A method, device and computer readable storage medium for detecting a falling object

Abstract

This application discloses a falling object detection method, apparatus, and computer-readable storage medium. The falling object detection method includes: acquiring a moving target region image in an image to be identified; tracking the moving target in the moving target region image to obtain the trajectory information of the moving target; and determining whether a falling object phenomenon has occurred in the image to be identified based on the moving target trajectory information; wherein the falling object phenomenon includes a change in the position of the moving target in a reference direction. Through the above methods, this application can improve the accuracy of falling object detection.

Description

Technical Field

[0001] This application relates to the field of target detection technology, specifically to a method, apparatus, and computer-readable storage medium for detecting falling objects. Background Technology

[0002] In fields such as power energy and petrochemicals, it is necessary to monitor areas where falling objects may occur, so as to issue timely warnings when such phenomena occur and thus ensure production safety. Currently, the accuracy of falling object detection methods relying on deep learning models is limited by the training sample set. However, it is difficult to collect training sample images of falling objects in application scenarios, resulting in poor detection accuracy of deep learning models. Summary of the Invention

[0003] This application provides a method, apparatus, and computer-readable storage medium for detecting falling objects, which can improve the accuracy of falling object detection.

[0004] To solve the above-mentioned technical problems, the technical solution adopted in this application is: to provide a falling object detection method, which includes: acquiring a moving target region image in an image to be identified; tracking the moving target in the moving target region image to obtain moving target trajectory information; and determining whether a falling object phenomenon has occurred in the image to be identified based on the moving target trajectory information; wherein, the falling object phenomenon includes a change in the position of the moving target in a reference direction.

[0005] To solve the above-mentioned technical problems, another technical solution adopted in this application is: to provide a falling object detection device, which includes a memory and a processor connected to each other, wherein the memory is used to store a computer program, and when the computer program is executed by the processor, it is used to implement the falling object detection method in the above-mentioned technical solution.

[0006] To solve the above-mentioned technical problems, another technical solution adopted in this application is to provide a computer-readable storage medium for storing a computer program, which, when executed by a processor, is used to implement the falling object detection method in the above-mentioned technical solution.

[0007] The beneficial effects of this application through the above scheme are: to obtain the image of the moving target region in the image to be identified, and then to track the moving target in the image of the moving target region to obtain the trajectory information of the moving target, thereby determining whether a falling object phenomenon has occurred in the image to be identified through the trajectory information of the moving target. Compared with the existing methods that rely on deep learning models for falling object detection, this method can greatly improve the accuracy of falling object detection. Attached Figure Description

[0008] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Wherein:

[0009] Figure 1 This is a flowchart illustrating an embodiment of the falling object detection method provided in this application;

[0010] Figure 2 This is a flowchart illustrating another embodiment of the falling object detection method provided in this application;

[0011] Figure 3 This is a flowchart illustrating another embodiment of the falling object detection method provided in this application;

[0012] Figure 4 The image to be identified provided in this application;

[0013] Figure 5 It is the mask image provided in this application;

[0014] Figure 6 This is a flowchart illustrating an embodiment of step 33 provided in this application;

[0015] Figure 7 It is the enhanced mask image provided in this application;

[0016] Figure 8 This is the target frame provided in this application;

[0017] Figure 9 This is the region image corresponding to the target bounding box provided in this application;

[0018] Figure 10 This is the region image corresponding to the enlarged target box provided in this application;

[0019] Figure 11 This is a schematic diagram of the structure of an embodiment of the falling object detection device provided in this application;

[0020] Figure 12 This is a schematic diagram of an embodiment of the computer-readable storage medium provided in this application. Detailed Implementation

[0021] The present application will now be described in further detail with reference to the accompanying drawings and embodiments. It should be particularly noted that the following embodiments are for illustrative purposes only and do not limit the scope of the application. Similarly, the following embodiments are only some, not all, embodiments of the present application, and all other embodiments obtained by those skilled in the art without inventive effort are within the scope of protection of the present application.

[0022] In this application, the reference to "embodiment" means that a specific feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0023] It should be noted that the terms "first," "second," and "third" in this application are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Therefore, a feature defined as "first," "second," or "third" may explicitly or implicitly include at least one of that feature. In the description of this application, "multiple" means at least two, such as two, three, etc., unless otherwise explicitly specified. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or apparatus that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or apparatuses.

