WO2023273011A9

WO2023273011A9 - Method, apparatus and device for detecting object thrown from height, and computer storage medium

Info

Publication number: WO2023273011A9
Application number: PCT/CN2021/123512
Authority: WO
Inventors: 蔡官熊; 赵晨; 方宝乐; 曾星宇; 赵瑞
Original assignee: 深圳市商汤科技有限公司
Priority date: 2021-06-29
Filing date: 2021-10-13
Publication date: 2023-02-09
Also published as: CN113409362A; WO2023273011A1; CN113409362B

Abstract

Disclosed in embodiments of the present disclosure are a method, apparatus and device for detecting an object thrown from height, and a computer storage medium, comprising: according to a time sequence of image acquisition, sequentially reading two adjacent images to be detected from multiple images to be detected, and generating an optical flow image corresponding to the two adjacent images by means of a preset optical flow model; determining optical flow images to be detected containing a thrown object from the multiple optical flow images, and determining position coordinates of the center point of the thrown object in each optical flow image to be detected; determining a motion trajectory of the thrown object according to the position coordinates of the center point of the thrown object in each optical flow image to be detected; and executing detection processing for an object thrown from height on the basis of the motion trajectory.

Description

High-altitude parabolic detection method, device, equipment, and computer storage medium

Cross References to Related Applications

This disclosure is based on the Chinese patent application with the application number 202110729203.4, the application date is June 29, 2021, and the application name is "High altitude parabolic detection method and device, equipment and computer storage medium", and claims the priority of the Chinese patent application , the entire content of this Chinese patent application is hereby incorporated by reference into this disclosure.

technical field

The present disclosure relates to the field of computer vision, and in particular to a method, device and equipment for detecting a high-altitude parabola, and a computer storage medium.

Background technique

In recent years, with the increase of urbanization rate, there are more and more high-rise buildings, and the dangers brought by high-altitude parabolic objects have been paid more and more attention. As an important technology in the intelligent video surveillance system, high-altitude parabolic detection detects dropped objects by analyzing the monitoring video of the building floor, so as to arm the floors of urban high-rise buildings and prevent possible high-altitude Real-time alarm for parabolic events.

Contents of the invention

Embodiments of the present disclosure provide a method, device, equipment, and computer storage medium for detecting high-altitude parabolic objects.

The disclosed technical solution is achieved in this way:

An embodiment of the present disclosure provides a high-altitude parabolic detection method, the method comprising:

sequentially reading two adjacent images to be tested from multiple frames of images to be tested according to the time sequence of image acquisition, and generating optical flow images corresponding to the two adjacent images to be tested through a preset optical flow model; Determining the optical flow images to be measured in which there are thrown objects from the optical flow images, and determining the position coordinates of the center points of the dropped objects in each optical flow image to be measured; according to the optical flow images in each optical flow image to be measured The coordinates of the center point of the dropped object determine the motion track of the dropped object; and the high-altitude parabolic detection process is performed based on the motion track.

In this way, on the one hand, the optical flow model based on deep learning is used to generate a dense optical flow map to detect moving objects, which not only has good robustness and high precision, but also takes less time and has less noise; on the other hand, On the basis of using the preset optical flow model for moving object detection, based on the robust single-frame plus multi-frame post-processing method, filter the single frame with disturbing objects, and find out the single-frame image with throwing objects , after removing single-frame false detections, track recovery and detection of high-altitude parabolic events are performed by combining the position information of the projectile on multiple frames.

An embodiment of the present disclosure provides a high-altitude parabolic detection device, including:

The reading part is configured to sequentially read two adjacent images to be tested from multiple frames of images to be tested according to the time sequence of image acquisition;

The generating part is configured to generate the optical flow images corresponding to the two adjacent images to be tested through a preset optical flow model;

The determining part is configured to determine from the optical flow images the optical flow images to be measured that there is a thrown object, and determine the position coordinates of the center point of the thrown object in each optical flow image to be measured;

The determining part is further configured to determine the trajectory of the dropped object according to the position coordinates of the center point of the dropped object in each optical flow image to be measured;

The processing part is configured to perform high-altitude parabolic detection processing based on the motion trajectory.

An embodiment of the present disclosure provides a high-altitude parabolic detection device. The high-altitude parabolic detection device includes a processor and a memory storing instructions executable by the processor. When the instructions are executed by the processor, the above-mentioned High-altitude parabolic detection method.

An embodiment of the present disclosure provides a computer-readable storage medium on which a program is stored, which is applied to a high-altitude parabolic detection device. When the program is executed by a processor, the above-mentioned high-altitude parabolic detection method is realized.

The present disclosure provides a computer program, including computer readable codes. When the computer readable codes run in electronic equipment and are executed by a processor in the electronic equipment, the above-mentioned high-altitude parabolic detection method is realized. .

The present disclosure provides a computer program product, which, when run on a computer, causes the computer to execute the above-mentioned high-altitude parabolic detection method.

According to the technical solution proposed by the embodiments of the present disclosure, the high-altitude parabolic detection equipment can sequentially read two adjacent images to be tested from multiple frames of images to be tested according to the time sequence of image acquisition, and generate adjacent images through the preset optical flow model. The optical flow images corresponding to the two images to be tested; from the optical flow images, determine the optical flow images to be tested that have thrown objects, and determine the coordinates of the center point position of the dropped objects in each optical flow image to be tested; The position coordinates of the center point of the dropped object in the photometric flow image determine the motion track of the dropped object; and the high-altitude parabolic detection process is performed based on the motion track. In this way, on the one hand, the present disclosure uses an optical flow model constructed based on deep learning to generate a dense optical flow map to detect moving objects, which not only has good robustness and high precision, but also takes less time and has less noise; On the one hand, on the basis of using the preset optical flow model for moving object detection, this disclosure also proposes a robust single-frame plus multi-frame post-processing method, which filters the single frame with disturbing objects to find out the For the single-frame image of the falling object, after removing the single-frame false detection, combined with the position information of the falling object on multiple frames for trajectory recovery and high-altitude parabolic event detection, which further improves the efficiency and accuracy of high-altitude parabolic detection.

Description of drawings

FIG. 1 is a schematic diagram of the implementation flow of the high-altitude parabolic detection method proposed by the embodiment of the present disclosure;

Fig. 2 is a schematic diagram of the implementation flow of the high-altitude parabolic detection method proposed by the embodiment of the present disclosure;

FIG. 3 is a schematic diagram of the trajectory of a dropped object proposed by an embodiment of the present disclosure;

FIG. 4 is a schematic diagram of the third implementation flow of the high-altitude parabolic detection method proposed by the embodiment of the present disclosure;

FIG. 5 is a schematic diagram 4 of the implementation flow of the high-altitude parabolic detection method proposed by the embodiment of the present disclosure;

FIG. 6 is a schematic diagram of the implementation flow of the high-altitude parabolic detection method proposed by the embodiment of the present disclosure;

FIG. 7 is a schematic diagram six of the implementation flow of the high-altitude parabolic detection method proposed by the embodiment of the present disclosure;

FIG. 8 is a schematic diagram of the implementation flow of the high-altitude parabolic detection method proposed by the embodiment of the present disclosure VII;

FIG. 9 is a schematic diagram of the eighth implementation flow of the high-altitude parabolic detection method proposed by the embodiment of the present disclosure;

FIG. 10 is a schematic diagram of the implementation flow diagram of the high-altitude parabolic detection method proposed by the embodiment of the present disclosure;

FIG. 11 is a schematic diagram of a scene of a high-altitude parabolic detection method proposed by an embodiment of the present disclosure;

12 is a schematic diagram of the composition and structure of the high-altitude parabolic detection device proposed by the embodiment of the present disclosure;

Fig. 13 is a schematic diagram of the composition and structure of a high-altitude parabolic detection device proposed by an embodiment of the present disclosure.

Detailed ways

In order to make the purpose, technical solutions and advantages of the present disclosure clearer, the present disclosure will be further described in detail below in conjunction with the accompanying drawings. All other embodiments obtained under the premise of creative labor belong to the protection scope of the present disclosure.

In the following description, references to "some embodiments" describe a subset of all possible embodiments, but it is understood that "some embodiments" may be the same subset or a different subset of all possible embodiments, and Can be combined with each other without conflict.

In the following description, the term "first\second\third" is only used to distinguish similar objects, and does not represent a specific ordering of objects. Understandably, "first\second\third" Where permitted, the specific order or sequencing may be interchanged such that the embodiments of the disclosure described herein can be practiced in sequences other than those illustrated or described herein.

Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this disclosure belongs. The terms used herein are only for the purpose of describing the embodiments of the present disclosure, and are not intended to limit the present disclosure.

Specifically, the high-altitude parabolic detection depends on the detection of moving objects in the image. In related technologies, the inter-frame difference method or the background difference method is often used to detect moving objects.

Among them, the principle of the inter-frame difference method is: when there is a moving object in the video, there will be a difference in grayscale between adjacent frames (or three adjacent frames), and the absolute value of the grayscale difference between two frames of images is calculated. value, the static object is all 0 on the difference image, and the moving object, especially the contour of the moving object, is non-zero due to the grayscale change. When the absolute value exceeds a certain threshold, it can be judged as a moving object. , so as to realize the detection function of the target, that is, the difference method between adjacent frames directly performs a difference operation on two adjacent frames of images, and takes the absolute value of the difference operation to form a moving object.

Although the frame-to-frame difference method can obtain the outer contour of the moving target, the method is simple and the computational complexity is small, but it has the defects of fixed camera, robust phase difference and low precision, and is generally only suitable for simple real-time motion detection.

Among them, the basic principle of the background subtraction method is to subtract the current frame in the image sequence from the background reference model (background image) that has been determined or obtained in real time, find the difference, and calculate the area where the pixel difference with the background image exceeds a certain threshold. It is the motion area, so as to determine the motion position, outline, size and other characteristics.

