Left-behind detection method, left-behind detection device, equipment and medium

Abstract

The embodiment of the present application provides a method for detecting items left behind, an apparatus for detecting items left behind, a device and a medium, and the method for detecting items left behind includes: determining the background type of the detected area; the background type includes a static background type and a dynamic background type; based on the background type of the detected area, determining a detection strategy for items left behind that matches the detected area, and based on the detection strategy for items left behind, performing item left behind detection on the detected area; when it is detected that there are items left behind in the detected area, performing category matching processing on the left behind items through an item category matching model, and performing item left behind alarm processing on the detected area based on the matching result. Through the embodiment of the present application, not only can the detection of items left behind in a dynamic background be realized, but also the specific category information of the left behind items can be identified, so that the result of the detection of items left behind is more accurate.

Description

Technical Field

[0001] The present application relates to the field of image processing, and in particular to a method for detecting leftover objects, a device, equipment and medium for detecting leftover objects. Background Art

[0002] At present, by performing item left behind detection on an area, it is possible to determine whether items have been left behind in the area and what items have been left behind. For example, by performing item left behind detection at the security checkpoint of a station, it is possible to detect whether a passenger undergoing security check has left behind any items they carry. In the prior art, a video can be shot of the detected area, and the shot video can be processed to determine whether items have been left behind in the detected area. However, the prior art only supports item left behind detection under static backgrounds, cannot adapt to dynamic backgrounds, and cannot determine the specific information of the left behind items, resulting in inaccurate item left behind detection results. Summary of the invention

[0003] The purpose of the embodiments of the present application is to provide a method for detecting left-behind items, a device, equipment and medium for detecting left-behind items, so as to solve the problem that the related art does not support left-behind item detection in a dynamic background and the left-behind item detection results are inaccurate.

[0004] In order to solve the above technical problems, the embodiments of the present application are implemented as follows: In a first aspect, an embodiment of the present application provides a method for detecting a left-behind object, comprising: Determine the background type of the detected area; the background type includes a static background type and a dynamic background type; Determine, based on the background type of the detected area, a left-behind detection strategy that matches the detected area, and perform left-behind detection on the detected area based on the left-behind detection strategy; When it is detected that there are items left behind in the detected area, the item category matching model is used to perform category matching processing on the left-behind items, and based on the matching result, an item left-behind alarm processing is performed for the detected area.

[0005] In a second aspect, an embodiment of the present application provides an object left behind detection device, comprising: A background type determination module, used to determine the background type of the detected area; the background type includes a static background type and a dynamic background type; A detection module, configured to determine a left-behind detection strategy that matches the detected area based on the background type of the detected area, and perform left-behind detection on the detected area based on the left-behind detection strategy; The category matching module is used to perform category matching processing on the left-behind items through the item category matching model when detecting that there are items left behind in the detected area, and to perform item left-behind alarm processing for the detected area based on the matching result.

[0006] In a third aspect, an embodiment of the present application provides an electronic device, comprising: a processor, a communication interface, a memory and a communication bus; wherein the processor, the communication interface and the memory communicate with each other through the bus; the memory is used to store computer programs; the processor is used to execute the programs stored in the memory to implement the method for detecting left-behind items as described in the first aspect.

[0007] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the method for detecting left-behind items as described in the first aspect is implemented.

[0008] In a fifth aspect, an embodiment of the present application provides a computer program product, including a computer program, which, when executed by a processor, implements the method for detecting left-behind items as described in the first aspect.

[0009] In the embodiment of the present application, first, the background type of the detected area is determined; the background type includes a static background type and a dynamic background type; then, based on the background type of the detected area, the detection strategy of the leftover items matching the detected area is determined, and based on the detection strategy of the leftover items, the detected area is subjected to the detection of the leftover items; finally, when the detected area is detected to have leftover items, the leftover items are subjected to category matching processing through the item category matching model, and based on the matching result, the leftover items are subjected to the detected area. It can be seen that through the embodiment of the present application, by determining the background type of the detected area, the leftover items detection strategy matching the detected area under the static background and the leftover items detection strategy matching the detected area under the dynamic background can be determined, and the leftover items detection is performed on the detected area using the corresponding leftover items detection strategy, and when the detected area is detected to have leftover items, the category of the leftover items is identified. In this way, not only the leftover items detection under the dynamic background can be realized, but also the specific category information of the leftover items can be identified, so that the result of the leftover items detection is more accurate. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0011] Figure 1 A schematic diagram of a method for detecting left-behind items provided in an embodiment of the present application; Figure 2 A schematic diagram of the network architecture of the target detection model provided in the embodiment of the present application; Figure 3 A schematic diagram of a process for detecting items left behind provided in an embodiment of the present application; Figure 4 A schematic diagram of a process for determining the background type of a detected area provided in an embodiment of the present application; Figure 5 A schematic diagram of a dynamic background object left behind detection process provided in an embodiment of the present application; Figure 6 A schematic diagram of a static background object left behind detection process provided in an embodiment of the present application; Figure 7 A schematic diagram of a process for matching the categories of items left behind provided in an embodiment of the present application; Figure 8 A schematic diagram of the module composition of the object left behind detection device provided in an embodiment of the present application; Fig. 9 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0012] The embodiments of the present application provide a method for detecting left-behind items, a device, equipment and medium for detecting left-behind items.

[0013] In order to enable those skilled in the art to better understand the technical solutions in this application, the technical solutions in the embodiments of this application will be clearly and completely described below in conjunction with the drawings in the embodiments of this application. Obviously, the described embodiments are only part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of this application.

[0014] Figure 1 A schematic diagram of a method for detecting a left-behind object provided in an embodiment of the present application is shown in FIG. Figure 1As shown, the embodiment of the present application provides a method for detecting items left behind, which can be executed by a server, which can be an independent server, a server cluster composed of multiple servers, or a cloud server for cloud computing processing. The method for detecting items left behind can specifically include the following steps: S102, determining the background type of the detected area; the background type includes a static background type and a dynamic background type.

[0015] In the embodiment of the present application, when performing object left behind detection on the detected area, it can be first determined whether the background type of the detected area is a static background type or a dynamic background type.