[0024] Please see Figure 1 , Figure 1 This is a flowchart illustrating an embodiment of the falling object detection method provided in this application. The method includes:

[0025] Step 11: Obtain the image of the moving target region in the image to be identified.

[0026] The image to be identified can be a monitoring image collected in real time by the monitoring device at the monitoring site. For example, if the falling object detection method in this embodiment is applied to the dripping detection of pipelines in a factory, the monitoring image collected by the monitoring device that monitors the pipeline can be used as the image to be identified, so as to use the image to be identified to detect dripping in the pipeline. The application scenario is not limited here.

[0027] Moving targets in the image to be identified can be detected to obtain a moving target region image. Specifically, the moving target region image can be a region image that includes the moving target. Moving target detection can be achieved using moving target detection algorithms in the field of target detection technology, such as Gaussian mixture background modeling method or inter-frame difference method, etc., which will not be described in detail or limited here.

[0028] Step 12: Track the moving target in the moving target region image to obtain the moving target trajectory information.

[0029] Determine whether the target classification module meets the preset accuracy condition. If the target classification module does not meet the preset accuracy condition, it means that the target classification module has low detection accuracy for falling objects. In this case, the moving target in the moving target area image can be tracked to obtain the trajectory information of the moving target. Specifically, the target tracking algorithm in the field of target detection technology can be used to achieve target tracking, which will not be described in detail or limited here.

[0030] Step 13: Determine whether a falling object has occurred in the image to be identified based on the trajectory information of the moving target.

[0031] Falling objects occur when the position of a moving target changes in a reference direction. After obtaining the trajectory information of the moving target, this information can be used to detect falling objects, thereby improving the accuracy of falling object detection. Specifically, the reference direction can be set according to actual needs; it can be the direction of gravity or a direction deviating from the direction of gravity at a certain angle, and is not limited here. The trajectory of the moving target can be obtained based on the trajectory information, thereby determining whether the position of the moving target has changed in the reference direction, and thus determining whether a falling object has occurred in the image to be identified.

[0032] This embodiment acquires an image of a moving target region in the image to be identified, and then tracks the moving target in the moving target region image to obtain the trajectory information of the moving target. Thus, it determines whether a falling object has occurred in the image to be identified by using the trajectory information of the moving target. Compared with the existing technology that relies on deep learning models for falling object detection, this method can greatly improve the accuracy of falling object detection.

[0033] Please see Figure 2 , Figure 2 This is a flowchart illustrating another embodiment of the falling object detection method provided in this application. The method includes:

[0034] Step 21: Obtain the image of the moving target region in the image to be identified.

[0035] Step 21 is the same as step 11 in the above embodiments, and will not be repeated here.

[0036] Step 22: Track the moving target in the moving target region image to obtain the moving target trajectory information.

[0037] Step 22 is the same as step 12 in the above embodiments, and will not be repeated here.

[0038] Step 23: Determine whether a falling object has occurred in the image to be identified based on the trajectory information of the moving target.

[0039] Step 23 is the same as step 13 in the above embodiments, and will not be repeated here.

[0040] Step 24: In response to determining that a falling object has occurred based on the trajectory information of the moving target, train the target classification module using the image of the moving target region, and return to the step of obtaining the image to be identified.

[0041] In response to determining that a falling object has occurred based on the trajectory information of the moving target, the target classification module can be trained using the image of the moving target region where the falling object occurred, and the process returns to the step of acquiring the image to be identified, continuing to execute the falling object detection task for the next image to be identified; in response to determining that no falling object has occurred based on the trajectory information of the moving target, the process returns to step 21. Specifically, incremental learning of the target classification module using the image of the moving target region where the falling object has occurred, detected based on the trajectory information of the moving target, allows the target classification module to simultaneously learn its falling object detection capabilities while maintaining accuracy. Furthermore, it eliminates the need for manual labeling of training samples, significantly saving costs and improving efficiency. Specifically, the target classification module can be a conventional model for target classification in the field of target detection technology; the specific structure and type of the target classification module will not be detailed or limited here.

[0042] Step 25: In response to the target classification module's accuracy in classifying and recognizing the moving target region image meeting the preset accuracy condition, the target classification module is used to determine whether a falling object phenomenon has occurred in the image to be identified.

