CN113723176B - A target object determination method, device, storage medium and electronic device - Google Patents

Abstract

Embodiments of the present invention provide a target object determination method, device, storage medium and electronic device, which relate to the technical field of object recognition technology. The method includes: acquiring target images of consecutive frames; performing object type filtering processing on the target image to obtain a target processing area; performing regional feature filtering processing on the target processing area to obtain a first recognition area; The motion feature condition of the first recognition area is subjected to object motion feature detection processing, and the initial object that satisfies the motion feature condition contained in the first recognition area is used as the first object; Difference recognition processing; in the case that the processing result is that the first object satisfies the pixel difference condition, determining the first object as the target object. Through the present invention, the problem of low object recognition accuracy is solved, thereby achieving the effect of improving the object recognition accuracy and recognition efficiency.

Description

A target object determination method, device, storage medium and electronic device

technical field

Embodiments of the present invention relate to the field of image recognition, and in particular, to a target object determination method, device, storage medium, and electronic device.

Background technique

With the development of my country's economy, target object recognition technology has been applied to many fields. However, the current target object recognition technology generally recognizes the target object in the target area through pixel recognition or trajectory prediction. The method will affect the recognition result of the target object when the difference between the pixels of the target object and the pixels of the surrounding environment is small, or the target object is in a static state, or the environmental noise and illumination where the target object is located are strong.

For example, in the field of traffic recognition, most of the existing detection methods of the intelligent detection device for thrown objects rely on the front-background separation technology to capture the foreground, and then set a time threshold to determine whether the thrown objects are generated. When the pixels between vehicles or between vehicles and the environment are close to each other, it is prone to false detection of vehicles, or it is easily affected by light reflection and noise, which can easily lead to false foregrounds, resulting in the identification of projectiles that cannot be detected in the static state. problem of low precision.

At present, there is no better solution to the above problems.

SUMMARY OF THE INVENTION

Embodiments of the present invention provide a target object recognition method, device, storage medium, and electronic device, so as to at least solve the problem of low target object recognition accuracy in the related art.

According to an embodiment of the present invention, a method for determining a target object is provided, including:

Get the target image of consecutive frames;

Performing object type filtering processing on the target image to obtain a target processing area;

Performing regional feature filtering processing on the target processing area to obtain a first recognition area, wherein the target processing area includes the first recognition area;

Based on the preset motion feature conditions, object motion feature detection processing is performed on the first recognition area, and the initial object included in the first recognition area that satisfies the motion feature conditions is used as the first object;

performing pixel difference recognition processing on the first object;

When the processing result is that the first object satisfies the pixel difference condition, the first object is determined as the target object.

In an exemplary embodiment, performing object type filtering processing on the target image to obtain the target processing area includes:

Performing environment type segmentation processing on the target image to obtain a first processing area;

Performing object feature recognition processing on the target image to obtain a second processing area;

Based on the second processing area, a region coincidence filtering process is performed on the first processing area to obtain the target processing area.

In an exemplary embodiment, performing an area coincidence filtering process on the first processing area based on the second processing area to obtain the target processing area includes:

performing overlap detection processing on the first processing region and the second processing region to obtain a region coincidence value;

In the case of determining that the area coincidence value is less than or equal to a preset value, determine that the first processing area is the target processing area;

When it is determined that the area coincidence value is greater than the preset value, a first filtering process is performed on the first processing area.

In an exemplary embodiment, the object motion feature detection process is performed on the first recognition region based on the preset motion feature conditions, and the initial objects included in the first recognition region that meet the motion feature conditions are As the first object includes:

performing motion state detection processing on the initial object;

In the case of determining that the initial object is in a stationary state, perform a stationary duration detection process on the initial object;

When it is determined that the stationary duration of the initial object satisfies a threshold, the initial object is used as the first object.

In an exemplary embodiment, the performing region feature filtering processing on the target processing region to obtain the first identification region includes:

Performing area feature detection processing on the target processing area to obtain area feature information of the target processing area;

Perform area feature comparison processing on the area feature information, and in the case that the area feature information is determined to meet the area threshold, the target processing area is used as the first identification area; when it is determined that the area feature information does not satisfy the area In the case of the threshold value, the second filtering process is performed on the target processing area.