[0016] In one embodiment, determining the background type of the detected area includes: Acquire an image sequence of the detected area, determine a baseline image and a plurality of reference images in the image sequence, and determine each first differential image between the baseline image and each reference image; Performing guided filtering processing on each first differential image to obtain a filtered image, and performing high-frequency information extraction processing on each first differential image to obtain a high-frequency information image; For each first differential image, a filtered image corresponding to the first differential image and a high-frequency information image corresponding to the first differential image are fused to obtain a fused image; based on the multiple fused images, the background type of the detected area is determined.

[0017] In an embodiment of the present application, when determining the background type of the detected area, an image sequence of the detected area can be first obtained, a baseline image and multiple reference images can be determined in the image sequence, and then the first differential image between the baseline image and each reference image can be determined respectively. Then, for each first differential image, the first differential image is subjected to guided filtering to obtain a filtered image, and for each first differential image, the first differential image is subjected to high-frequency information extraction to obtain a high-frequency information image corresponding to the first differential image. Then, for each first differential image, the filtered image corresponding to the first differential image and the high-frequency information image corresponding to the first differential image are fused to obtain a fused image; finally, based on multiple fused images, the background type of the detected area is determined.

[0018] In an exemplary implementation, the detected area may be, for example, a station entrance, a ticket inspection hall, a shopping mall, a library, and the like. The detection of items left behind in the detected area may be a real-time detection of items left behind in the detected area, or an item left behind in the detected area. The real-time detection of items left behind in the detected area may be a detection of items left behind in the detected area at fixed time intervals. When performing item left behind detection in the detected area, an image sequence of the detected area may be first obtained, and the image sequence of the detected area may be part or all of the video in the surveillance video of the detected area. If it is a real-time detection of items left behind, a surveillance video of the detected area between the current moment and the moment with a fixed time interval from the current moment may be obtained. After obtaining the surveillance video of the detected area, the first frame of the video may be selected in the surveillance video as a reference image, and then from a certain moment of the surveillance video, for example, starting from the 60th second, a video frame may be extracted every 5 seconds, and a total of 10 video frames may be extracted, and the extracted 10 video frames may be used as reference images. Then, the first differential image between the benchmark image and each reference image is determined respectively, and 10 first differential images are obtained. For each first differential image, two processes are performed, one of which is to output the filtered image through guided filtering, and the other is to output the high-frequency information image corresponding to the differential image after Fourier transform, high-pass filtering and Fourier inverse transform. Outputting the filtered image through guided filtering is to perform guided filtering processing on the first differential image to obtain the filtered image. Outputting the high-frequency information image corresponding to the differential image after Fourier transform, high-pass filtering and Fourier inverse transform is to convert the first differential image to the frequency domain through Fourier transform, apply a high-pass filter to remove the low-frequency components, retain the high-frequency components (edges and details), and then convert the first differential image back to the spatial domain through Fourier inverse transform to obtain the high-frequency information image corresponding to the first differential image. Then, the filtered image and the high-frequency information image corresponding to each first differential image are fused to obtain the corresponding fused image. The fusion process can be a grayscale image superposition process, and accordingly, each first differential image between the benchmark image and each reference image is a grayscale image, and the filtered image and the high-frequency information image corresponding to the first differential image are also grayscale images. Finally, the background type of the detected area is determined according to the fused images corresponding to the first differential images.

[0019] In one implementation, the fused image can also be subjected to threshold processing and dilation and erosion. Threshold processing, also known as image binarization, is one of the simplest methods for image segmentation. Threshold processing is to divide the pixel values ​​in the image into two categories according to the set threshold value, thereby generating a binary image. Dilation and erosion are commonly used image processing operations, mainly used to process binary images. Dilation increases the boundaries of objects in the image, expands the object area, and helps to connect the holes between objects or increase the size of the objects. Erosion reduces the boundaries of objects in the image, which helps to remove small details or disconnect small channels between objects.

[0020] In one embodiment, determining the background type of the detected area based on multiple fused images includes: Determining a reference fusion image from the multiple fusion images, performing difference processing on the reference fusion image and each of the multiple fusion images except the reference fusion image, to obtain a plurality of second difference images; Determine the magnitude relationship between an average quantity value of white pixels in a plurality of second differential images and a first preset quantity value threshold; If the judgment result is that the average quantity value of the white pixels in the plurality of second differential images is less than the first preset quantity value threshold, it is determined that the background type of the detected area is a static background type; Otherwise, it is determined that the background type of the detected area is a dynamic background type.

[0021] In the embodiment of the present application, when determining the background type of the detected area based on multiple fused images, a reference fused image can be first determined in the multiple fused images, and then the second differential images between the reference fused image and each of the other fused images in the multiple fused images except the reference fused image can be respectively determined. Then, the number of white pixels in each second differential image is counted, and the average number of white pixels in the multiple second differential images is calculated, and the size relationship between the average number of white pixels in the multiple second differential images and the first preset number value threshold is determined. If the judgment result is that the average number of white pixels in the multiple second differential images is less than the first preset number value threshold, the background type of the detected area is determined to be a static background type; if the judgment result is that the average number of white pixels in the multiple second differential images is greater than or equal to the first preset number value threshold, the background type of the detected area is determined to be a dynamic background type.

[0022] Continuing with the above example, a reference fusion image is determined from the 10 fusion images determined above, and the reference fusion image is subjected to differential processing with each of the other fusion images to obtain 9 second differential images. The number of white pixels in each second differential image is counted, the average number of white pixels in the 9 second differential images is calculated, and the size relationship between the average number of white pixels in the 9 second differential images and the first preset number threshold is determined. Assuming that the image size is 1920*1080, the first preset number threshold is 500, and the average number of white pixels in the 9 second differential images is 650, it can be seen that the average number of white pixels in the 9 second differential images is greater than the first preset number threshold, then it can be determined that the background type of the detected area is a static background type. Assuming that the average number of white pixels in the 9 second differential images is 400, which is less than the first preset number threshold, then it can be determined that the background type of the detected area is a dynamic background type.

[0023] S104, determining a left-behind detection strategy that matches the detected area based on the background type of the detected area, and performing left-behind detection on the detected area based on the left-behind detection strategy.