The background subtraction method can obtain the entire area of the moving target. The advantage is fast and accurate, but the robustness is poor, the camera must be fixed, and it is difficult to use inter-frame information, and image post-processing is difficult, especially in weak light or rainy and snowy weather. more affected.

In order to overcome the defects of the above-mentioned inter-frame difference method and background difference method, the moving object detection based on the optical flow method is further proposed in related technologies, such as the algorithm using the optical flow constraint equation, the traditional (Lucas Kanade, KL) algorithm, the The optical flow method mainly uses the changes of pixels in the image sequence in the time domain and the correlation between adjacent frames, and calculates the motion information of objects between adjacent frames according to the correspondence between the previous frame and the current frame.

Although it can detect moving objects in the entire area, it is suitable for camera still and moving situations, with good robustness and high precision, but most of the optical flow calculation methods have a huge amount of calculation and complex structure, resulting in high calculation time and high cost. It is difficult to guarantee speed and accuracy at the same time.

It can be seen that under the current mass video data, the traditional high-altitude parabolic detection method has been difficult to meet the high requirements of high-altitude parabolic detection in terms of speed and accuracy. In view of this, how to ensure the high requirements of high-altitude parabolic detection in terms of speed and accuracy is an urgent problem to be solved, which is the content to be discussed in the embodiments of the present disclosure, and will be described in conjunction with the following specific embodiments.

Embodiments of the present disclosure provide a high-altitude parabolic detection method, device, equipment, and computer storage medium. On the one hand, an optical flow model based on deep learning is used to generate a dense optical flow graph to detect moving objects, which is not only robust Good, high precision, and less time-consuming, less noise; on the other hand, on the basis of using the preset optical flow model for moving object detection, this disclosure also proposes a robust single-frame plus multi-frame post-processing The method is to filter the single frame with disturbing objects, find out the single frame image with the thrown object, and remove the false detection of single frame, and combine the position information of the dropped object on multiple frames to recover the trajectory and detect the high-altitude parabolic event , which further improves the efficiency and accuracy of high-altitude parabolic detection.

The high-altitude parabolic detection method proposed in the embodiment of the present disclosure is applied to high-altitude parabolic detection equipment. The following describes the exemplary application of the high-altitude parabolic detection equipment proposed by the embodiments of the present disclosure. The high-altitude parabolic detection equipment proposed by the embodiments of the present disclosure can be implemented as mobile terminals, notebook computers, tablet computers, desktop computers, servers, and various industrial equipment, etc. .

In the following, the technical solutions in the embodiments of the present disclosure will be clearly and completely described with reference to the drawings in the embodiments of the present disclosure.

An embodiment of the present disclosure provides a high-altitude parabolic detection method. FIG. 1 is a schematic diagram of the implementation flow of the high-altitude parabolic detection method proposed by the embodiment of the present disclosure. As shown in FIG. 1, in the embodiment of the present disclosure, the high-altitude parabolic The detection method may include the following steps:

S100. According to the time sequence of image acquisition, sequentially read two adjacent images to be tested from the multiple frames of images to be tested, and generate optical flow images corresponding to the two adjacent images to be tested by using a preset optical flow model.

It should be understood that, in the embodiments of the present disclosure, the high-altitude parabolic detection process may be performed on historical events, or the high-altitude parabolic detection process may be performed on current events in real time.

In some embodiments, the image to be tested refers to an image that requires moving object detection. Wherein, the image may be collected in real time or stored locally.

In some embodiments, based on different configurations of image acquisition and monitoring equipment, the image to be tested may be an RGB color image, a grayscale image, or other sensor image data (such as an infrared image).

Among them, the high-altitude parabolic detection equipment can be equipped with image acquisition and monitoring equipment, such as a camera, through which the image to be tested can be collected in real time; or, a video historically collected by the camera can be stored locally, and the video stream can be read and analyzed locally. Get the image to be tested.

Among them, FIG. 2 is a schematic diagram of the second implementation flow of the high-altitude parabolic detection method proposed by the embodiment of the present disclosure. As shown in FIG. 2, the method for obtaining the image to be tested may include the following steps:

S101. Acquire an initial image, and determine a target detection area from the initial image based on a preset polygonal outline.

It should be understood that, for any image acquisition monitoring device, the range of images it can acquire may include target monitoring objects (such as buildings) and greenery or fixed equipment downstairs, such as light poles and numbers. Among them, green facilities such as trees will have a certain impact on the detection results of moving objects. The impact can be that fallen leaves are easily judged as high-altitude objects, and additional processing tasks are brought to image post-processing, slowing down the high-altitude Speed of parabola detection processing.

Therefore, in the embodiment of the present disclosure, the monitoring screen of each image acquisition and monitoring device can be analyzed, and the area without obstructions can be marked as the monitoring area of the target monitoring object (such as a building), and then when performing subsequent image analysis, Only analyze the monitoring area, ignore the image content of other non-monitoring areas, and speed up the processing speed of high-altitude parabolic detection.

In some embodiments, the initial image refers to an image within a monitoring range that is collected in real time by an image acquisition monitoring device, or a video within a monitoring range that is locally historically stored.

In the embodiment of the present disclosure, a polygonal contour C, that is, a preset polygonal contour, can be manually determined first, and then the region of interest R is demarcated from the initial image by the preset polygonal contour as the target detection region, ignoring the Other interfering graphic content.

Here, the high-altitude parabolic detection device can perform binarization processing on the initial image based on the preset polygonal outline, first set the image pixels of the target detection area corresponding to the preset polygonal outline to 1, and set the image pixels outside the target detection area to 0, Then perform pixel point multiplication between the binarized image and the original image, so that the image of the target detection area retains the original pixel value, and the image pixels outside this area are all 0, so as to determine the target detection area existing in the initial image .

It can be seen that, except for the selected target drop area, the pixels in other areas are all 0, and the obtained target detection area can cover the target monitoring object (such as a building) to the greatest extent, that is, the maximum coverage of the potential trajectory of the dropped object, and also Avoids the effects of trees, sky, light poles, and some other distractions.

S102. Generate a minimum detection frame based on the target detection area, and perform image segmentation processing on the initial image based on the minimum detection frame to obtain an image to be tested.

It should be understood that in order to cover the potential trajectory of the dropped object to the greatest extent, the target dropped object may be an irregular polygon in the actual situation, so before the model is input, the area of the target dropped object needs to be regularized. Among them, the minimum detection frame of the Bounding box corresponding to the target drop area can be generated.

Among them, the edge position coordinates of the target drop area can be taken to generate a minimum detection frame. It can take the leftmost edge coordinate x ₁ , the rightmost edge coordinate x ₂ , the uppermost edge coordinate y ₁ and the lowermost edge coordinate y ₂ of the target drop area, and then generate the minimum border (x ₁ , y ₁ , x ₂ , y ₂ ).

In an implementation manner of the embodiment of the present disclosure, image segmentation processing may be performed on the initial image based on the minimum detection frame, that is, image frame cropping, and then the image to be tested may be obtained.

It can be seen that, in the embodiment of the present disclosure, based on S101a-S101b, the negative impact on false detection of dropped objects can be reduced, the workload of image post-processing can be reduced, and the speed of high-altitude parabolic detection processing can be further accelerated.

In the embodiment of the present disclosure, the preset optical flow model is a model with high precision and fast inference speed constructed and trained based on deep learning technology through convolutional neural network and recurrent neural network. The input of the optical flow model is two adjacent frames of images to be tested, and the output is an optical flow map that can reflect the speed and direction of the moving object, so as to realize the detection of the moving object in the dropped image.

In some embodiments, the optical flow image is an image that reflects the moving speed and moving direction of the moving object. In order to determine the moving object that may exist in the video stream, an optical flow map can be generated based on adjacent frames in the video stream, which can reflect the motion The movement of the object from the point of the previous frame image to the second frame image. It can be seen that all existing moving objects can be found based on the optical flow map.

In an implementation manner of an embodiment of the present disclosure, each adjacent two frames of images to be tested among the multiple frames of images to be tested obtained through the processing of S101a-S101b may be sequentially formed into an image sample pair according to the time sequence of image acquisition, For example, the image sample pair (I _k , I _k+1 ), and then input the image sample pair into the preset optical flow model to obtain a dense optical flow map for reflecting moving objects, that is, to generate adjacent The optical flow images corresponding to the two images to be tested.

Here, the following formula can be used to represent the current optical flow image:

f _k ＝(u _k , v _k ) (1)

Among them, in the formula (1), u _k represents the moving speed of the moving object in the vertical direction, v _k represents the moving speed in the horizontal direction, and k represents the frame sorting value of the image, such as the kth optical flow image.

In some embodiments, if the high-altitude parabolic detection process is performed on historical events, S101a-S101b can be performed on each frame of image in the locally stored historical video to obtain the image to be tested corresponding to each frame of image, Then, the images to be tested obtained after image segmentation are read in batches, and image sample pairs are sequentially constructed from two adjacent frames of images to be tested based on the image acquisition time, and input into the preset optical flow model for moving object detection to generate each adjacent The optical flow images corresponding to the two images to be tested. Correspondingly, if the high-altitude parabolic detection processing is performed on the currently occurring event, then execute S101a-S101b in real time according to the acquisition time to obtain the image to be tested, and input the two adjacent frames of the image to be tested into the preset optical flow model for motion in real time Object detection to generate optical flow images corresponding to every two adjacent images to be tested.

S110. Determine, from the optical flow images, the optical flow images to be measured that contain the dropped objects, and determine the position coordinates of the center points of the dropped objects in each optical flow image to be measured.

It is understandable that there are actually many scenes of high-altitude parabolic objects. Different natural conditions such as day, night, and rain affect the detection accuracy of dropped objects. At the same time, there are some natural objects, such as swinging leaves and flying objects. Birds, moving people, etc. will affect the accurate detection of dropped objects. Therefore, in order to overcome the detection error caused by other non-high-altitude throwing moving objects to the high-altitude parabola detection during the detection process, in the embodiment of the present disclosure, the optical flow images corresponding to the two adjacent images to be tested are generated based on the preset optical flow model After that, the single-frame post-processing method can be used to filter the interference images or abnormal images of non-high-altitude throwing objects to obtain images with real throwing objects and less noise, and reduce the false detection rate of high-altitude parabolic detection.