In an exemplary embodiment, before the region feature filtering process is performed on the target processing region to obtain the first identification region, the method further includes:

Get the target image of consecutive frames;

Perform image object detection processing on the target image to obtain the target processing area.

In an exemplary embodiment, when the processing result is that the first object satisfies the pixel difference condition, after determining the first object as the target object, the method further includes:

The target object is stored in the target storage area.

According to another embodiment of the present invention, an apparatus for determining a target object is provided, comprising:

The image acquisition module is used to acquire target images of consecutive frames;

an object type filtering module for performing object type filtering processing on the target image to obtain a target processing area;

a regional feature filtering module, configured to perform regional feature filtering processing on the target processing region to obtain a first recognition region, wherein the target processing region includes the first recognition region;

A motion feature detection module, configured to perform object motion feature detection processing on the first recognition area based on preset motion feature conditions, and use the initial object included in the first recognition area that satisfies the motion feature conditions as the first object;

a pixel difference recognition module, configured to perform pixel difference recognition processing on the first object;

A target object determination module, configured to determine the first object as the target object when the processing result is that the first object satisfies the pixel difference condition.

According to yet another embodiment of the present invention, a computer-readable storage medium is also provided, and a computer program is stored in the computer-readable storage medium, wherein the computer program is configured to execute any one of the above methods when running steps in the examples.

According to yet another embodiment of the present invention, there is also provided an electronic device comprising a memory and a processor, wherein the memory stores a computer program, the processor is configured to run the computer program to execute any of the above Steps in Method Examples.

According to the present invention, since the object type processing is performed on the images of consecutive frames first, the interference factors in the process of target object recognition are reduced, and the area feature filtering processing performed subsequently can perform the target processing area that does not meet the requirements. Filtering, thereby reducing the amount of calculation in the recognition process, and performing object motion feature detection processing can avoid false recognition of stationary objects and improve the accuracy of recognition. Therefore, the problem of low recognition accuracy of the target object can be solved, and the effect of improving the recognition accuracy and recognition efficiency of the target object can be achieved.

Description of drawings

1 is a block diagram of a hardware structure of a mobile terminal of a method for determining a target object according to an embodiment of the present invention;

2 is a flowchart of a method for determining a target object according to an embodiment of the present invention;

3 is a structural block diagram of an apparatus for determining a target object according to an embodiment of the present invention;

Fig. 4 is a flow chart according to a specific embodiment of the present invention;

FIG. 5 is a schematic structural diagram according to a specific embodiment of the present invention.

Detailed ways

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings and in conjunction with the embodiments.

It should be noted that the terms "first", "second" and the like in the description and claims of the present invention and the above drawings are used to distinguish similar objects, and are not necessarily used to describe a specific sequence or sequence.

The method embodiments provided in the embodiments of the present application may be executed in a mobile terminal, a computer terminal, or a similar computing device. Taking running on a mobile terminal as an example, FIG. 1 is a block diagram of a hardware structure of a mobile terminal according to a method for determining a target object according to an embodiment of the present invention. As shown in FIG. 1 , the mobile terminal may include one or more (only one is shown in FIG. 1 ) processor 102 (the processor 102 may include, but is not limited to, a processing device such as a microprocessor MCU or a programmable logic device FPGA) and a memory 104 for storing data, wherein the above-mentioned mobile terminal may further include a transmission device 106 and an input and output device 108 for communication functions. Those of ordinary skill in the art can understand that the structure shown in FIG. 1 is only a schematic diagram, which does not limit the structure of the above-mentioned mobile terminal. For example, the mobile terminal may also include more or fewer components than those shown in FIG. 1 , or have a different configuration than that shown in FIG. 1 .