[0024] In an embodiment of the present application, after determining the background type of the detected area, a left-behind object detection strategy that matches the detected area can be determined based on the background type of the detected area, and based on the left-behind object detection strategy, the detected area can be subjected to left-behind object detection.

[0025] In one embodiment, based on the background type of the detected area, determining a left-behind detection strategy that matches the detected area, and performing left-behind detection on the detected area based on the left-behind detection strategy includes: In the case where it is determined that the background type of the detected area is a static background type, any first image in the image sequence is acquired, a third differential image between the first image and the reference image is determined, and based on the contour information of the target object in the third differential image, it is determined whether an object is left behind in the detected area; the first image contains the target object but does not contain a target object corresponding to the target object; When it is determined that the background type of the detected area is a dynamic background type, any second image in the image sequence is obtained, and it is determined whether the distance between the target item and the target object corresponding to the target item in the second image is greater than a preset distance threshold. Based on the determination result, it is determined whether an item is left behind in the detected area.

[0026] In the embodiment of the present application, when the background type of the detected area is determined to be a static background type, the object left behind detection strategy matching the detected area can be determined to be a static background object left behind detection strategy, and the static background object left behind detection strategy specifically includes: first, obtaining any first image in the image sequence, the first image is any image in the image sequence that contains the target object and does not contain the target object corresponding to the target object. Then, determine the third differential image between the first image and the reference image, and determine whether an object is left behind in the detected area based on the contour information of the target object in the third differential image.

[0027] In the case where the background type of the detected area is determined to be a dynamic background type, the left-behind item detection strategy matching the detected area can be determined to be a dynamic background left-behind item detection strategy, and the dynamic background left-behind item detection strategy specifically includes: first, obtaining any image in the image sequence as the second image, and determining whether the distance between the target item and the target object corresponding to the target item in the second image is greater than a preset distance threshold. Then, based on the determination result, determining whether there is left-behind item in the detected area.

[0028] In one embodiment, determining whether an object is left behind in the detected area based on the contour information of the target object in the third differential image includes: determining whether the contour area of ​​the target object in the third differential image is within a preset contour area range; When it is determined that the contour area of ​​the target object in the third differential image is within a preset contour area range, the first image is determined as a first target image, and it is determined whether the image sequence includes a plurality of first target images whose quantity values ​​exceed a second preset quantity value threshold and are continuous in time; When it is determined that the image sequence contains a plurality of first target images whose quantity values ​​exceed a second preset quantity value threshold and are continuous in time, it is determined that an object is left behind in the detected area, and the target object is determined to be a left-behind object.

[0029] In the embodiment of the present application, when determining whether an item is left behind in the detected area, first, it can be determined whether the contour area of ​​the target item in the third differential image is within the preset contour area range. If it is determined that the contour area of ​​the target item in the third differential image is within the preset contour area range, the first image can be determined as the first target image. Then, it is determined whether the image sequence contains multiple first target images whose quantity values ​​exceed the second preset quantity value threshold and are continuous in time. If it is determined that the image sequence contains multiple first target images whose quantity values ​​exceed the second preset quantity value threshold and are continuous in time, it can be determined that an item is left behind in the detected area, and the target item can be determined as a left-behind item.

[0030] In one embodiment, determining whether an item is left behind in the detected area based on the determination result includes: In the case where it is determined that the distance between the target item and the target object corresponding to the target item in the second image is greater than a preset distance threshold, determining the second image as a second target image, and determining whether the image sequence includes a plurality of second target images whose quantity value exceeds a second preset quantity value threshold and are continuous in time; When it is determined that the image sequence contains a plurality of second target images whose quantity values ​​exceed the second preset quantity value threshold and are continuous in time, it is determined that an object is left behind in the detected area, and the target object is determined as a left-behind object.

[0031] In an embodiment of the present application, if it is determined that the distance between the target item in the second image and the target object corresponding to the target item is greater than a preset distance threshold, the second image can be determined as a second target image, and then it is determined whether the image sequence contains multiple second target images whose quantity values ​​exceed the second preset quantity threshold and are continuous in time. If it is determined that the image sequence contains multiple second target images whose quantity values ​​exceed the second preset quantity threshold and are continuous in time, then it can be determined that an item has been left behind in the detected area, and the target item can be determined as a left-behind item.

[0032] In one embodiment, before acquiring any second image in the image sequence, the method further includes: Through the target detection model, the target object detection process is performed on the detected area, and it is determined whether the detection result contains the target object; If the detection result contains the target object, the target object is bound to the corresponding target object using the identification information; Otherwise, for each image in the image sequence, the moving object in the detected area and the area in the image where the moving object is located are identified, and the area is occluded; based on the image sequence after occlusion processing, the detected area is detected for objects left behind.

[0033] In an embodiment of the present application, before acquiring any second image in the image sequence, the target object detection processing can also be performed on the detected area through the target detection model, and it can be determined whether the target object is included in the detection result. The target object can be an object that may be left behind. If the target object is included in the detection result, the target object can be bound to the corresponding target object by using the identification information through a multi-target tracking algorithm. If the target object is not included in the detection result, for each image in the image sequence, the moving object in the detected area and the area in the image where the moving object is located are identified, and the area where the moving object is located is occluded. Based on the image sequence after the occlusion processing, the detected area is detected for left-behind items.