It should be understood that if there are too many types of moving objects in the current optical flow image, it indicates that there is too much noise in the image, or if the motion response intensity of the moving object in the current optical flow image is too large, it indicates that there is an abnormality in the frame image, such as a bird Flying in front of the camera, this type of image has a greater negative impact on the false detection caused by the detection of dropped objects, so this type of optical flow image has little effect in the high-altitude drop detection process, and will slow down the high-altitude detection process. speed.

Therefore, in the embodiment of the present disclosure, the obtained optical flow images can be further analyzed and processed, and the current optical flow images determined to be excessively noisy, with too many types of moving objects or abnormal optical flow images can be filtered, and the current optical flow images can be filtered out. There is an optical flow image to be measured of a thrown object.

In an implementation manner of the embodiments of the present disclosure, a single-frame binarization method may be used to determine whether there is a real falling object in the current optical flow image.

In some embodiments, for any optometry flow image determined to contain a dropped object, the position coordinates of the center point of the dropped object in any optical flow image to be measured may be continuously determined. Among them, the position clustering method can be introduced to determine the position coordinates of the center point of the dropped object.

According to the method of S100-S110, the optical flow image generation of the real-time collected image to be tested can be sequentially performed, whether there is a thrown object in the corresponding optical flow image, and the position coordinates of the center point of the thrown object in the optical flow image to be tested can be determined, etc. By repeating the process, multiple frames of the optical flow images to be measured and the coordinates of the central point of the dropped object in each optical flow image to be measured can be obtained.

S120. Determine the trajectory of the dropped object according to the position coordinates of the center point of the dropped object in each optical flow image to be measured.

S130. Perform high-altitude parabolic detection processing based on the motion trajectory.

In the embodiment of the present disclosure, the trajectory restoration of the dropped object can be performed based on a multi-frame post-processing method, wherein the trajectory restoration of the dropped object can be performed based on the position coordinates of the center point of the dropped object in each optical flow image to be measured.

Among them, since the coordinates of the center point of the dropped object in each optical flow image to be measured are independent and single, the coordinates of the center point of the dropped object in each optical flow image to be measured can be first calculated according to the coordinates of each optical flow image to be measured Interpolation processing is performed in order of generation time, and then the center point position coordinates of the dropped objects in each optical flow image to be measured are mapped on a plane image, and the center point position coordinates of the dropped objects are connected to facilitate The trajectory corresponding to the dropped object is formed.

Exemplarily, FIG. 3 is a schematic diagram of a motion trajectory of a dropped object proposed by an embodiment of the present disclosure, and the motion trajectory is formed by position coordinates of multiple center points of the dropped object in multiple consecutive frames of optical flow images to be measured.

It can be understood that, since some buildings or trees may be blocked during the object throwing process, there are determined multiple frames of images to be tested in which there are dropped objects and the frames are not continuous, but based on The degree of frame density can also restore the trajectory of the dropped object based on the coordinates of the center point of the dropped object in most of the optical flow images to be tested.

In an implementation of the embodiment of the present disclosure, after restoring the trajectory corresponding to the dropped object, a trajectory straight line fitting method can be introduced to detect the high-altitude parabolic event of the dropped object, so as to further determine the corresponding motion of the dropped object Whether the event is a high-altitude parabolic event. Among them, the angle between the trajectory after the straight line fitting and the vertical direction (the slope of the trajectory after the straight line fitting), the height of the object falling, and other information can be used to detect the high-altitude parabolic event of the dropped object.

Exemplarily, the event that a person raises an apple with his hands above his head in front of the camera is also a motion event of throwing an object, but this event is not a high-altitude parabolic event in essence; while a person standing on the eighth floor of a building will The motion event of a dropped object such as an apple falling to the ground is considered a high-altitude parabolic event.

An embodiment of the present disclosure provides a high-altitude parabolic detection method. The high-altitude parabolic detection device sequentially reads two adjacent images to be tested from multiple frames of images to be tested according to the time sequence of image collection, and uses the preset optical flow model Generate optical flow images corresponding to two adjacent images to be tested; determine from the optical flow images the optical flow images to be measured where there is a thrown object, and determine the position coordinates of the center point of the dropped object in each optical flow image to be measured; The trajectory of the dropped object is determined according to the position coordinates of the center point of the dropped object in each optical flow image to be measured; and the high-altitude parabolic detection process is performed based on the trajectory. In this way, on the one hand, the present disclosure uses an optical flow model constructed based on deep learning to generate a dense optical flow map to detect moving objects, which not only has good robustness and high precision, but also takes less time and has less noise; On the one hand, on the basis of using the preset optical flow model for moving object detection, this disclosure also proposes a robust single-frame plus multi-frame post-processing method, which filters the single frame with disturbing objects to find out the For the single-frame image of the falling object, after removing the single-frame false detection, combined with the position information of the falling object on multiple frames for trajectory recovery and high-altitude parabolic event detection, which further improves the efficiency and accuracy of high-altitude parabolic detection.

Based on the above-mentioned embodiments, in yet another implementation of the embodiment of the present disclosure, FIG. 4 is a schematic diagram of the implementation process of the high-altitude parabolic detection method proposed in the embodiment of the present disclosure. As shown in FIG. The method for determining the optical flow image to be tested in which there is a thrown object in the flow image also includes the following steps:

S111. For any optical flow image, generate a binary image corresponding to any optical flow image; wherein the binary image includes a foreground moving object with a first pixel value and a background non-moving object with a second pixel value.

In the embodiment of the present disclosure, after the optical flow images corresponding to two adjacent images to be tested are generated based on the preset optical flow model, a single-frame binarization method can be used to determine whether there is a real falling object in the current optical flow image Judgment is made to determine the optical flow image to be measured where there is a thrown object.

Wherein, for any optical flow image, the optical flow image may first be converted into a binary image, that is, a grayscale image including only the first pixel value and the second pixel value in the image. For example, a binary image corresponding to an optical flow image includes a pixel value of 0 and a pixel value of 255.

In an implementation of the embodiment of the present disclosure, in order to distinguish between moving objects and non-moving objects in the image, it is possible to choose to determine the region where the pixel value in the image is greater than or equal to the critical pixel gray value as the region where the moving object exists, and the The pixel value of the area in the binary image is set as the first pixel value, otherwise, the area where the pixel value in the image is smaller than the critical pixel gray value is determined as the area where the non-moving object is located, and the pixel value of the area in the binary image If the value is set to the second pixel value, the binary image corresponding to any optical flow image can be obtained.

Here, the first pixel value in the binary image may be 255, representing a foreground moving object, and the second pixel value may be 0, representing a background non-pixel object.

Among them, FIG. 5 is a schematic diagram of the fourth implementation flow of the high-altitude parabolic detection method proposed by the embodiment of the present disclosure. As shown in FIG. 5, the method for generating the binary image corresponding to any optical flow image may include the following steps:

S111a. Perform single-channel grayscale conversion processing on any optical flow image to obtain a single-channel grayscale image corresponding to any optical flow image.

S111b. Perform normalization processing on the single-channel grayscale image to obtain a normalized grayscale image corresponding to any optical flow image.

S111c. Perform binarization processing on the normalized grayscale image to obtain a binary image corresponding to any optical flow image.

In the embodiment of the present disclosure, when converting the current optical flow image to a binary image, the current optical flow image can be converted to a grayscale image first, and then the optical flow image can be obtained by binarizing the grayscale image The corresponding binary image.

It should be understood that since the optical flow image is a multi-channel image, conversion of the multi-channel optical flow image to a single-channel grayscale image is performed first.

Specifically, the grayscale image conversion can be performed based on the following formula:

Among them, in formula (2),

Characterize the single-channel grayscale image corresponding to the current optical flow image (or the k-th frame optical flow image).

Then for the single-channel grayscale image corresponding to the current optical flow image

Perform normalization processing so that the gray value of the pixels is evenly distributed between 0-255 values, and obtain a normalized gray image to reduce interference to subsequent processing. Further, the normalized grayscale image can be transformed into a binary image.

Here, the critical pixel gray value for image binarization can be set in advance, and the pixel whose pixel value is greater than the critical pixel gray value is determined as a foreground moving object with a pixel value of 255 in the binary image. Correspondingly, the pixel value Pixels smaller than the critical pixel gray value are determined as background moving objects with a pixel value of 0 in the binary image.

For example, the binarized image corresponding to the current optical flow image can be obtained by performing image binarization processing such as the Otsu method OSTU

S112. In response to the pixel ratio of the foreground moving object in the binary image being less than or equal to a preset ratio threshold, determine any optical flow image as an optical flow image to be measured with a falling object.

In the embodiment of the present disclosure, for any optical flow image, after obtaining the binarized image corresponding to the optical flow image f _k

Afterwards, the number of pixels occupied by the foreground moving object in the binary image can be used to characterize the probability of moving objects in the optical flow image, then the current optical flow can be further calculated based on the ratio of the number of pixels of the foreground moving object in the binary image. Whether there is a falling object in the image is judged.

Considering the noise caused by too many moving objects and the negative impact of high-altitude parabolic detection caused by excessive motion response strength of moving objects, the preset ratio threshold of the number of pixels of moving objects in binary images can be set, that is, moving objects in The ratio of the number of pixels in the binary image to the entire image cannot exceed the specified threshold. At this time, it is judged whether there is a thrown object in the current optical flow image based on the comparison result of the pixel data ratio of the foreground moving object in the binary image and the preset ratio threshold.

Among them, when the proportion of the number of pixels of the foreground moving object in the binary image is less than or equal to the preset proportion threshold, it can be determined that there is a throwing object in the current optical flow image; otherwise, if the foreground moving object is in the binary image If the proportion of the number of pixels in is greater than the preset proportion threshold, it can be determined that there is no thrown object in the current optical flow image.