[0043] As the above-mentioned object detection task based on target trajectory information continues, the images of moving target regions used to train the target classification module are continuously accumulated, which improves the detection accuracy of the target classification module for falling objects. Until the target classification module meets the preset accuracy condition, it means that the target classification module has been relatively accurate in detecting falling objects. At this time, the target classification module can be directly used to determine whether falling objects have occurred in the image to be identified, thereby improving the detection efficiency.

[0044] Specifically, the detection accuracy of the target classification module can be tested using a preset test set to determine whether the target classification module meets the preset accuracy conditions. In other embodiments, other accuracy detection methods can also be used for testing, and no limitation is made to the accuracy detection method here.

[0045] This embodiment detects moving targets in the image to be identified, obtaining an image of the moving target region. When the target classification module does not meet the preset accuracy conditions, it determines whether a falling object has occurred in the image based on the trajectory information of the moving target, thereby improving the accuracy of falling object detection. Simultaneously, when the moving target region image where a falling object has occurred is detected using the trajectory information of the moving target, the target classification module is trained using this image. This allows the target classification module to learn the ability to detect falling objects while maintaining the accuracy of falling object detection, and it eliminates the need for manual labeling of training samples, greatly saving costs and improving efficiency. Furthermore, when the target classification module meets the preset accuracy conditions, it can be directly used to perform falling object detection, further improving the efficiency of falling object detection.

[0046] Please see Figure 3 , Figure 3 This is a flowchart illustrating another embodiment of the falling object detection method provided in this application. The method includes:

[0047] Step 31: Obtain the image to be recognized.

[0048] Step 31 is the same as obtaining the content of the image to be identified in step 11 of the above embodiment, and will not be repeated here.

[0049] Step 32: Perform moving target detection on the image to be recognized to obtain a mask image.

[0050] Moving target detection is performed on the image to be recognized to obtain a mask image. The mask image can be used to distinguish between the foreground region and the background region. The foreground region includes moving target detection points. Specifically, the image to be recognized can be input into the moving target detection model to obtain the mask image. The moving target detection model can be a model used for moving target detection in the field of target detection technology. The specific structure of the moving target detection model is not limited here.

[0051] like Figure 4 As shown, Figure 4 For an image to be identified in a specific application scenario, moving target detection can be performed on it to obtain results such as... Figure 5 The mask image shown is in which the white star points in the mask image are moving target detection points, the area formed by the white star points is the foreground area, and the black area is the background area.

[0052] Step 33: Based on the mask image, crop out the moving target region image from the image to be identified.

[0053] Based on the mask image, the moving target region image is cropped from the image to be identified; specifically, such as... Figure 6As shown, step 33 includes steps 61 to 63 below.

[0054] Step 61: Perform image processing on the mask image to obtain an enhanced mask image.

[0055] Image processing of a mask image yields an enhanced mask image. This processing allows for the measurement or extraction of features within the mask image. Simultaneously, denoising the mask image clusters moving target detection points in the foreground region, facilitating subsequent image analysis. Specifically, the enhanced mask image may include clustered moving target detection points, and image processing may include morphological processing such as dilation, erosion, opening, or closing operations. Figure 5 Taking the mask image shown as an example, by performing image processing on the mask image, we can obtain... Figure 7 The enhanced mask image shown.

[0056] Step 62: Extract the contours of the clustered moving target detection points to obtain target boxes that can be selected for moving targets.

[0057] Contour extraction is performed on the clustered moving target detection points to obtain target bounding boxes that can select the moving targets; still as described above... Figure 7 Taking the enhanced mask image shown as an example, contour extraction can be performed on it to obtain the following: Figure 8 The target bounding box is shown.

[0058] Step 63: Obtain the moving target region image from the region corresponding to the target bounding box cropped from the image to be recognized.

[0059] After obtaining the bounding box of the moving target, the region corresponding to the bounding box is cropped from the image to be identified, thereby obtaining the moving target region image. In a specific implementation, when there is no moving target in the image to be identified, at least one region image can be randomly extracted from the image to be identified as the moving target region image. The method of random extraction is not limited here.

[0060] Specifically, before step 63, target frames that meet at least one preset filtering condition can be filtered out. The preset filtering condition may include: the target frame is outside the preset detection area, or the target frame is a shadow area, or the size of the target frame does not meet the preset falling object size. The preset detection area may be a detection area preset according to application requirements or experience. For example, if falling object detection is to be performed on the corner of the pipe, the preset detection area can be defined as the corner area of ​​the pipe in the pipe monitoring screen. By judging whether the target frame appears in the preset detection area, it can be determined whether the moving target is the target to be detected. If the target frame is not in the preset detection area, it can be directly filtered out.