[0034] In an exemplary implementation, the target object is a person carrying a target object. When the background type of the detected area is a static background type, when the person and the object appear in the picture at the same time, the difference image between the image containing the person and the object and the above-mentioned reference image will contain more person foreground information. When the person and the object are not in the same picture, the difference image between the image containing only the object but not the person carrying the object and the reference image will only retain the contour information of the object. Based on this, it can be determined whether the area of ​​the object contour is within the preset contour area range. Assuming that the size of each image in the image sequence is 1920*1080, the preset contour area range is 500Pixel² (pixel square)-10000Pixel², when the area of ​​the object contour is greater than 500Pixel² and less than 10000Pixel², the area of ​​the object contour is within the preset contour area range, then it can be further determined whether the number of images in the image sequence that only contain the object but not the person carrying the object exceeds the second preset number value threshold and multiple images are continuous in time, that is, whether the state that the person and the object are not in the same picture has lasted for a period of time. Assume that the frame rate of the surveillance video of the detected area is 24fps, the value of the second preset number value threshold is 480, the number of images containing only the object but not the person carrying the object is 600, and the 600 images are continuous in time, then it is determined that the separation of the person and the object has lasted for a period of time, so it can be determined that the object is left behind. While judging whether the area of ​​the object contour is within the preset contour area range, it can also be judged whether the size of the object contour is within the preset contour size range. The contour size can include the length and width of the contour. The preset contour size range can be, for example, 20Pixel (pixel)-200Pixel, which means that the number of pixels in the horizontal direction is between 20-200, and the number of pixels in the vertical direction is between 20-200. When the length and width of the object contour are respectively greater than 20Pixel and less than 200Pixel, the size of the object contour is within the preset contour size range.

[0035] By judging the relationship between the contour area of ​​the object in the differential image and the preset contour area range, and the relationship between the size of the object's contour and the preset contour size range, it can be determined whether the object is an object that may be left behind. For example, backpacks, school bags, handbags / bags / boxes, luggage bags / boxes, wallets, umbrellas, mobile phones, cups, bottles, bowls, buckets, books, laptops, monitors, cartons, etc. are all objects that pedestrians may carry with them and are easy to leave behind. Then, by judging the relationship between the number value of the image that only contains the object but not the person carrying the object and the second preset number value threshold, or judging the duration of the state that the person and the object are not in the same picture, it can be determined that the person carrying the object is separated from the object for a long time, thereby further determining that the object is left behind.

[0036] When the background type of the detected area is a dynamic background type, the target object detection model can be used to detect the target object in the detected area. The target object can be the above-mentioned items that may be left behind. Determine whether the detection result contains the above-mentioned items that may be left behind. If the detection result contains the above-mentioned items that may be left behind, then track the items and the person carrying the items through a multi-target tracking algorithm, and use the tracking ID number to bind the person to the item. The basic principle of binding people and items is one person to multiple objects. If the average distance between an object and a person in multiple consecutive images is the smallest, the same ID is used to bind the object to the person. If the detection result does not include the above-mentioned items that may be left behind, then it can be determined whether the detection result includes moving objects such as bicycles, motorcycles, cars, cats, dogs, etc. If the detection result includes the above-mentioned moving objects, then the area in the image where the moving object is located can be identified in each image in the image sequence, and the area can be blocked to exclude the moving object in the detected area, which is equivalent to converting the background type of the detected area from a dynamic background type to a static background type, and then using the static background object left detection to perform item left detection on the detected area. It is also possible to first mark the target frame of the moving object in each image in the image sequence, and then use the static background object left detection to perform item left detection on the detected area. During the detection, the target frame area of ​​the moving object in the differential image between each image and the reference image can be blocked to exclude the interference of the moving object on the item left detection.

[0037] After determining that the detection results include the above-mentioned items that may be left behind, and binding the items with the corresponding people carrying the items, you can first obtain any image in the image sequence as the second image, or you can use the image at the current moment as the second image to determine whether the distance between the item and the person carrying the item in the second image is greater than the preset distance threshold. There are two situations at this time. The first is that the item and the person carrying the item are in the same picture, and the second is that the item and the person carrying the item are not in the same picture. When the item and the person carrying the item are not in the same picture, set the distance between the item and the person carrying the item to 1920, in pixels. Assuming that the preset distance threshold is 480 pixels and the distance between the object and the person carrying the object is 1200 pixels, then the second image can be determined as the second target image. Next, determine whether the image sequence contains multiple second target images whose quantity value exceeds the second preset threshold and are continuous in time, that is, whether the state in which the distance between the object and the person carrying the object is greater than the preset distance threshold has continued for a period of time. If it is determined that the quantity value of the second target images in the image sequence is 600, which exceeds the second preset quantity value threshold of 480, and multiple second target images are continuous in time, it means that the object and the person carrying the object have been separated for a long time, then it can be determined that the object has been left behind.

[0038] In one implementation, the target detection model is an improved yolov8s network. Since targets such as mobile phones, bottles, bowls, and headphones are extremely small (possibly less than 10*10 pixels) at medium and high positions, the original yolov8s network is prone to missed detection. Assuming that the size of the input image is 640×640, the detection feature map size corresponding to P3 / 8-small (detection layer) is 80×80, which is used to detect small targets larger than 8×8. The detection feature map size corresponding to P4 / 16-medium (detection layer) is 40×40, which is used to detect medium targets larger than 16×16. The detection feature map size corresponding to P5 / 32-large (detection layer) is 20×20, which is used to detect large targets larger than 32×32. The original yolov8s network may have poor detection capabilities for small targets.

[0039] Figure 2 A schematic diagram of the network architecture of the target detection model provided in the embodiment of the present application is shown in FIG. Figure 2 As shown in the figure, by adding a 160×160 detection feature map to the original yolov8 network ( Figure 2The upsample and concat marked in dark gray in the figure are used to detect small targets larger than 4×4. The original yolov8 network output layer includes three output layers. In order to improve the detection ability of small targets, a fourth output layer is added on the basis of the original three output layers ( Figure 2 In addition, the original yolov8 network uses the SPPF module for multi-scale feature map fusion, which is fast but has poor detection effect on very small targets. The original yolov7 multi-scale fusion module SPPCSPC has good detection effect on very small targets, but its speed is slow. Based on this, a Fast-SPPCSPC module is added to the original yolov8 network. The Fast-SPPCSPC module can be obtained by replacing the 3*3 convolution layer with 1024 output channels in the original yolov7 multi-scale fusion module SPPCSPC with 1*1 convolution and 3*3 convolution. The channel transformation is realized by 1*1 convolution, so that the channels processed by 3*3 convolution are fewer, so that the efficiency of target detection can be improved to a certain extent. By adding a 160×160 detection feature map, the fourth output layer and the Fast-SPPCSPC module to the original yolov8 network, a target detection model is obtained. This target detection model can effectively detect tiny targets and improve the efficiency of target detection to a certain extent.