Based on the above operations, images with falling objects in the optical flow image can be screened out, and used as the optical flow image to be measured for subsequent high-altitude parabolic detection processing.

It can be seen that in the embodiment of the present disclosure, single-frame binarization processing can be performed on the current optical flow image, and based on the ratio of the number of pixels of the foreground moving object in the binarized image to the optical flow image with excessive noise or abnormal Filtering is performed to filter out the optical flow images to be tested that have dropped objects, thereby improving the accuracy and speed of high-altitude detection processing.

Based on the above-mentioned embodiment, in another embodiment of the present disclosure, FIG. 6 is a schematic diagram of the implementation process of the high-altitude parabolic detection method proposed in the embodiment of the present disclosure. As shown in 6, determine each optical flow image to be measured. The method for the center point position coordinates of the dropped object may comprise the following steps:

S113. For any optical flow image to be measured, obtain the initial position coordinate set of the foreground moving object in the binary image, and analyze the foreground motion in any optical flow image to be measured based on the preset clustering algorithm and the initial position coordinate set Objects are classified to obtain at least one moving object and coordinate subsets corresponding to each moving object.

S114. For any moving object, calculate the coordinate average value corresponding to the coordinate subset of any moving object, and determine the coordinate average value as the center point position coordinate of any moving object.

S115. Determine the target pixel value at the coordinates of the center point of any moving object in the normalized grayscale image corresponding to any optical flow image to be measured, and determine the moving object with the largest target pixel value as the thrown object , and the center point position coordinates corresponding to the moving object with the largest target pixel value are determined as the center point position coordinates of the dropped object.

In the embodiment of the present disclosure, for any optical flow image to be measured with a falling object, a clustering algorithm may be introduced to determine the position coordinates of the center point of the falling object in the optical flow image to be measured.

In an implementation of the embodiment of the present disclosure, when it is determined that there is a falling object in the current optical flow image, all pixel position coordinates of the foreground moving object in the binary image can be obtained to form a two-dimensional array of position coordinates , which is the set of initial position coordinates.

It should be understood that there may be more than one moving object in the current optical flow image, including real falling objects, falling leaves, etc., and there is not only one real falling object. In order to determine the position of the dropped object, in the embodiment of the present disclosure, the position of the dropped object may be determined through a position clustering algorithm, such as not limited to the Kmeans clustering algorithm.

Among them, the position clustering algorithm can be applied to perform preliminary classification on the moving objects existing in the current optical flow image. Wherein, based on the pixel position coordinates of the foreground moving objects as clustering parameters, the foreground moving objects can be clustered into at least one category, such as throwing objects into one category, and leaves into one category. And determine the clusters corresponding to the real dropped objects.

In an implementation manner of an embodiment of the present disclosure, for any optical flow image to be measured, a set of initial position coordinates of the foreground moving object in the binary image can be obtained, and the initial position coordinates can be classified based on a preset position clustering algorithm The set is clustered to obtain at least one set of coordinate subsets, and each coordinate subset corresponds to a type of moving object, thereby implementing a position-based clustering algorithm to classify moving objects in the optical flow image to be measured.

Here, after completing the classification of the moving objects in the optical flow image to be tested and obtaining the coordinate subsets of each type of moving objects, the mean value calculation can be performed based on the coordinate subsets of each type of moving objects, and the determined coordinate mean value can be used as each The center point position coordinates of a class of moving objects, and from the normalized grayscale image corresponding to the optical flow image to be measured, determine the pixel gray value at the corresponding position based on the center point position coordinates.

In the embodiments of the present disclosure, in order to realize timely alarm processing for high-altitude parabolic events, the parabolic interval of high-altitude parabolic events can be quickly determined by reducing the amount of data without affecting the detection accuracy of high-altitude parabolic events. Among them, on the premise of not considering the current number of dropped objects and the types of dropped objects, the moving object with the largest optical flow amplitude in each optical flow image to be tested can be used as the dropped object, and its corresponding center point position The coordinates are used as the coordinates of the center point of the dropped object, and based on the coordinates of the center point of the dropped object in each optical flow image to be measured, the interval of the parabolic event is quickly determined, and the alarm processing of the high-altitude parabolic event is performed in time.

Among them, based on the characteristic that the larger the response intensity/optical flow amplitude, the larger the pixel value reflected on the grayscale image, the normalized grayscale image corresponding to each moving object in the optical flow image to be measured can be determined After the gray value of the pixel at the coordinates of the center point position in , determine the type of moving object with the largest pixel gray value as the dropped object, and determine the center point position coordinates of this type of moving object as the center point of the dropped object Position coordinates.

It can be understood that, in the case where there is only one falling object, the response intensity in the current optical flow image, that is, the class with the largest optical flow amplitude may be the real falling object, and the weaker response intensity should be one of them. Non-throwing objects such as leaves; in the case that there may be multiple throwing objects, the response intensity in the current optical flow image, that is, the class with the largest optical flow amplitude may be one of the multiple throwing objects .

It should be understood that during the high-altitude parabolic detection process, after the high-altitude parabolic alarm is issued, the user can view the parabolic event interval, such as a surveillance video corresponding to the parabolic event, although the optical flow to be measured may be taken when there may be multiple thrown objects The position coordinates of the center point of the moving object with the largest optical flow amplitude in the image are used as the position coordinates of the parabolic object, but after the parabolic event interval is determined, you can view the event interval corresponding to each falling object that may exist in the interval Parabolic events to view.

In this way, the position coordinates of the center point of the moving object with the largest optical flow amplitude in each optical flow image to be measured are used as the position coordinates of the center point of the dropped object, and based on the position coordinates of the center point of the dropped object in each optical flow image to be measured Quickly determining the interval of parabolic events can not only effectively filter the interference objects in the image, but also improve the speed of high-altitude parabolic detection on the basis of reducing the amount of calculation and processing data.

Among them, FIG. 7 is a schematic diagram six of the implementation flow of the high-altitude parabolic detection method proposed by the embodiment of the present disclosure. As shown in FIG. Before describing the trajectory of the dropped object, that is, before step 130, the method for performing high-altitude parabolic detection processing also includes the following steps:

Step 140, from each optical flow image to be measured, determine the optical flow image to be measured from the start optical flow image to the end optical flow image to be measured of the parabolic event corresponding to the dropped object.

It can be understood that the image acquisition and monitoring equipment collects images continuously in real time, and there may be image acquisition and monitoring equipment that collects images of multiple parabolic events that occur at different times; that is, the optical flow to be measured belonging to the same parabolic event The images are generally frame-continuous, or, when there are a few projectiles that are blocked by some objects during the throwing process, the frame interval between the optical flow images to be measured belonging to the same parabolic event is less than a certain threshold. Correspondingly, if the frame interval between the optical flow images to be measured is too far apart, the two frames of images may correspond to parabolic events occurring at different times.

In the embodiment of the present disclosure, different parabolic events occurring at different times can be divided based on the frame interval value of the optical flow image to be measured, and the start frame and end frame corresponding to the parabolic events occurring at different times can be determined.

In one embodiment, the frame interval threshold between different parabolic events can be set in advance, and the optical flow image to be measured with a thrown object obtained after single frame processing is sequentially compared with the frame interval value of the previous frame of the optical flow image to be measured to determine whether the two frames of optical flow images to be measured belong to the same parabolic event, so as to update the start frame and end frame of the parabolic event.

Exemplarily, the first image and the second image of two frames of optical flow images to be measured are taken as an example for description.

Wherein, after it is determined that the first image belongs to the first parabolic event and the first image is determined as the end frame corresponding to the first parabolic event, the frame interval value between the current second image and the first image may be calculated. When the frame interval value is less than or equal to the preset frame interval threshold, it is determined that the second image and the first image belong to the same parabolic event, then the end frame of the first parabolic event is updated to the second image; correspondingly, in the If the frame interval value is greater than the preset frame interval threshold, it is determined that the second image and the first image do not belong to the same parabolic event, then it indicates that the end frame of the first parabolic event is the first image, and the second image belongs to a new parabolic event That is, the second parabolic event, at this time, the second image may be determined as the starting frame of the new second parabolic event.

Correspondingly, after determining the parabolic event to which the second image belongs, it is possible to continue to determine the next frame of the optical flow image to be measured, that is, the parabolic event to which the third image belongs, if the frame interval between the third image and the second image is less than the preset If the frame interval threshold is set, then the third image and the second image belong to the same second parabolic event, and at this time, the end frame of the second parabolic event can be determined as the third image. Repeat this step to continue to determine the frame interval value between the next frame of the optical flow image to be measured and the third image. If the frame interval value is still smaller than the preset frame interval threshold, update the end frame of the second parabolic event to the next Frame the optical flow image to be measured, repeat this step until the frame interval between the optical flow image to be measured and the previous optical flow image to be measured is greater than the preset frame interval threshold, then the optical flow image of the previous frame to be measured is The second parabolic event finally ends the frame, so that a complete parabolic event can be determined from the beginning to the end of the optical flow image to be measured, that is, the interval in which a parabolic event occurs can be determined.

For example, the logic code for determining the parabolic event to which the optical flow image to be measured belongs is as follows, initialize the parabolic event start frame subscript d _s =0, end frame subscript d _e =0, denote δ _e as the frame interval between parabolic events Threshold, and initialize starts=[], ends=[] at the same time, used to record the start and end frames of each parabolic event.

Based on the above logic code, the interval of each parabolic event can be returned, that is, the list starts and ends from the start frame to the end frame.

Here, based on the chronological order of the optical flow images to be measured, the determination process of the above-mentioned motion events to be measured can be cycled, and the start frame and the end frame corresponding to different parabolic events can be determined, that is, the starting frame of the optical flow image to be measured to the end of the parabolic event Optical flow image.