[0061] The shadow region can be an area misidentified as a foreground region by the moving object detection model. It actually belongs to neither the background nor the foreground region, meaning it does not contain the moving object to be detected and can be filtered out. The preset falling object size can be used to represent the size of the falling object (the target object for falling object detection). The falling object can be selected according to actual needs, such as a droplet or other falling objects; there is no limitation here. The preset falling object size can be set based on experience or actual conditions. It can determine whether the target box might contain a falling object based on the preset falling object size, thereby filtering out target boxes that obviously do not contain falling objects from a size perspective. In this embodiment, the size can include two dimensions: height and width. For example, the width range in the preset falling object size is within [w1, w2], and the height range is within [h1, h2]. If the height of the target box is not within [w1, w2] or the width of the target box is not within [h1, h2], it means that the size of the target box does not meet the preset falling object size, and the target box can be filtered out. Understandably, if all target boxes are filtered out using at least one of the above preset filtering conditions, the following falling object detection steps can be skipped, and the process can directly return to step 31.

[0062] Furthermore, after filtering out target boxes that meet at least one of the above preset filtering conditions, the size of the filtered target boxes can be adjusted to a preset size range. The preset size range can be set according to the actual situation. The preset size range may include the minimum input size and the maximum input size. The above scheme for adjusting the size of the filtered target boxes to the preset size range may include: in response to the target box size being smaller than the minimum input size, increasing the size of the target box so that the size of the increased target box is between the minimum input size and the maximum input size. This can avoid the situation where the target box size is too small, resulting in the cropped moving target area image being too small, thereby affecting the accuracy of falling object detection.

[0063] Taking the minimum width (wmin) and minimum height (hmin) within the preset size range as an example, if the width (w) of the filtered target box is less than the minimum width (wmin), it means the size of the filtered target box is less than the minimum input size. In this case, the left and right borders of the target box can be increased by (wmin-w) / 2 respectively to make the target box size reach the minimum input size. If the height (h) of the target box is less than the minimum height (hmin), the top and bottom borders of the target box can be increased by (hmin-h) / 2 respectively to make the target box size reach the minimum input size. Alternatively, the maximum width and maximum height within the preset size range can be used to calculate the upper limit of the target box's height or width increase. Then, the width of the target box can be adjusted within the range formed by (wmin-w) / 2 and the upper limit of the width increase, and the height can be adjusted within the range formed by (hmin-h) / 2 and the upper limit of the height increase. If the expanded target box's border exceeds the image boundary of the image to be recognized, pixels can be used to fill the area exceeding the image boundary.

[0064] like Figure 9 As shown, Figure 9 The image of the region corresponding to the target bounding box before adjustment, after resizing, can be obtained as follows: Figure 10 The image shows the region corresponding to the adjusted target bounding box.

[0065] Step 34: In response to the fact that the accuracy of the target classification module in classifying and recognizing the moving target region image does not meet the preset accuracy condition, the target classification module is used to classify the moving target region image.

[0066] In response to the target classification module not meeting the preset accuracy conditions, the target classification module can be used to classify the moving target region image first, so as to classify the moving target region image as a foreground image or a background image. Specifically, the preset accuracy conditions include: the detection accuracy of the target classification module in classifying and recognizing the moving target region image is greater than or equal to the preset accuracy threshold. For example, the accuracy of the target classification module can be detected by using a test set test method, or other accuracy test methods can also be used, which are not limited here.

[0067] Furthermore, the target classification module here can be a neural network model that only has the ability to recognize foreground and background images. It has not been trained on training images containing falling objects and does not have the ability to detect falling objects in foreground images. When the target classification module does not meet the preset accuracy conditions, the target classification module can first use its ability to classify images of moving target regions to preliminarily determine whether the moving target region image is a background image. In response to the moving target region image being a background image, the step of obtaining the image to be recognized is directly returned.

[0068] Step 35: In response to the moving target region image being the foreground image, track the moving target in the moving target region image to obtain the moving target trajectory information.

[0069] The steps described above for tracking moving targets in the moving target region image to obtain moving target trajectory information are the same as step 12 in the above embodiment, and will not be repeated here.

[0070] Step 36: Determine whether a falling object has occurred in the image to be identified based on the trajectory information of the moving target.