The memory 104 can be used to store computer programs, for example, software programs and modules of application software, such as a computer program corresponding to a method for determining a target object in an embodiment of the present invention, the processor 102 runs the computer program stored in the memory 104, Thereby, various functional applications and data processing are performed, that is, the above-mentioned method is realized. Memory 104 may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 104 may further include memory located remotely from the processor 102, and these remote memories may be connected to the mobile terminal through a network. Examples of such networks include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

Transmission means 106 are used to receive or transmit data via a network. The specific example of the above-mentioned network may include a wireless network provided by a communication provider of the mobile terminal. In one example, the transmission device 106 includes a network adapter (Network Interface Controller, NIC for short), which can be connected to other network devices through a base station so as to communicate with the Internet. In one example, the transmission device 106 may be a radio frequency (Radio Frequency, RF for short) module, which is used to communicate with the Internet in a wireless manner.

A method for determining a target object is provided in this embodiment. FIG. 2 is a flowchart of a method for determining a target object according to an embodiment of the present invention. As shown in FIG. 2 , the process includes the following steps:

Step S202, acquiring target images of consecutive frames;

In this embodiment, the purpose of acquiring the target images of consecutive frames is to compare and identify the images of the consecutive frames, so as to determine whether the target object in the target images meets the motion condition, and improve the recognition accuracy.

Wherein, the target image of consecutive frames may be (but not limited to) two consecutive frames, or may be images of more than two consecutive frames, and the target image may be (but not limited to) a photo, a video, or other types of images. multimedia image data.

It should be noted that the target object can be (but not limited to) other things such as road, sky, people, cars, and throwing objects, wherein the throwing objects can be solid or non-solid things such as plastic bags, paper scraps, and cans.

Step S204, performing object type filtering processing on the target image to obtain the target processing area;

In this embodiment, the object type filtering process is performed on the target image to filter the area containing non-target objects, thereby reducing the amount of data processing in the recognition process and improving the target object recognition efficiency.

The object type filtering process may (but is not limited to) compare regions identified through different methods, and then filter regions containing non-target objects, or may select objects included in the target image to Determine the initial recognition area, and then perform recognition and filtering on the selected objects, and can also filter in other ways; the target processing area can be a square, circular or other polygonal identification frame area, or it can be intercepted from the target image. It can also be other forms of regions.

For example, each frame image of the test video stream is input into the trained MASK-RCNN or other instance segmentation model to classify each pixel of the target image into road, sky, people, cars, throwing objects through the instance segmentation model , a total of 6 other irregular areas containing suspected thrown objects, and then use the trained yolov5 deep learning detection model to detect people and vehicles to detect the image to obtain the rectangular area of people and vehicles in the image, and then multiple The irregular area is compared with the rectangular areas of people and vehicles, and if it is determined that the irregular area overlaps with the rectangular areas of people and vehicles, the divided area of the throwing object is eliminated, so as to obtain irregular throwing that does not overlap with people and vehicles. Object area, this area is the pre-trigger area of the projectile (equivalent to the aforementioned target processing area).

Step S206, performing regional feature filtering processing on the target processing area to obtain a first recognition area, wherein the target processing area includes the first recognition area;

In this embodiment, the area is filtered to filter objects that are too large and too small, thereby reducing the amount of calculation in the recognition process and improving the recognition efficiency; at the same time, reducing the interference of non-target objects and improving the recognition accuracy.

Wherein, the regional feature can be (but not limited to) the size, format, pixel, etc. of the target processing area, and the regional feature filtering process can (but is not limited to) filter out objects that are too large and too small; the first recognition region can be ( But not limited to) there may be more than one, or there may be one.

It should be noted that in the process of determining the target processing area, the space between the target processing area and the target object is small, so the size of the target processing area can reflect the size of the target object. The target processing area can be filtered to filter the non-target objects.

Step S208, based on the preset motion feature conditions, perform object motion feature detection processing on the first recognition area, and use the initial object included in the first recognition area that meets the motion feature conditions as the first object;

In this embodiment, determining the motion state of the initial object is to recognize objects with different motion states according to different requirements, so as to adapt to different recognition environments.