In some embodiments, after determining the interval in which the parabolic event corresponding to the thrown object occurs, the parabolic object in each frame of the optical flow image to be measured from the start optical flow image to the end optical flow image of the parabolic event can be The position coordinates of the center point of the object are interpolated according to the time sequence, and the coordinates of the center point position in each optical flow image to be measured on the parabolic event interval of the dropped object are mapped on a plane image and connected, so that Form the trajectory of a parabolic object.

It can be seen that in the embodiments of the present disclosure, the start frame and end frame of different motion events to be detected can be determined based on the frame interval difference, so as to accurately divide different motion events to be measured and improve the accuracy of parabolic detection processing.

Based on the above-mentioned embodiments, in yet another implementation of the embodiment of the present disclosure, FIG. 8 is a schematic diagram of the implementation process of the high-altitude parabolic detection method proposed in the embodiment of the present disclosure. As shown in FIG. 8, the high-altitude parabolic detection equipment is based on the motion The method for track execution high-altitude parabolic detection processing also includes the following steps:

S121. Perform a straight line fitting process on the motion trajectory of the dropped object, obtain a fitted straight line corresponding to the motion trajectory, and determine an included angle between the fitted straight line and the vertical direction.

S122. Based on the position coordinates of the center point of the dropped object in each optical flow image to be measured in the parabolic event, determine the pixel value of the center point position from the corresponding normalized grayscale image, and perform a calculation on the pixel value of each center point position Accumulation processing to obtain the pixel accumulation value.

S123. In the parabolic event, the coordinates of the center point of the dropped object in the initial optical flow image to be measured and the coordinates of the center point of the dropped object in the optical flow image to be measured are calculated to obtain the coordinate difference value.

S124. Execute a high-altitude parabola detection process based on the angle between the fitted straight line and the vertical direction, the pixel accumulation value, and the coordinate difference.

In the embodiment of the present disclosure, according to the time sequence of each optical flow image to be measured in the optical flow image to be measured from the beginning of the parabolic event to the end of the optical flow image to be measured, the position coordinates of the center point of the thrown object in each optical flow image After performing interpolation processing to obtain the trajectory of the dropped object, the high-altitude parabolic detection of the parabolic event can be realized based on the trajectory.

It should be understood that the trajectory of the dropped object does not satisfy a certain function rule, but there is a certain regularity. In the embodiment of the present disclosure, a straight line fitting method can be introduced to perform a straight line fitting process on the trajectory of the dropped object , to get the fitted straight line corresponding to the motion trajectory.

In one embodiment, since the parabolic event is not a real high-altitude parabolic event, it may be similar to a person passing an object by hand, or lifting an object above his head, such as an event where a person stands at the window and eats an apple, and the hand picks up the apple Put it on your mouth, bite your hand and put it down. Therefore, the angle between the fitted line and the vertical direction can be calculated, which can be used as one of the factors for judging whether a parabolic event is a high-altitude parabolic event.

For example, FIG. 9 is a schematic diagram of the angle between the trajectory fitting line and the vertical direction proposed by the embodiment of the present disclosure. As shown in FIG. 9 , the angle between the fitted line and the vertical direction is θ.

In another embodiment, it is also possible to determine the center of the thrown object from the normalized grayscale images corresponding to each optical flow image to be measured in the optical flow image to be measured from the beginning of the parabolic event to the end of the optical flow image to be measured The pixel value at the coordinates of the point position, and these pixel values are cumulatively summed to obtain the pixel cumulative value, and the pixel cumulative value is used as one of the factors for judging whether the parabolic event is a high-altitude parabolic event.

In another embodiment, it should be understood that a real high-altitude parabolic event must have a certain height parabolic range. For example, a parabolic event with a height of more than three floors belongs to a high-altitude parabolic event. Therefore, the ordinate of the dropped object in the initial optical flow image of the parabolic event determined based on the above-mentioned embodiment is the maximum ordinate, that is, the parabolic The highest point of the event; and the ordinate of the end frame of the parabolic event, that is, the minimum ordinate, that is, the lowest point of the parabolic event, perform a difference operation, and then determine the parabola based on the coordinate difference between the maximum ordinate and the minimum ordinate The height difference between the lowest point and the highest point of the event, that is, the height of the parabola, can also be used as one of the factors for judging whether the parabolic event is a high-altitude parabolic event.

In the embodiment of the present disclosure, it is possible to further determine whether the parabolic event belongs to High-altitude parabolic events are judged.

Among them, FIG. 10 is a schematic diagram of the implementation process of the high-altitude parabolic detection method proposed by the embodiment of the present disclosure. As shown in FIG. The method for performing high-altitude parabolic detection processing on parabolic events may include the following steps:

S124a. In response to the fact that the angle between the fitted straight line and the vertical direction is less than a preset angle threshold, the pixel accumulation value is less than a preset pixel threshold, and the coordinate difference is greater than a preset height threshold, determine that the dropped object is a high-altitude thrown object, And the corresponding parabolic event is a high-altitude parabolic event.

S124b. In response to the angle between the fitted straight line and the vertical direction being greater than or equal to the preset angle threshold, or the pixel accumulation value is greater than or equal to the preset pixel threshold, and the coordinate difference is less than or equal to the preset height threshold, determine the thrown The falling object is not a high-altitude throwing object, and the corresponding parabolic event is not a high-altitude parabolic event.

In one implementation of the embodiment of the present disclosure, the included angle threshold, pixel threshold and height threshold of the high-altitude parabolic event are preset, and the angle between the straight line and the vertical direction after the parabolic event is fitted is compared with the preset included angle threshold, Compare the cumulative value of the pixel value at the center point of the dropped object with the preset pixel threshold, and compare the height difference between the lowest point and the highest point of the parabolic event with the preset height threshold, and combine the comparison of the three judgment factors As a result, it is judged whether the dropped object is a high-altitude thrown object, in other words, it is determined whether the parabolic event corresponding to the dropped object is a high-altitude parabolic event.

Among them, after the parabolic event is fitted, the angle between the straight line and the vertical direction is less than the preset angle threshold, the accumulated value of the pixel value at the center point of the dropped object is less than the preset pixel threshold, and the height of the lowest point and the highest point of the parabolic event The difference is greater than the preset height threshold, that is, when the three judging factors meet the corresponding preset threshold conditions at the same time, it is determined that the dropped object is a high-altitude thrown object, and the corresponding parabolic event is a high-altitude parabolic event.

In another embodiment of the present disclosure, when at least one of the three judging factors does not meet the corresponding preset threshold condition, it is determined that the thrown object is not a true high-altitude parabolic object, and the corresponding parabolic event Not for parabolic events.

Among them, when the angle between the straight line and the vertical direction after fitting the motion event to be measured is greater than or equal to the preset angle threshold, it is determined that the thrown object does not belong to a high-altitude thrown object; or, at the center point of the thrown object When the accumulated pixel value is greater than or equal to the preset pixel threshold, it is determined that the thrown object does not belong to a high-altitude thrown object; or when the height difference between the lowest point and the highest point of the parabolic event is less than or equal to the preset height threshold , to determine that the dropped object is not a high-altitude dropped object. Alternatively, when the angle between the straight line and the vertical direction after the fitting of the motion event to be measured is greater than or equal to the preset angle threshold and the accumulated value of the pixel value at the center point of the dropped object is greater than or equal to the preset pixel threshold, determine whether the throwing The falling object is not a high-altitude throwing object; or the angle between the straight line and the vertical direction after fitting the motion event to be measured is greater than or equal to the preset angle threshold and the height difference between the lowest point and the highest point of the parabolic event is less than or If it is equal to the preset height threshold, it is determined that the dropped object does not belong to a high-altitude dropped object; or the cumulative value of the pixel value at the center point of the dropped object is greater than or equal to the preset pixel threshold and the height of the lowest point and the highest point of the parabolic event When the difference is less than or equal to the preset height threshold, it is determined that the dropped object does not belong to a high-altitude thrown object; or the angle between the straight line and the vertical direction is greater than or equal to the preset angle threshold after the motion event to be measured is fitted and If the cumulative value of the pixel value at the center point of the dropped object is greater than or equal to the preset pixel threshold and the height difference between the lowest point and the highest point of the parabolic event is less than or equal to the preset height threshold, it is determined that the dropped object does not belong to high-altitude throwing Object, the corresponding parabolic event is not a high-altitude parabolic event.

It can be seen that in the embodiments of the present disclosure, the event-based detection method can realize accurate determination of high-altitude parabolic events based on the motion trajectory after the trajectory is completely restored by combining the detection information of the dropped object on multiple frames.

Based on the above-mentioned embodiments, in yet another implementation of the embodiments of the present disclosure, the method for performing high-altitude parabolic detection processing mainly includes four parts: data preprocessing, dense optical flow calculation, single-frame post-processing and multi-frame post-processing.

Among them, data preprocessing specifically includes the following steps:

S201. Acquire an initial image, and determine a target detection area from the initial image based on a preset polygonal outline.

S202. Generate a minimum detection frame based on the target detection area, and perform image segmentation processing on the initial image based on the minimum detection frame to obtain an image to be tested.

Among them, the dense optical flow calculation specifically includes the following steps:

S203. Read two adjacent images to be tested sequentially in each cycle of the execution cycle, and generate optical flow images corresponding to the two adjacent images to be tested by using a preset optical flow model.

Wherein, the single frame post-processing specifically includes the following steps:

S204. Perform single-channel grayscale conversion processing on the optical flow image to obtain a single-channel grayscale image corresponding to the optical flow image.

S205. Perform normalization processing on the single-channel grayscale image to obtain a normalized grayscale image corresponding to the optical flow image.

S206. Perform binarization processing on the normalized grayscale image to obtain a binary image corresponding to the optical flow image; wherein, the binary image includes a foreground moving object with a first pixel value and a background non-moving object with a second pixel value .

S207. Determine whether the pixel proportion of the foreground moving object in the binary image is less than or equal to a preset proportion threshold? If yes, execute S208; if not, skip to execute S204, and continue to process the next frame of optical flow image.