[0071] In response to the moving target region image being the foreground image, the moving target in the moving target region image is tracked to obtain the moving target trajectory information. Then, based on the moving target trajectory information, it is determined whether a falling object phenomenon has occurred in the image to be identified. The falling object phenomenon includes a change in the position of the moving target in a reference direction, which includes the direction of gravity. The falling object phenomenon may include a dripping phenomenon. It is understood that this embodiment only uses a dripping phenomenon as an example for illustration. In other embodiments, the falling object phenomenon may also be the falling of other objects besides liquids, which is not limited here.

[0072] Specifically, based on the trajectory information of the moving target, it can be determined whether the moving target in the image of the moving target region has a displacement in the direction of gravity; if a displacement has occurred in the direction of gravity, it is determined that a dripping phenomenon has occurred in the image to be identified; it can be understood that, in an ideal state, the dripping object is in free fall in a direction completely coinciding with the direction of gravity. In a non-ideal state, affected by the ambient wind, the dripping object may move in a direction deviating from the direction of gravity by a certain angle. However, overall, the dripping object has a displacement in the direction of gravity. Therefore, it can be determined whether the moving target has a displacement in the direction of gravity by using the trajectory information, and thus determine whether it is a dripping object; in other embodiments, the reference direction can also be a direction deviating from the direction of gravity by a certain angle, which is not limited here.

[0073] Step 37: In response to determining that a falling object has occurred based on the trajectory information of the moving target, train the target classification module using the image of the moving target region, and return to the step of obtaining the image to be identified.

[0074] In response to determining that a falling object has occurred based on the trajectory information of the moving target, the target classification module is trained using the image of the moving target area, and the process returns to the step of acquiring the image to be identified. Specifically, after determining that a falling object has occurred based on the trajectory information of the moving target, an alarm message can be generated and fed back to the user so that the user can take appropriate measures in a timely manner, thereby ensuring production safety.

[0075] Step 38: In response to the target classification module's accuracy in classifying and recognizing the moving target region image meeting the preset accuracy condition, the target classification module is used to determine whether a falling object phenomenon has occurred in the image to be identified.

[0076] Step 38 is the same as step 25 in the above embodiment, and will not be repeated here; it can be understood that after the target classification module determines that a falling object has occurred, alarm information can also be generated and fed back to the user so that the user can take corresponding measures in a timely manner, thereby ensuring production safety.

[0077] This embodiment can determine whether a falling object has occurred in the image to be identified by judging whether the moving target in the moving target region image has shifted in the direction of gravity when the target classification module does not meet the preset accuracy conditions, thereby improving the accuracy of falling object detection. Moreover, before cropping the moving target region image from the image to be identified, the target box can be filtered using preset filtering conditions to remove target boxes that are obviously unlikely to contain falling objects, thereby greatly saving subsequent detection time and improving detection efficiency. At the same time, before using the moving target trajectory information for falling object detection, the classification results of the moving target region image by the target classification module can be used to select whether to use trajectory information for falling object detection, which can improve detection accuracy and efficiency. In addition, adjusting the size of the target box obtained after filtering to a preset size range can prevent the moving target region image from being too small, which would lead to low recognition accuracy, thereby further ensuring the accuracy of falling object detection.

[0078] Please see Figure 11 , Figure 11 This is a schematic diagram of an embodiment of the falling object detection device provided in this application. The falling object detection device 110 includes a memory 111 and a processor 112 connected to each other. The memory 111 is used to store a computer program. When the computer program is executed by the processor 112, it is used to implement the falling object detection method in the above embodiment.

[0079] Please see Figure 12 , Figure 12 This is a schematic diagram of an embodiment of a computer-readable storage medium provided in this application. The computer-readable storage medium 120 is used to store a computer program 121. When the computer program 121 is executed by a processor, it is used to implement the falling object detection method in the above embodiment.

[0080] The computer-readable storage medium 120 can be any medium capable of storing program code, such as a server, USB flash drive, external hard drive, read-only memory (ROM), random access memory (RAM), magnetic disk, or optical disk.

[0081] In the several embodiments provided in this application, it should be understood that the disclosed methods and devices can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For instance, the division of modules or units is only a logical functional division, and there may be other division methods in actual implementation. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed.

[0082] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment, depending on actual needs.

