characterized in that: The determining, based on the first object display attribute corresponding to the operation object, a result processing method corresponding to the fluorescent clothing wearing result includes: When the first object display attribute corresponding to the operation object includes an unblocked state, a non-overexposed state, and a non-dark state, determining that the result processing method corresponding to the fluorescent clothing wearing result is directly adopted; When the first object display attribute corresponding to the operation object includes at least one of an obscured state, an overexposed state, and an overly dark state, the result processing method corresponding to the fluorescent clothing wearing result is determined to be re-identification.

6. The fluorescent clothing wearing recognition method based on target detection according to claim 5 is characterized in that: The result processing method is direct adoption; the processing of the fluorescent clothing wearing result based on the result processing method includes: In a case where the fluorescent clothing wearing result includes that the fluorescent clothing is not worn or the fluorescent clothing is not worn correctly, fluorescent clothing wearing warning information is generated and displayed.

7. The fluorescent clothing wearing recognition method based on target detection according to claim 5 is characterized in that: The result processing method is re-identification; the processing of the fluorescent clothing wearing result based on the result processing method includes one or more of the following: When the first object display attribute includes an obscured state, obtaining a third scene video frame including the work object, and re-identifying the third scene video frame to obtain an updated fluorescent clothing wearing result; When the first object display attribute includes an overexposed state and / or an overly dark state, the shooting parameters of the camera device are adjusted, and the fourth scene video frame is captured by the adjusted camera device, and the fourth scene video frame is re-identified to obtain an updated fluorescent clothing wearing result.

8. A fluorescent clothing wearing recognition device based on target detection, characterized in that: include: A video frame acquisition module is configured to acquire a first on-site video frame of a work site, wherein the first on-site video frame includes at least one work object located at the work site and in motion; a fluorescent clothing wearing recognition module, configured to perform fluorescent clothing wearing recognition on the first scene video frame based on a pre-trained fluorescent clothing wearing recognition model, and obtain a fluorescent clothing wearing result corresponding to at least one of the work subjects; wherein the fluorescent clothing wearing recognition model is configured to identify whether the work subject in the first scene video frame is wearing and correctly wearing the fluorescent clothing; an object display attribute determination module, configured to determine, based on the first live video frame, a first object display attribute of at least one of the operation objects in the first live video frame; wherein the first object display attribute is used to characterize display clarity and / or display completeness of the operation object in the first live video frame; The wearing result processing module is used to determine, for at least one of the operation objects, a result processing method corresponding to the fluorescent clothing wearing result according to the first object display attribute corresponding to the operation object, and process the fluorescent clothing wearing result based on the result processing method.

9. An electronic device, characterized in that: The electronic device comprises: at least one processor; and a memory communicatively connected to the at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the fluorescent clothing wearing recognition method based on target detection according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to implement the fluorescent clothing wearing recognition method based on target detection according to any one of claims 1 to 7 when executed.

Images