[0040] S106, when it is detected that there are items left behind in the detected area, the item category matching model is used to perform category matching processing on the left-behind items, and based on the matching result, an item left-behind alarm processing is performed for the detected area.

[0041] In an embodiment of the present application, when it is detected that there are items left behind in the detected area, the left-behind items are subjected to category matching processing through the item category matching model to obtain the left-behind item category matching results, and based on the left-behind item category matching results, the left-behind item alarm processing is performed for the detected area.

[0042] In one implementation, the item category matching model can be a clip vit-B / 32 model. The clip model is an English image-text matching model. In order to allow the clip model to match the Chinese category, a large number of commonly used pictures of objects of various categories can be selected during initialization, such as school bags, backpacks, mobile phones, wallets, notebooks, etc. First, a picture-text mapping dictionary between Chinese categories and pictures is established, and then multiple pictures of objects of various categories are encoded through the clip model to generate a feature file corresponding to each picture, and the feature file is cached locally. After detecting that there are items left in the detected area and completing the location of the left items, the left items are cropped from the image where they are located. When cropping, the cropping frame can be appropriately enlarged, for example, by increasing the width by 0.1 times on the left and right, and by 0.1 times on the top and bottom. Since the clip model input picture size requirement is 224*224, the size of the cropped left items should be restored to 224*224 here. After the picture of the abandoned item is input into the clip model for encoding processing, the feature vector of the abandoned item is matched with the feature file cached locally. If there is a feature file whose cosine similarity exceeds the preset cosine similarity threshold, it is considered that the category of the abandoned item is matched successfully. The preset cosine similarity threshold is, for example, 0.4. After the match is successful, an alarm containing the coordinates and category information of the item can be sent directly to the item abandoned detection platform. If the match fails, in addition to sending an alarm containing the coordinate information of the item and the unknown category to the platform, the picture of the abandoned item can also be automatically recovered. After the picture of the abandoned item is manually labeled, the image-text mapping dictionary and feature file are updated, and the same category of abandoned items can be matched successfully when they appear in the future.

[0043] In the embodiment of the present application, first, the background type of the detected area is determined; the background type includes a static background type and a dynamic background type; then, based on the background type of the detected area, the detection strategy of the leftover items matching the detected area is determined, and based on the detection strategy of the leftover items, the detected area is subjected to the detection of the leftover items; finally, when the detected area is detected to have leftover items, the leftover items are subjected to category matching processing through the item category matching model, and based on the matching result, the leftover items are subjected to the detected area. It can be seen that through the embodiment of the present application, by determining the background type of the detected area, the leftover items detection strategy matching the detected area under the static background and the leftover items detection strategy matching the detected area under the dynamic background can be determined, and the leftover items detection is performed on the detected area using the corresponding leftover items detection strategy, and when the detected area is detected to have leftover items, the category of the leftover items is identified. In this way, not only the leftover items detection under the dynamic background can be realized, but also the specific category information of the leftover items can be identified, so that the result of the leftover items detection is more accurate.

[0044] Figure 3 A schematic diagram of the object left behind detection process provided in the embodiment of the present application is shown in FIG. Figure 3 As shown, in an exemplary implementation, the object left behind detection process specifically includes the following steps: S301: Acquire a surveillance video of the area to be detected, and determine a baseline image and multiple reference images in the surveillance video.

[0045] S302: Determine whether the background type of the detected area is a static background type based on the benchmark image and multiple reference images.

[0046] If the background type of the detected area is a static background type, execute S303; if the background type of the detected area is a dynamic background type, execute S308.

[0047] S303: Using the static background object left behind detection strategy, detect whether there is any object left behind in the detected area.

[0048] If an object is left behind in the detected area, execute S304; otherwise, no processing is performed.

[0049] S304: Identify the position of the abandoned object in the image and crop the abandoned object in the image, and input the cropped image of the abandoned object into the category matching model.

[0050] S305: Determine whether the category matching model has identified the category of the left-behind item.

[0051] If the category of the left-behind item is identified, execute S306; otherwise, execute S307.

[0052] S306: reporting the coordinate information and category information of the left-behind item to the left-behind item detection platform, and issuing a left-behind item alarm.

[0053] S307: Send an alarm containing the coordinate information of the object and the unknown category to the platform, manually mark the category of the leftover object, and update the category matching model.

[0054] S308: Using the dynamic background object left behind detection strategy, detect whether an object is left behind in the detected area.

[0055] If an object is left behind in the detected area, execute S305; otherwise, no processing is performed.

[0056] Figure 4 A schematic diagram of the process of determining the background type of the detected area provided in an embodiment of the present application, such as Figure 4 As shown, in an exemplary implementation, determining the background type of the detected area specifically includes the following steps: S401: Acquire a surveillance video of the area to be detected, and determine a baseline image and multiple reference images in the surveillance video.

[0057] S402: Determine a difference image between the benchmark image and each reference image.

[0058] S403: performing guided filtering and Fourier transform processing on each differential image respectively, to obtain a filtered image and a Fourier transform processed image corresponding to each differential image.

[0059] S404: performing grayscale image fusion processing on the filtered images corresponding to each differential image and the images processed by Fourier transform corresponding to each differential image to obtain fused images corresponding to each differential image.

[0060] S405: Perform threshold processing and dilation erosion on each fused image to obtain a plurality of differential binary images.

[0061] S406: Determine a reference image from the plurality of differential binary images, and determine differential images between the reference image and each of the other differential binary images.

[0062] S407: Calculate the average value of white pixels in the difference image between the reference image and other difference binary images.

[0063] S408: Determine whether the average value is lower than a preset quantity value threshold.

[0064] If yes, execute S409; otherwise, execute S410.

[0065] S409: Determine that the background type of the detected area is a static background type.

[0066] S410: Determine that the background type of the detected area is a dynamic background type.

[0067] Figure 5 A schematic diagram of a dynamic background object left behind detection process provided in an embodiment of the present application is shown in FIG. Figure 5 As shown, in an exemplary implementation, the dynamic background object left behind detection process specifically includes the following steps: S501: Acquire surveillance video of the detected area.