Wherein, in the case that the pixel ratio of the foreground moving object in the binary image is less than or equal to the preset ratio threshold, it is determined that the current optical flow image is the optical flow image to be measured with the dropped object; otherwise, it does not exist.

S208. Obtain a set of initial position coordinates of the foreground moving object in the binary image.

S209. Classify the foreground moving objects based on the preset clustering algorithm and the initial position coordinate set, and obtain the coordinate subsets corresponding to each moving object.

S210. For any moving object, calculate the coordinate average value corresponding to the coordinate subset of any moving object, and determine the coordinate average value as the center point position coordinate of any moving object.

S211. Determine the target pixel value at the coordinates of the center point of any moving object in the normalized grayscale image corresponding to any optical flow image to be measured, and determine the moving object with the largest target pixel value as the thrown object , and the center point position coordinates corresponding to the moving object with the largest target pixel value are determined as the center point position coordinates of the dropped object.

Wherein, multi-frame post-processing includes the following steps:

S213. Calculate the frame interval value between the current optical flow image to be measured and the previous optical flow image to be measured in history.

S214. Determine whether the frame interval value is less than or equal to a preset interval threshold? If yes, execute S215; if not, skip to execute S213.

Among them, if the frame interval value between the current optical flow image to be measured and the previous optical flow image to be measured in history is less than or equal to the preset interval threshold, it indicates that the current optical flow image to be measured and the previous optical flow image in history belong to the same parabolic event , update the end frame of the parabolic event based on the current optical flow image to be measured; otherwise, if it does not belong to the same parabolic event, the current optical flow image to be measured is the start frame of the new parabolic event.

S215. Determine whether the start frame to the end frame of the parabolic event is determined? If yes, execute S216; if not, skip to execute S213, and continue the loop process from S213 to S214 until the start frame to end frame of the parabolic event is obtained.

S216 , according to the frame sequence from the first frame to the end frame of the parabolic event, perform interpolation processing on the coordinates of the center point of the dropped object in each optical flow image to be measured, to obtain the trajectory of the dropped object.

S217. Fitting a straight line to the trajectory of the dropped object, calculating the angle θ between the fitted straight line and the vertical direction; summing the gray values of the pixels at the central point of the dropped object on each frame of the parabolic event to obtain the pixel Accumulated value V; Calculate the height difference between the lowest point and the highest point of the parabolic event, that is, the parabolic height H;

S218. Determine whether θ is smaller than threshold T ₁ , whether V is smaller than threshold T ₂ , and whether H is greater than threshold T ₃ ; if yes, execute S219 ; if not, skip to execute S220 .

Among them, T ₁ is the preset angle threshold, T ₂ is the preset pixel threshold, and T ₃ is the preset height threshold.

S219, the dropped object is a high-altitude thrown object, and the parabolic event is a high-altitude parabolic event.

Among them, after the parabolic event is fitted, the angle between the straight line and the vertical direction is less than the preset angle threshold, the accumulated value of the pixel value at the center point of the dropped object is less than the preset pixel threshold, and the height of the lowest point and the highest point of the parabolic event The difference is greater than the preset height threshold, that is, when the three judging factors meet the corresponding preset threshold conditions at the same time, it is determined that the dropped object is a high-altitude thrown object, and the parabolic event is a high-altitude parabolic event.

S220. The dropped object is not a high-altitude thrown object, and the parabolic event is not a high-altitude parabolic event.

Wherein, after the parabolic event is fitted, the angle between the straight line and the vertical direction is greater than or equal to the preset angle threshold, or, the accumulated value of the pixel value at the center point of the thrown object is greater than or equal to the preset pixel threshold, or, the parabolic event The height difference between the lowest point and the highest point is less than or equal to the preset height threshold, that is, when at least one of the three judgment factors does not meet the corresponding preset threshold condition, it is determined that the thrown object is not a high-altitude thrown object, Parabolic events are not high-altitude parabolic events.

FIG. 11 is a schematic diagram of the scene of the high-altitude parabolic detection method proposed by the embodiment of the present disclosure. As shown in FIG. 11 , in the process of high-altitude parabolic detection for buildings, the image is first collected by the camera to obtain the initial image, and then through preprocessing, Including determining the region of interest based on the preset polygonal outline, and cropping the image based on the minimum frame to obtain the nth frame of the image to be tested and the (n+1)th frame of the image to be tested; then input the adjacent two frames of the image to be tested by the volume The preset optical flow model constructed by the product neural network and the recurrent neural network is used to obtain a dense optical flow map; continue to perform single-frame post-processing on each frame of optical flow image, including checking whether there are falling objects in the image based on image binarization Judgment, and determine the position coordinates of the center point of the thrown object based on position clustering for the optical flow image to be measured with the presence of the dropped object; loop the above process to obtain multiple frames of the optical flow image to be measured, and perform multiple frames of the optical flow to be measured The image is post-processed in multiple frames until the start frame to the end frame of the parabolic event corresponding to the dropped object is obtained, and the motion track of the dropped object is restored, and based on the motion track, whether the dropped object is a high-altitude dropped object is detected. When the dropped object is a high-altitude throwing object, that is, the corresponding throwing time is a high-altitude parabolic event, and a time alarm is performed for the high-altitude parabolic event.

It can be seen that the high-altitude parabolic detection method proposed in the embodiment of the present disclosure, on the one hand, uses the optical flow model constructed based on deep learning to generate a dense optical flow map to detect moving objects, which is not only robust and accurate, but also time-consuming Shorter and less noise; on the other hand, on the basis of using the preset optical flow model for moving object detection, a robust single-frame plus multi-frame post-processing method is also proposed, based on image binarization and position The clustering algorithm filters the single frame with disturbing objects and obtains the accurate position of the dropped object, and then combines the dropped object detection information on multiple frames to recover and detect the trajectory of the high-altitude parabolic event, which further improves the high-altitude parabolic detection. efficiency and precision.

Based on the above-mentioned embodiments, in an embodiment of the present disclosure, FIG. 12 is a schematic diagram of the composition and structure of the high-altitude parabolic detection device proposed by the embodiment of the present disclosure. As shown in FIG. 12 , the high-altitude parabolic detection device 10 includes an acquisition part 11 , a generation part 12 , a determination part 13 , and a processing part 14 .

The acquisition part 11 is configured to sequentially read two adjacent images to be tested from multiple frames of images to be tested according to the time sequence of image acquisition;

The generating part 12 is configured to generate optical flow images corresponding to the two adjacent images to be tested through a preset optical flow model;

The determining part 13 is configured to determine from the optical flow images the optical flow images to be measured that there is a thrown object, and determine the position coordinates of the center point of the thrown object in each optical flow image to be measured;

The determining part 13 is further configured to determine the trajectory of the dropped object according to the position coordinates of the center point of the dropped object in each optical flow image to be measured;

The processing part 14 is configured to perform high-altitude parabolic detection processing based on the motion trajectory.

In some embodiments, the determining part 13 is configured to, for any optical flow image, generate a binary image corresponding to any optical flow image; wherein, the binary image includes the first pixel value of the foreground moving object and the second The background non-moving object of the pixel value; and in response to the pixel ratio of the foreground moving object in the binary image being less than or equal to a preset ratio threshold, determining that any optical flow image is the existence of the throwing object The optical flow image to be measured.

In some embodiments, the determining part 13 is configured to perform single-channel grayscale conversion processing on any optical flow image to obtain a single-channel grayscale image corresponding to any optical flow image; performing normalization processing on the image to obtain a normalized grayscale image corresponding to any optical flow image; and performing binarization processing on the normalized grayscale image to obtain the grayscale image corresponding to any optical flow image Binary image.

In some embodiments, the determination part 13 is configured to obtain, for any optical flow image to be measured, a set of initial position coordinates of the foreground moving object in the binary image, and based on a preset clustering algorithm and The set of initial position coordinates classifies the foreground moving objects in any of the optical flow images to be measured, and obtains at least one moving object and respective coordinate subsets corresponding to each moving object; and for any moving object, Calculating the average value of coordinates corresponding to the coordinate subset of any moving object, and determining the average value of coordinates as the position coordinates of the center point of any moving object; and determining any optical flow to be measured In the normalized grayscale image corresponding to the image, the target pixel value at the center point position coordinates of any moving object, and the moving object with the largest target pixel value is determined as the thrown object, and the The position coordinates of the center point corresponding to the moving object with the largest target pixel value are determined as the position coordinates of the center point of the dropped object.

In some embodiments, the determining part 13 is configured to determine the trajectory of the thrown object according to the center point position coordinates of the thrown object in the optical flow images to be measured, In the optical flow image, determine the optical flow image to be measured from the start to the optical flow image to be measured of the parabolic event corresponding to the dropped object.

In some embodiments, the determining part 13 is configured to interpolate the position coordinates of the center point of the thrown object in each optical flow image to be measured in the parabolic event in chronological order, to obtain the The trajectory of the falling object.

In some embodiments, the optical flow image to be measured includes at least a first image and a second image, the first image is the optical flow image to be measured at the current end of the first parabolic event; the first image and the second The images are two consecutive frames of optical flow images to be measured, and the determining part 13 is configured to calculate a frame interval value between the second image and the first image; and in response to the frame interval value being less than or equal to a preset Setting an interval threshold, updating the second image to the optical flow image to be measured at the end of the first parabolic event; in response to the frame interval value being greater than the preset interval threshold, determining the second image is the initial optical flow image to be measured for the second parabolic event.

In some embodiments, the processing part 14 is configured to perform a straight line fitting process on the motion trajectory of the thrown object, obtain a fitted straight line corresponding to the motion trajectory, and determine the relationship between the fitted straight line and The included angle in the vertical direction; and based on the center point position coordinates of the thrown object in each optical flow image to be measured in the parabolic event, determine the center point position pixel value from the corresponding normalized grayscale image , and perform cumulative processing on the pixel values of each central point position to obtain the pixel cumulative value; In the optical flow image to be measured, the position coordinates of the central point of the dropped object are subjected to a difference operation of the ordinate to obtain a coordinate difference; and based on the angle between the fitted straight line and the vertical direction, the pixel cumulative value And the coordinate difference value executes the high-altitude parabola detection process.