For example, it can identify the throwing objects that are located on the road and are in a stationary state, or identify objects that are in the process of throwing and moving.

Among them, the motion feature conditions include static duration, position changes in consecutive frames of images, linear velocity and direction, angular velocity, action duration and other conditions; the initial object can be (but not limited to) plastic bags, paper scraps, cans, etc. Solid or non-solid, moving or stationary projectiles.

Step S2010, performing pixel difference identification processing on the first object;

In this embodiment, the pixel difference recognition process is performed on the first object to distinguish the first object from the environment where the first object is located, so as to avoid recognition errors caused by environmental factors (such as ground, trees, etc.), and improve recognition precision.

Wherein, the pixel difference identification processing may be (but not limited to) comparing the pixels of the first object with the environment pixels, or may be comparing the identification detection result of the first object with the environment identification result, or may be performed in other ways. deal with.

Step S2012, when the processing result is that the first object satisfies the pixel difference condition, determine the first object as the target object.

In this embodiment, when it is confirmed that the pixel difference between the first object and the surrounding environment is large, it can be determined that the first object is a throwing object.

For example, when it is confirmed that the suspected thrown object is significantly different from the surrounding image pixels, the influence of the road surface on the recognition can be excluded, so that the object is determined to be a thrown object, which avoids the influence of environmental factors on the recognition process and improves the recognition accuracy. .

Through the above steps, since the object type processing is first performed on the images of consecutive frames, the interference factors in the process of target object recognition are reduced, and the subsequent region feature filtering processing can be performed to target processing regions that do not meet the requirements. Filtering, thereby reducing the amount of calculation in the recognition process, and performing object motion feature detection processing can avoid false recognition of stationary objects and improve the accuracy of recognition. Accurate identification solves the problem of low identification accuracy of the target object, and improves the identification accuracy and identification efficiency of the target object.

In an optional embodiment, performing object type filtering processing on the target image to obtain the target processing area includes:

Step S2042, performing environment type segmentation processing on the target image to obtain a first processing area;

In this embodiment, the purpose of performing environment type segmentation on the target image is to separate image regions containing different contents, so that individual regions can be identified, thereby reducing the amount of calculation in the identification process.

The environment type segmentation processing may be performed by a preset neural network model, or by a predetermined segmentation algorithm, or by other methods; the first processing area may be (but not limited to) is one, or at least two; the first processing area may be a regular polygon or circle, such as a square, triangle, circle, ellipse, etc., or an irregular polygon.

For example, each frame image of the test video stream is input into the trained MASK-RCNN or other instance segmentation model to classify each pixel of the target image into road, sky, people, cars, throwing objects through the instance segmentation model , and other 6 types of processing areas.

When it needs to be explained, using the image instance segmentation algorithm to extract the spill area can also avoid false detection caused by unstable background modeling.

Step S2044, performing object feature recognition processing on the target image to obtain a second processing area;

In this embodiment, the object feature recognition is performed on the target image to determine the region in the target image that contains the non-target object, so that the region containing the non-target object is screened out in the subsequent processing process.

Among them, the object feature recognition processing may be to confirm the area of non-target objects such as people and cars in the target image, or it may be to confirm the area of non-target objects such as buildings and trees, and it may also be to confirm other areas. The second processing area may (but It is not limited to one, but can also be at least two; the first processing area can be a regular polygon or circle, such as a square, triangle, circle, ellipse, etc., or an irregular polygon.

For example, using the trained yolov5 deep learning detection model to detect people and cars, detect the image to obtain the rectangular area of people and cars in the image.

It should be noted that the addition of the human-vehicle detection model is to filter out the false positives of people and vehicles, especially when the vehicle is mistakenly detected as a thrown object/remaining object in a traffic jam.

Step S2046, based on the second processing area, perform area coincidence filtering processing on the first processing area to obtain a target processing area.

In this embodiment, the area overlap filtering on the first processing area is to filter the first processing area that overlaps with the second processing area, so as to reduce the area containing non-target objects and reduce the amount of calculation.