[0068] S502: Input the surveillance video of the detected area into the target detection model, and obtain the detection result through the target detection model.

[0069] S503: Determine whether the detection result contains any items that may be left behind.

[0070] If included, execute S504; otherwise, execute S509.

[0071] S504: Binding the object and the person carrying the object through a multi-target tracking algorithm.

[0072] S505: Obtain any one image in the image sequence, and determine whether the distance between the object in the image and the person bound to the object exceeds a preset distance threshold.

[0073] If yes, execute S506; otherwise, execute S508.

[0074] S506: Determine the image as a target image, and determine whether the number of target images in the image sequence exceeds a preset number threshold and the plurality of target images are continuous in time.

[0075] If yes, execute S507; otherwise, execute S508.

[0076] S507: Determine whether an item is left behind in the detected area.

[0077] S508: Determine that no items are left in the detected area.

[0078] S509: Determine whether the detection result contains a moving object.

[0079] If included, execute S510; otherwise, do nothing.

[0080] S510: Marking the area where the moving object is located in the image, performing occlusion processing on the area where the moving object is located, converting the dynamic background into a static background, and performing object left behind detection on the detected area using the static background object left behind detection process.

[0081] Figure 6 A schematic diagram of a static background object left behind detection process provided in an embodiment of the present application is shown in FIG. Figure 6 As shown, in an exemplary implementation, the static background object left behind detection process specifically includes the following steps: S601: Acquire any image in the surveillance video of the detected area, and determine a reference image in the surveillance video.

[0082] S602: Determine a differential image between the image and a reference image.

[0083] S603: Determine whether the outline size of the object in the differential image is within a preset outline size range, and whether the outline area of ​​the object is within a preset outline area range.

[0084] If yes, execute S604; otherwise, execute S606.

[0085] S604: Determine the image as a target image, and determine whether the number of target images in the image sequence exceeds a preset number threshold and the plurality of target images are continuous in time.

[0086] If yes, execute S605; otherwise, execute S606.

[0087] S605: Determine whether an object is left behind in the detected area.

[0088] S606: Determine that no items are left in the detected area.

[0089] Figure 7 A schematic diagram of the process of matching the categories of the leftover items provided in the embodiment of the present application is shown in FIG. Figure 7 As shown, in an exemplary implementation, the legacy item category matching process specifically includes the following steps: S701: Determine the location information of the left-behind object in the image, and crop to obtain a picture of the left-behind object.

[0090] S702: Input the image of the abandoned item into the clip model for category matching processing to obtain a category matching result.

[0091] S703: Determine whether the category matching result contains category information of the left-behind items.

[0092] If included, execute S704; otherwise, execute S705.

[0093] S704: Report the coordinate information and category information of the left-behind item to the left-behind item detection platform, and issue a left-behind item alarm.

[0094] S705: Send an alarm containing the coordinate information of the object and the unknown category to the platform.

[0095] S706: Manually label the category of the abandoned item and update the category matching model.

[0096] It should be noted that the above Figure 3-Figure 7 For the specific implementation of each step in Figure 1 The specific implementation methods in the present application will not be described in detail here.

[0097] Corresponding to the object left behind detection method provided in the above embodiment, based on the same technical concept, the embodiment of the present application also provides an object left behind detection device, Figure 8 A schematic diagram of the module composition of the object left behind detection device provided in an embodiment of the present application is shown in FIG. Figure 8 As shown, the device for detecting left-behind items comprises: Background type determination module 801, determines the background type of the detected area; the background type includes static background type and dynamic background type; A detection module 802 is used to determine a left-behind detection strategy that matches the detected area based on the background type of the detected area, and perform left-behind detection on the detected area based on the left-behind detection strategy; The category matching module 803 is used to perform category matching processing on the left-behind items through the item category matching model when it is detected that there are items left behind in the detected area, and to perform item left-behind alarm processing for the detected area based on the matching result.

[0098] In a possible implementation, the background type determination module 801 is specifically configured to: Acquire an image sequence of the detected area, determine a baseline image and a plurality of reference images in the image sequence, and determine each first differential image between the baseline image and each of the reference images; Performing guided filtering processing on each of the first differential images to obtain a filtered image, and performing high-frequency information extraction processing on each of the first differential images to obtain a high-frequency information image; For each of the first differential images, a filtered image corresponding to the first differential image and a high-frequency information image corresponding to the first differential image are fused to obtain a fused image; based on the multiple fused images, the background type of the detected area is determined.

[0099] In a possible implementation, the background type determination module 801 is further specifically configured to: Determining a reference fused image from the plurality of fused images, performing difference processing on the reference fused image and each of the other fused images from the plurality of fused images except the reference fused image, to obtain a plurality of second difference images; Determine the magnitude relationship between an average quantity value of white pixels in a plurality of the second differential images and a first preset quantity value threshold; If the judgment result is that the average quantity value of the white pixels in the plurality of the second differential images is less than the first preset quantity value threshold, it is determined that the background type of the detected area is a static background type; Otherwise, it is determined that the background type of the detected area is a dynamic background type.

[0100] In a possible implementation, the detection module 802 is specifically configured to: In the case where it is determined that the background type of the detected area is a static background type, obtaining any first image in the image sequence, determining a third differential image between the first image and the reference image, and determining whether an object is left behind in the detected area based on contour information of the target object in the third differential image; the first image includes the target object but does not include a target object corresponding to the target object; When it is determined that the background type of the detected area is a dynamic background type, any second image in the image sequence is acquired to determine whether the distance between the target item and the target object corresponding to the target item in the second image is greater than a preset distance threshold, and based on the determination result, determine whether an item is left behind in the detected area.

[0101] In a possible implementation, the detection module 802 is further specifically configured to: Determining whether the contour area of ​​the target object in the third differential image is within a preset contour area range; In the case where it is determined that the contour area of ​​the target object in the third differential image is within a preset contour area range, determining the first image as a first target image, and determining whether the image sequence includes a plurality of first target images whose quantity values ​​exceed a second preset quantity value threshold and are continuous in time; When it is determined that the image sequence contains a plurality of first target images whose quantity values ​​exceed the second preset quantity value threshold and are continuous in time, it is determined that an object is left behind in the detected area, and the target object is determined to be a left-behind object.