In some embodiments, the processing part 14 is configured to respond to the angle between the fitted straight line and the vertical direction being smaller than a preset angle threshold, the pixel accumulation value being smaller than the preset pixel threshold, and the If the coordinate difference is greater than a preset height threshold, it is determined that the dropped object is a high-altitude projectile, and the corresponding parabolic event is a high-altitude parabolic event.

In some embodiments, the acquisition part 11 is further configured to acquire an initial image, and determine a target detection area from the initial image based on a preset polygonal outline; and generate a minimum detection frame based on the target detection area, and based on The minimum detection frame performs image segmentation processing on the initial image to obtain the image to be tested.

In the embodiment of the present disclosure, further, FIG. 13 is a schematic diagram of the composition and structure of the high-altitude parabolic detection device proposed in the embodiment of the present disclosure. As shown in FIG. 13 , the high-altitude parabolic detection device 20 proposed in the embodiment of the present disclosure may also include processing processor 21, a memory 22 storing instructions executable by the processor 21, further, the living body detection device 20 may also include a communication interface 23, and a bus 24 for connecting the processor 21, the memory 22 and the communication interface 23.

In an embodiment of the present disclosure, the above-mentioned processor 21 may be an application specific integrated circuit (Application Specific Integrated Circuit, ASIC), a digital signal processor (Digital Signal Processor, DSP), a digital signal processing device (Digital Signal Processing Device, DSPD ), Programmable Logic Device (ProgRAMmable Logic Device, PLD), Field Programmable Gate Array (Field Prog RAMmable Gate Array, FPGA), Central Processing Unit (Central Processing Unit, CPU), controller, microcontroller, microprocessor at least one of the It can be understood that, for different devices, the electronic device used to implement the above processor function may also be other, which is not specifically limited in this embodiment of the present disclosure. The living body detection device 20 may also include a memory 22, which may be connected to the processor 21, wherein the memory 22 is used to store executable program codes, the program codes include computer operation instructions, and the memory 22 may include a high-speed RAM memory, or may Also included is non-volatile memory, eg, at least two disk memories.

In the embodiment of the present disclosure, the bus 24 is used to connect the communication interface 23 , the processor 21 and the memory 22 and communicate with each other among these devices.

In an embodiment of the present disclosure, the memory 22 is used to store instructions and data.

Further, in the embodiment of the present disclosure, the above-mentioned processor 21 is configured to sequentially read two adjacent images to be tested from multiple frames of images to be tested according to the time sequence of image acquisition, and use the preset optical flow The model generates the optical flow images corresponding to the two adjacent images to be tested; from the optical flow images, it is determined that there is an optical flow image to be measured with a thrown object, and it is determined that in each optical flow image to be measured, the The coordinates of the center point of the object; determining the motion track of the dropped object according to the coordinates of the center point of the dropped object in the optical flow images to be measured; and performing high-altitude parabolic detection processing based on the motion track.

In practical applications, the above-mentioned memory 22 can be a volatile memory (volatile memory), such as a random access memory (Random-Access Memory, RAM); or a non-volatile memory (non-volatile memory), such as a read-only memory (Read-Only Memory, ROM), flash memory (flash memory), hard disk (Hard Disk Drive, HDD) or solid-state hard drive (Solid-State Drive, SSD); Provide instructions and data.

In addition, each functional module in this embodiment may be integrated into one recommendation unit, each unit may exist separately physically, or two or more units may be integrated into one unit. The above-mentioned integrated units can be implemented in the form of hardware or in the form of software function modules.

If the integrated unit is implemented in the form of a software function module and is not sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this embodiment is essentially or Part of the prior art contribution or all or part of the technical solution can be embodied in the form of software products, the computer software products are stored in a storage medium, including a number of instructions to make a high-altitude parabolic detection equipment (can It is a personal computer, a server, or a network device, etc.) or a processor (processor) that executes all or part of the steps of the method of this embodiment. The aforementioned storage medium includes: various media capable of storing program codes such as U disk, mobile hard disk, ROM, RAM, magnetic disk or optical disk.

An embodiment of the present disclosure provides a high-altitude parabolic detection device. The high-altitude parabolic detection device sequentially reads two adjacent images to be tested from multiple frames of images to be tested according to the time sequence of image acquisition, and uses preset light The flow model generates the optical flow images corresponding to the two adjacent images to be tested; from the optical flow images, determine the optical flow images to be tested that have thrown objects, and determine the position of the center point of the dropped objects in each optical flow image to be tested Coordinates; determine the motion track of the dropped object according to the position coordinates of the center point of the dropped object in each optical flow image to be measured; perform high-altitude parabolic detection processing based on the motion track. In this way, on the one hand, the present disclosure uses an optical flow model constructed based on deep learning to generate a dense optical flow map to detect moving objects, which not only has good robustness and high precision, but also takes less time and has less noise; On the one hand, on the basis of using the preset optical flow model for moving object detection, this disclosure also proposes a robust single-frame plus multi-frame post-processing method, which filters the single frame with disturbing objects to find out the For the single-frame image of the falling object, after removing the single-frame false detection, combined with the position information of the falling object on multiple frames for trajectory recovery and high-altitude parabolic event detection, which further improves the efficiency and accuracy of high-altitude parabolic detection.

An embodiment of the present disclosure provides a computer-readable storage medium on which a program is stored, and when the program is executed by a processor, the above-mentioned high-altitude parabolic detection method is implemented.

Specifically, the program instructions corresponding to a high-altitude parabolic detection method in this embodiment can be stored on a storage medium such as an optical disc, a hard disk, and a USB flash drive. When the program instructions corresponding to a high-altitude parabolic detection method in the storage medium When read or executed by an electronic device, the following steps are included:

sequentially reading two adjacent images to be tested from multiple frames of images to be tested according to the time sequence of image acquisition, and generating optical flow images corresponding to the two adjacent images to be tested through a preset optical flow model;

Determining the optical flow images to be measured in which there are dropped objects from the optical flow images, and determining the position coordinates of the center points of the dropped objects in each optical flow image to be measured;

determining the trajectory of the dropped object according to the position coordinates of the center point of the dropped object in each optical flow image to be measured;

A high-altitude parabolic detection process is performed based on the motion trajectory.

Correspondingly, an embodiment of the present disclosure further provides a computer program product, where the computer program product includes computer-executable instructions, and the computer-executable instructions are used to implement the steps in the high-altitude parabolic detection method proposed by the embodiments of the present disclosure.

Those skilled in the art should understand that the embodiments of the present disclosure may be provided as methods, systems, or computer program products. Accordingly, the present disclosure may take the form of a hardware embodiment, a software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present disclosure may take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, optical storage, etc.) having computer-usable program code embodied therein.

The present disclosure is described with reference to the implementation flow diagrams and/or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present disclosure. It should be understood that each process and/or block in the schematic flowchart and/or block diagram, and a combination of processes and/or blocks in the schematic flowchart and/or block diagram can be realized by computer program instructions. These computer program instructions can be provided to a general purpose computer, a special purpose computer, an embedded processor or a processor of other programmable parabolic detection equipment to produce a machine, so that the instructions executed by the computer or other programmable parabolic detection equipment processor Means for realizing the functions specified in one or more steps of the flowchart and/or one or more blocks of the block diagram are produced.

These computer program instructions may also be stored in a computer readable memory capable of directing a computer or other programmable parabolic detection device to operate in a specific manner, such that the instructions stored in the computer readable memory produce an article of manufacture comprising instruction means, the The instruction means implements the functions specified in implementing one or more procedures of the flowchart and/or one or more blocks of the block diagram.

These computer program instructions can also be loaded on a computer or other programmable parabolic detection equipment, so that a series of operation steps are performed on the computer or other programmable equipment to generate computer-implemented processing, thereby executing on the computer or other programmable equipment The instructions provide steps for implementing the functions specified in one or more processes of the flowchart diagrams and/or one or more blocks of the block diagrams.

The above descriptions are only preferred embodiments of the present disclosure, and are not intended to limit the protection scope of the present disclosure.

Industrial Applicability

In the embodiment of the present disclosure, by sequentially reading two adjacent images to be tested from multiple frames of images to be tested according to the time sequence of image acquisition, and using a preset optical flow model to generate the corresponding two adjacent images to be tested The optical flow image of the object; from the optical flow image, determine the optical flow image of the object to be measured, and determine the coordinates of the center point of the object in each optical flow image to be measured; according to the optical flow image of each object to be measured The position coordinates of the center point of the object determine the motion track of the dropped object; the high-altitude parabolic detection process is performed based on the motion track. The above scheme further improves the efficiency and accuracy of high-altitude parabolic detection.