[0083] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0084] If the technical solution of this application involves personal information, the product using this technical solution has clearly informed the user of the personal information processing rules and obtained the user's voluntary consent before processing the personal information. If the technical solution of this application involves sensitive personal information, the product using this technical solution has obtained the user's separate consent before processing the sensitive personal information, and also meets the requirement of "express consent". For example, at personal information collection devices such as cameras, clear and prominent signs are set up to inform users that they have entered the scope of personal information collection and that personal information will be collected. If an individual voluntarily enters the collection scope, it is deemed that they have agreed to the collection of their personal information; or on the personal information processing device, with clear signs / information informing users of the personal information processing rules, authorization is obtained from the individual through pop-up information or by asking the individual to upload their personal information; wherein, the personal information processing rules may include information such as the personal information processor, the purpose of personal information processing, the processing method, and the types of personal information processed.

[0085] The above description is merely an embodiment of this application and does not limit the patent scope of this application. Any equivalent structural or procedural transformations made using the content of this application's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of this application.

Claims

1. A method for detecting falling objects, characterized in that, include: Acquire images of the moving target region in the image to be identified; In response to the fact that the accuracy of the target classification module in classifying and recognizing the moving target region image does not meet the preset accuracy condition, the target classification module is used to classify the moving target region image. In response to the classification result indicating that the moving target region image is a foreground image, the moving target in the moving target region image is tracked to obtain the moving target trajectory information. Based on the moving target trajectory information, it is determined whether a falling object phenomenon has occurred in the image to be identified; wherein, the falling object phenomenon includes a change in the position of the moving target in the reference direction; In response to determining that a falling object has occurred based on the trajectory information of the moving target, the target classification module is trained using the image of the moving target region, and the process returns to the step of obtaining the image to be identified. In response to the target classification module's accuracy in classifying and recognizing the moving target region image meeting a preset accuracy condition, the target classification module is used to determine whether a falling object phenomenon has occurred in the image to be identified.

2. The method for detecting falling objects according to claim 1, characterized in that, The preset accuracy condition includes: the accuracy of the target classification module in classifying and recognizing the moving target region image is greater than or equal to the preset accuracy threshold.

3. The method for detecting falling objects according to claim 1, characterized in that, The reference direction includes the direction of gravity, and the falling object phenomenon includes the dripping phenomenon.

4. The method for detecting falling objects according to claim 1, characterized in that, The step of acquiring the image of the moving target region in the image to be identified includes: Moving target detection is performed on the image to be identified to obtain a mask image; wherein, the mask image is used to distinguish between the foreground region and the background region, and the foreground region includes moving target detection points; Based on the mask image, the moving target region image is cropped from the image to be identified.

5. The method for detecting falling objects according to claim 4, characterized in that, The step of cropping the moving target region image from the image to be identified based on the mask image includes: The mask image is processed to obtain an enhanced mask image; wherein the enhanced mask image includes clustered moving target detection points; Contour extraction is performed on the clustered moving target detection points to obtain a target bounding box that can select the moving target. The moving target region image is obtained from the region corresponding to the target bounding box cropped from the image to be identified.

6. The method for detecting falling objects according to claim 5, characterized in that, Before the step of cropping the region corresponding to the target bounding box from the image to be recognized, the method further includes: Filter out target boxes that meet at least one preset filtering condition, wherein the preset filtering condition includes: the target box is outside the preset detection area, or the target box is a shadow area, or the size of the target box does not meet the preset falling object size.

7. The method for detecting falling objects according to claim 6, characterized in that, Before the step of cropping the region corresponding to the target bounding box from the image to be recognized, the method further includes: Adjust the size of the target box obtained after filtering to the preset size range.

8. The method for detecting falling objects according to claim 7, characterized in that, The preset size range includes a minimum input size and a maximum input size. The step of adjusting the size of the filtered target box to the preset size range includes: In response to the target box size being smaller than the minimum input size, the target box size is increased so that the increased target box size is between the minimum input size and the maximum input size.

9. The method for detecting falling objects according to claim 1, characterized in that, The step of acquiring the image of the moving target region in the image to be identified further includes: When the moving target is not present in the image to be identified, at least one region image is randomly extracted from the image to be identified as the moving target region image.

10. A falling object detection device, characterized in that, The device includes an interconnected memory and a processor, wherein the memory is used to store a computer program, which, when executed by the processor, is used to implement the falling object detection method according to any one of claims 1-9.

11. A computer-readable storage medium for storing a computer program, characterized in that, When the computer program is executed by the processor, it is used to implement the falling object detection method according to any one of claims 1-9.

Images