[0102] In a possible implementation, the detection module 802 is further specifically configured to: In a case where it is determined that the distance between the target item and the target object corresponding to the target item in the second image is greater than a preset distance threshold, determining the second image as a second target image, and determining whether the image sequence includes a plurality of second target images whose quantity value exceeds the second preset quantity value threshold and are continuous in time; When it is determined that the image sequence contains a plurality of second target images whose quantity values ​​exceed the second preset quantity value threshold and are continuous in time, it is determined that an object is left behind in the detected area, and the target object is determined to be the left-behind object.

[0103] In a possible implementation, the detection module 302 is further specifically configured to: Using the target detection model, the detected area is subjected to target object detection processing, and it is determined whether the detection result contains the target object; If the detection result includes a target object, the target object is bound to the corresponding target object using identification information; Otherwise, for each image in the image sequence, identify the moving object in the detected area and the area in the image where the moving object is located, and perform occlusion processing on the area; based on the image sequence after occlusion processing, perform object abandonment detection on the detected area.

[0104] In the embodiment of the present application, first, the background type of the detected area is determined; the background type includes a static background type and a dynamic background type; then, based on the background type of the detected area, the detection strategy for items left behind that matches the detected area is determined; then, based on the detection strategy for items left behind, the detected area is subjected to item left behind detection; finally, when it is detected that there are items left behind in the detected area, the item category matching model is used to perform category matching processing on the left behind items, and based on the matching result, the detected area is subjected to item left behind alarm processing. It can be seen that through the embodiment of the present application, by determining the background type of the detected area, the detection strategy for items left behind that matches the detected area under the static background and the detection strategy for items left behind that matches the detected area under the dynamic background can be determined, and the corresponding detection strategy for items left behind can be used to perform item left behind detection on the detected area, and when it is detected that there are items left behind in the detected area, the category of the left behind items can be identified. In this way, not only can the detection of items left behind under the dynamic background be realized, but also the specific category information of the left behind items can be identified, so that the result of the detection of items left behind is more accurate.

[0105] The device for detecting left-behind items provided in the embodiments of the present application can implement each process in the embodiments corresponding to the above-mentioned method for detecting left-behind items, and will not be described again here to avoid repetition.

[0106] It should be noted that the object left behind detection device provided in the embodiment of the present application and the object left behind detection method provided in the embodiment of the present application are based on the same inventive concept, so the specific implementation of this embodiment can refer to the implementation of the aforementioned object left behind detection method, and the repeated parts will not be repeated.

[0107] Corresponding to the object left behind detection method provided in the above embodiment, based on the same technical concept, the embodiment of the present application further provides an electronic device, which is used to execute the above object left behind detection method. Fig. 9 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application, such as Fig. 9As shown, the electronic device may have relatively large differences due to different configurations or performances, and may include one or more processors 901 and memory 902, and the memory 902 may store one or more storage applications or data. Among them, the memory 902 can be a temporary storage or a permanent storage. The application stored in the memory 902 may include one or more modules (not shown in the figure), and each module may include a series of computer executable instructions in the electronic device. Furthermore, the processor 901 can be configured to communicate with the memory 902 to execute a series of computer executable instructions in the memory 902 on the electronic device. The electronic device may also include one or more power supplies 903, one or more wired or wireless network interfaces 904, one or more input and output interfaces 905, and one or more keyboards 906.

[0108] In a specific embodiment, the electronic device includes a processor, a communication interface, a memory and a communication bus; wherein the processor, the communication interface and the memory communicate with each other through the bus; the memory is used to store a computer program; the processor is used to execute the program stored in the memory to implement the following method steps: Determine the background type of the detected area; the background type includes a static background type and a dynamic background type; Determine, based on the background type of the detected area, a left-behind detection strategy that matches the detected area, and perform left-behind detection on the detected area based on the left-behind detection strategy; When it is detected that there are items left behind in the detected area, the item category matching model is used to perform category matching processing on the left-behind items, and based on the matching result, an item left-behind alarm processing is performed for the detected area.

[0109] In the embodiment of the present application, first, the background type of the detected area is determined; the background type includes a static background type and a dynamic background type; then, based on the background type of the detected area, the detection strategy of the leftover items matching the detected area is determined, and based on the detection strategy of the leftover items, the detected area is subjected to the detection of the leftover items; finally, when the detected area is detected to have leftover items, the leftover items are subjected to category matching processing through the item category matching model, and based on the matching result, the leftover items are subjected to the detected area. It can be seen that through the embodiment of the present application, by determining the background type of the detected area, the leftover items detection strategy matching the detected area under the static background and the leftover items detection strategy matching the detected area under the dynamic background can be determined, and the leftover items detection is performed on the detected area using the corresponding leftover items detection strategy, and when the detected area is detected to have leftover items, the category of the leftover items is identified. In this way, not only the leftover items detection under the dynamic background can be realized, but also the specific category information of the leftover items can be identified, so that the result of the leftover items detection is more accurate.

[0110] The present application also provides a computer-readable storage medium, wherein the storage medium stores a computer program, and when the computer program is executed by a processor, the following method steps are implemented: Determine the background type of the detected area; the background type includes a static background type and a dynamic background type; Determine, based on the background type of the detected area, a left-behind detection strategy that matches the detected area, and perform left-behind detection on the detected area based on the left-behind detection strategy; When it is detected that there are items left behind in the detected area, the item category matching model is used to perform category matching processing on the left-behind items, and based on the matching result, an item left-behind alarm processing is performed for the detected area.

[0111] In the embodiment of the present application, first, the background type of the detected area is determined; the background type includes a static background type and a dynamic background type; then, based on the background type of the detected area, the detection strategy of the leftover items matching the detected area is determined, and based on the detection strategy of the leftover items, the detected area is subjected to the detection of the leftover items; finally, when the detected area is detected to have leftover items, the leftover items are subjected to category matching processing through the item category matching model, and based on the matching result, the leftover items are subjected to the detected area. It can be seen that through the embodiment of the present application, by determining the background type of the detected area, the leftover items detection strategy matching the detected area under the static background and the leftover items detection strategy matching the detected area under the dynamic background can be determined, and the leftover items detection is performed on the detected area using the corresponding leftover items detection strategy, and when the detected area is detected to have leftover items, the category of the leftover items is identified. In this way, not only the leftover items detection under the dynamic background can be realized, but also the specific category information of the leftover items can be identified, so that the result of the leftover items detection is more accurate.