Claims

A high-altitude parabolic detection method, said method comprising:

sequentially reading two adjacent images to be tested from multiple frames of images to be tested according to the time sequence of image acquisition, and generating optical flow images corresponding to the two adjacent images to be tested through a preset optical flow model;

Determining the optical flow images to be measured in which there are dropped objects from the optical flow images, and determining the position coordinates of the center points of the dropped objects in each optical flow image to be measured;

determining the trajectory of the dropped object according to the position coordinates of the center point of the dropped object in each optical flow image to be measured;

A high-altitude parabolic detection process is performed based on the motion trajectory.
The method according to claim 1, wherein said determining from said optical flow image that there is an optical flow image to be measured of a thrown object comprises:

For any optical flow image, generate a binary image corresponding to any optical flow image; wherein, the binary image includes a foreground moving object with a first pixel value and a background non-moving object with a second pixel value;

In response to the pixel proportion of the foreground moving object in the binary image being less than or equal to a preset proportion threshold, it is determined that any optical flow image is the optical flow image to be tested in which the falling object exists.
The method according to claim 2, wherein said generating a binary image corresponding to any optical flow image comprises:

performing single-channel grayscale conversion processing on any optical flow image to obtain a single-channel grayscale image corresponding to any optical flow image;

Performing normalization processing on the single-channel grayscale image to obtain a normalized grayscale image corresponding to any optical flow image;

Binarizing the normalized grayscale image to obtain the binary image corresponding to any optical flow image.
The method according to claim 3, wherein said determining the position coordinates of the center point of said dropped object in each optical flow image to be measured comprises:

For any optical flow image to be measured, the initial position coordinate set of the foreground moving object in the binary image is obtained, and based on the preset clustering algorithm and the initial position coordinate set, the Classifying the foreground moving objects in the stream image to obtain at least one moving object and coordinate subsets corresponding to each moving object;

For any moving object, calculate an average value of coordinates corresponding to the coordinate subset of any moving object, and determine the average value of coordinates as the position coordinates of the center point of any moving object;

Determine the target pixel value at the coordinates of the center point of any moving object in the normalized grayscale image corresponding to any of the optical flow images to be measured, and determine the moving object with the largest target pixel value The center point position coordinates corresponding to the dropped object and the moving object with the largest target pixel value are determined as the center point position coordinates of the dropped object.
The method according to claim 3, wherein, before determining the trajectory of the dropped object according to the position coordinates of the center point of the dropped object in each optical flow image to be measured, the method further comprises:

From the optical flow images to be measured, determine the optical flow image to be measured from the start to the optical flow image to be measured of the parabolic event corresponding to the dropped object;

The determining the trajectory of the dropped object according to the position coordinates of the center point of the dropped object in each optical flow image to be measured includes:

Interpolation processing is performed on the position coordinates of the center point of the dropped object in each optical flow image to be measured in the parabolic event in chronological order to obtain the trajectory of the dropped object.
The method according to claim 5, wherein the optical flow image to be measured comprises at least a first image and a second image, and the first image is the current end optical flow image to be measured of the first parabolic event; the first image and the second image is two consecutive frames of optical flow images to be measured;

The determining from the optical flow images to be measured from the start optical flow image to the optical flow image to be measured of the parabolic event corresponding to the thrown object to the end optical flow image to be measured includes:

calculating a frame interval value between the second image and the first image;

In response to the frame interval value being less than or equal to a preset interval threshold, updating the second image to the end-to-be-measured optical flow image of the first parabolic event;

In response to the frame interval value being greater than the preset interval threshold, the second image is determined as the initial optical flow image to be measured of the second parabolic event.
The method according to claim 5 or 6, wherein said performing high-altitude parabolic detection processing based on said motion trajectory comprises:

performing a straight line fitting process on the trajectory of the dropped object, obtaining a fitted straight line corresponding to the motion trajectory, and determining an angle between the fitted straight line and the vertical direction;

Based on the position coordinates of the center point of the dropped object in each optical flow image to be measured in the parabolic event, the pixel value of the center point position is determined from the corresponding normalized grayscale image, and the pixel value of the center point position is determined for each center point position pixel The values are accumulated and processed to obtain the pixel accumulated value;

In the parabolic event, the coordinates of the center point position of the dropped object in the initial optical flow image to be measured and the coordinates of the center point position of the dropped object in the end optical flow image to be measured are ordinated The difference operation to obtain the coordinate difference;

The high-altitude parabola detection process is performed based on the angle between the fitted straight line and the vertical direction, the pixel accumulation value and the coordinate difference value.
The method according to claim 7, wherein, performing the high-altitude parabola detection process based on the angle between the fitted straight line and the vertical direction, the pixel accumulation value and the coordinate difference value includes:

In response to the angle between the fitted straight line and the vertical direction being less than a preset angle threshold, the pixel accumulation value being less than the preset pixel threshold, and the coordinate difference being greater than a preset height threshold, determining that the thrown The falling object is a high-altitude throwing object, and the corresponding parabolic event is a high-altitude parabolic event.
The method according to any one of claims 1 to 8, wherein the method further comprises:

Acquiring an initial image, and determining a target detection area from the initial image based on a preset polygonal outline;

A minimum detection frame is generated based on the target detection area, and image segmentation processing is performed on the initial image based on the minimum detection frame to obtain the image to be tested.
A high-altitude parabolic detection device, the high-altitude parabolic detection device includes:

The reading part is configured to sequentially read two adjacent images to be tested from multiple frames of images to be tested according to the time sequence of image acquisition;

The generating part is configured to generate the optical flow images corresponding to the two adjacent images to be tested through a preset optical flow model;

The determining part is configured to determine from the optical flow images the optical flow images to be measured that there is a thrown object, and determine the position coordinates of the center point of the thrown object in each optical flow image to be measured;

The determining part is further configured to determine the trajectory of the dropped object according to the position coordinates of the center point of the dropped object in each optical flow image to be measured;

The processing part is configured to perform high-altitude parabolic detection processing based on the motion trajectory.
The high-altitude parabolic detection device according to claim 10, wherein,

The determining part is further configured to generate a binary image corresponding to any optical flow image for any optical flow image; wherein, the binary image includes a foreground moving object with a first pixel value and a non-moving background with a second pixel value an object; and in response to a pixel ratio of the foreground moving object in the binary image being less than or equal to a preset ratio threshold, determining that any optical flow image is the optical flow image to be measured where there is the falling object.
The high-altitude parabolic detection device according to claim 11, wherein,

The determining part is further configured to perform single-channel grayscale conversion processing on any optical flow image to obtain a single-channel grayscale image corresponding to any optical flow image; and perform normalization processing on the single-channel grayscale image , obtaining a normalized grayscale image corresponding to any optical flow image; and performing binarization processing on the normalized grayscale image to obtain the binary image corresponding to any optical flow image.
The high-altitude parabolic detection device according to claim 12, wherein,

The determining part is further configured to obtain, for any optical flow image to be measured, a set of initial position coordinates of the foreground moving object in the binary image, and based on a preset clustering algorithm and the set of initial position coordinates Classifying the foreground moving objects in any of the optical flow images to be measured to obtain at least one moving object and respective coordinate subsets corresponding to each moving object; and for any moving object, calculating any of the moving objects The coordinate average value corresponding to the coordinate subset of the object, and determining the coordinate average value as the position coordinate of the center point of any moving object; and determining the normalization corresponding to any optical flow image to be measured In the post-grayscale image, the target pixel value at the coordinates of the center point of any moving object, and determine the moving object with the largest target pixel value as the dropped object, and the moving object with the largest target pixel value The position coordinates of the center point corresponding to the moving object are determined as the position coordinates of the center point of the dropped object.
The high-altitude parabolic detection device according to claim 12, wherein,

The determining part is further configured to, from the optical flow images to be measured, before determining the trajectory of the dropped object according to the position coordinates of the center point of the dropped object in the optical flow images to be measured, determining the optical flow image to be measured from the start to the end of the optical flow image to be measured of the parabolic event corresponding to the dropped object;

The determining part is further configured to interpolate the position coordinates of the center point of the dropped object in each optical flow image to be measured in the parabolic event in chronological order, so as to obtain the trajectory of the dropped object.
The high-altitude parabolic detection device according to claim 14, wherein the optical flow image to be measured includes at least a first image and a second image, and the first image is the optical flow image to be measured at the current end of the first parabolic event; The first image and the second image are two consecutive frames of optical flow images to be measured;

The determining part is configured to calculate a frame interval value between the second image and the first image; and in response to the frame interval value being less than or equal to a preset interval threshold, update the second image as The end optical flow image to be measured of the first parabolic event; and in response to the frame interval value being greater than the preset interval threshold, determining the second image as the initial optical flow to be measured for a second parabolic event stream images.
The high-altitude parabolic detection device according to claim 14 or 15, wherein,

The processing part is further configured to perform a straight line fitting process on the motion trajectory of the dropped object, obtain a fitted straight line corresponding to the motion trajectory, and determine the angle between the fitted straight line and the vertical direction ; and based on the center point position coordinates of the thrown object in each optical flow image to be measured in the parabolic event, determine the center point position pixel value from the corresponding normalized grayscale image, and calculate each center point The position pixel value is accumulated and processed to obtain the pixel accumulation value; and in the parabolic event, the position coordinates of the center point of the thrown object in the initial optical flow image to be measured and the position coordinates of the center point of the object in the optical flow image to be measured at the end The center point position coordinates of the dropped object are subjected to a difference operation of the ordinate to obtain a coordinate difference; and based on the angle between the fitted straight line and the vertical direction, the pixel accumulation value and the coordinate difference The high-altitude parabolic detection process is performed.
The high-altitude parabolic detection device according to claim 16, wherein,

The processing part is further configured to respond to the fact that the angle between the fitted straight line and the vertical direction is smaller than a preset angle threshold, the pixel accumulation value is smaller than the preset pixel threshold, and the coordinate difference is larger than a preset A height threshold is set to determine that the dropped object is a high-altitude projectile, and the corresponding parabolic event is a high-altitude projectile event.
The high-altitude parabolic detection device according to any one of claims 10 to 17, wherein,

The acquisition part 11 is further configured to acquire an initial image, and determine a target detection area from the initial image based on a preset polygonal outline; and generate a minimum detection frame based on the target detection area, and generate a minimum detection frame based on the minimum detection frame The initial image is subjected to image segmentation processing to obtain the image to be tested.
A high-altitude parabolic detection device, the high-altitude parabolic detection device includes a processor, a memory storing instructions executable by the processor, and when the instructions are executed by the processor, any one of claims 1 to 9 can be realized. method described in the item.
A computer-readable storage medium, on which a program is stored, which is applied to a high-altitude parabolic detection device, and when the program is executed by a processor, the method according to any one of claims 1 to 9 is realized.
A computer program, comprising computer readable code, when the computer readable code runs in an electronic device and is executed by a processor in the electronic device, it realizes any one of claims 1 to 9 High-altitude parabolic detection method.
A computer program product which, when run on a computer, causes the computer to execute the method for detecting high-altitude parabolic objects according to any one of claims 1 to 9.