[0112] The embodiment of the present application also provides a computer program product. When the computer program is executed by a processor, each process of the above-mentioned method for detecting left-behind objects is implemented and the same technical effect can be achieved. To avoid repetition, it will not be described here.

[0113] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, devices, or computer program products. Therefore, the present application may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0114] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0115] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0116] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0117] In a typical configuration, an electronic device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0118] Memory may include non-permanent storage in a computer-readable medium, in the form of random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of a computer-readable medium.

[0119] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0120] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0121] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, devices or computer program products. Therefore, the present application may take the form of a complete hardware embodiment, a complete software embodiment or an embodiment combining software and hardware. Moreover, the present application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0122] The above is only an embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of the claims of the present application.

Claims

1. A method for detecting left-behind items, characterized in that: The method comprises: Determine the background type of the detected area; the background type includes a static background type and a dynamic background type; Determine, based on the background type of the detected area, a left-behind detection strategy that matches the detected area, and perform left-behind detection on the detected area based on the left-behind detection strategy; When it is detected that there are items left behind in the detected area, the item category matching model is used to perform category matching processing on the left-behind items, and based on the matching result, an item left-behind alarm processing is performed for the detected area.

2. The method according to claim 1, characterized in that The determining the background type of the detected area includes: Acquire an image sequence of the detected area, determine a baseline image and a plurality of reference images in the image sequence, and determine each first differential image between the baseline image and each of the reference images; Performing guided filtering processing on each of the first differential images to obtain a filtered image, and performing high-frequency information extraction processing on each of the first differential images to obtain a high-frequency information image; For each of the first differential images, a filtered image corresponding to the first differential image and a high-frequency information image corresponding to the first differential image are fused to obtain a fused image; based on the multiple fused images, the background type of the detected area is determined.

3. The method according to claim 2, characterized in that The determining the background type of the detected area based on the multiple fused images includes: Determining a reference fused image from the plurality of fused images, performing difference processing on the reference fused image and each of the other fused images from the plurality of fused images except the reference fused image, to obtain a plurality of second difference images; Determine the magnitude relationship between an average quantity value of white pixels in a plurality of the second differential images and a first preset quantity value threshold; If the judgment result is that the average quantity value of the white pixels in the plurality of the second differential images is less than the first preset quantity value threshold, it is determined that the background type of the detected area is a static background type; Otherwise, it is determined that the background type of the detected area is a dynamic background type.

4. The method according to claim 2, characterized in that: The step of determining a left-behind detection strategy matching the detected area based on the background type of the detected area, and performing left-behind detection on the detected area based on the left-behind detection strategy, includes: In the case where it is determined that the background type of the detected area is a static background type, obtaining any first image in the image sequence, determining a third differential image between the first image and the reference image, and determining whether an object is left behind in the detected area based on contour information of the target object in the third differential image; the first image includes the target object but does not include a target object corresponding to the target object; When it is determined that the background type of the detected area is a dynamic background type, any second image in the image sequence is acquired to determine whether the distance between the target item and the target object corresponding to the target item in the second image is greater than a preset distance threshold, and based on the determination result, determine whether an item is left behind in the detected area.

5. The method according to claim 4, characterized in that The determining whether an object is left behind in the detected area based on the contour information of the target object in the third differential image includes: Determining whether the contour area of ​​the target object in the third differential image is within a preset contour area range; In the case where it is determined that the contour area of ​​the target object in the third differential image is within a preset contour area range, determining the first image as a first target image, and determining whether the image sequence includes a plurality of first target images whose quantity values ​​exceed a second preset quantity value threshold and are continuous in time; When it is determined that the image sequence contains a plurality of first target images whose quantity values ​​exceed the second preset quantity value threshold and are continuous in time, it is determined that an object is left behind in the detected area, and the target object is determined to be a left-behind object.

6. The method according to claim 5, characterized in that The determining, based on the determination result, whether an object is left behind in the detected area includes: In a case where it is determined that the distance between the target item and the target object corresponding to the target item in the second image is greater than a preset distance threshold, determining the second image as a second target image, and determining whether the image sequence includes a plurality of second target images whose quantity value exceeds the second preset quantity value threshold and are continuous in time; When it is determined that the image sequence contains a plurality of second target images whose quantity values ​​exceed the second preset quantity value threshold and are continuous in time, it is determined that an object is left behind in the detected area, and the target object is determined to be the left-behind object.

7. The method according to claim 4, characterized in that Before acquiring any second image in the image sequence, the method further includes: Using the target detection model, the detected area is subjected to target object detection processing, and it is determined whether the detection result contains the target object; If the detection result includes a target object, the target object is bound to the corresponding target object using identification information; Otherwise, for each image in the image sequence, identify the moving object in the detected area and the area in the image where the moving object is located, and perform occlusion processing on the area; based on the image sequence after occlusion processing, perform object abandonment detection on the detected area.

8. A device for detecting left-behind objects, characterized in that: include: A background type determination module, used to determine the background type of the detected area; The background types include static background types and dynamic background types; A detection module, configured to determine a left-behind detection strategy that matches the detected area based on the background type of the detected area, and perform left-behind detection on the detected area based on the left-behind detection strategy; The category matching module is used to perform category matching processing on the left-behind items through the item category matching model when detecting that there are items left behind in the detected area, and to perform item left-behind alarm processing for the detected area based on the matching result.

9. An electronic device, characterized in that: It includes a processor, a communication interface, a memory and a communication bus; wherein the processor, the communication interface and the memory communicate with each other through the bus; the memory is used to store computer programs; the processor is used to execute the programs stored in the memory to implement the steps of the method for detecting left-behind items as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the method for detecting left-behind items according to any one of claims 1 to 7 are implemented.

Images