For example, if the first processing area overlaps with the second processing area, the first processing area is eliminated to obtain a throwing object area that does not overlap with people and vehicles, that is, the throwing object pre-triggering area.

It should be noted that the execution order of step S2042 and step S2044 can be interchanged, that is, step S2044 may be executed first, and then step S2042 may be executed.

In an optional embodiment, based on the second processing area, performing area coincidence filtering processing on the first processing area to obtain the target processing area includes:

Step S20462, performing overlap detection processing on the first processing region and the second processing region to obtain a region coincidence value;

Step S20464, in the case that the determined area coincidence value is less than or equal to the preset value, determine the first processing area as the target processing area;

Step S20466, in the case that the determined area coincidence value is greater than the preset value, perform a first filtering process on the first processing area.

In this embodiment, when it is determined that the area coincidence value is greater than the preset value, it is considered that the first processing area and the second processing area have a greater degree of coincidence and contain more non-target objects, so they are filtered.

Wherein, the coincidence detection process may be (but not limited to) the calculation of the overlapped pixel quantity between the first processing area and the second processing area or the calculation of the overlapped boundary quantity between the first processing area and the second processing area, or It is to calculate the overlapping area between the first processing area and the second processing area, etc. The first filtering processing on the first processing area may (but is not limited to) exclude the first processing area or perform other processing.

In an optional embodiment, based on a preset motion feature condition, an object motion feature detection process is performed on the first recognition area, and an initial object included in the first recognition area that satisfies the motion feature condition is used as the first object including: :

Step S2082, performing motion state detection processing on the initial object;

Step S2084, in the case of determining that the initial object is in a static state, perform static duration detection processing on the initial object;

Step S2086, in the case that it is determined that the stationary duration of the initial object satisfies the threshold, the initial object is taken as the first object.

In this embodiment, the motion state detection may be performed by detecting the target images of consecutive frames. When the position of the initial object in the target images of consecutive frames does not change, it is confirmed that the target object is in a static state, otherwise it is in motion. After determining that the initial object is in a static state, the threshold value of the static duration is determined to avoid the judgment error caused by the secondary movement of the object affected by the surrounding environment, such as plastic being blown by the wind.

In an optional embodiment, performing area feature filtering processing on the target processing area to obtain the first identification area includes:

Step S2062, performing area feature detection processing on the target processing area to obtain area feature information of the target processing area;

Step S2064, performing area feature comparison processing on the area feature information, and in the case that the area feature information satisfies the area threshold, the target processing area is used as the first identification area; when it is determined that the area feature information does not meet the area threshold, the target processing area is determined. The processing area is subjected to a second filtering process.

In this embodiment, the area feature detection processing is performed on the target processing area in order to filter the area that is too large or too small, so as to reduce the interference caused by the non-target objects in the area that is too large or too small to the recognition process, and improve the detection efficiency.

In an optional embodiment, before the region feature filtering process is performed on the target processing region to obtain the first identification region, the method further includes:

Step S202, acquiring target images of consecutive frames;

Step S2040, performing image object detection processing on the target image to obtain a target processing area.

In this embodiment, the target processing area is directly determined by the image target detection processing, which can avoid the data classification process caused by cutting the target image, thereby improving the image recognition efficiency.

The image target detection processing may be implemented by a preset algorithm, or by a neural network model, or by other means.

In an optional embodiment, when the processing result is that the first object satisfies the pixel difference condition, after the first object is determined as the target object, the method further includes:

Step S2016, the target object is stored in the target storage area.

In this embodiment, the purpose of storing the target object in the target storage area is to facilitate information backtracking at a later stage and realize data management.

The target storage area may be (but not limited to) an internal storage device such as a hard disk and flash, or an external storage device such as a management cloud.

From the description of the above embodiments, those skilled in the art can clearly understand that the method according to the above embodiment can be implemented by means of software plus a necessary general hardware platform, and of course can also be implemented by hardware, but in many cases the former is better implementation. Based on this understanding, the technical solutions of the present invention can be embodied in the form of software products in essence or the parts that make contributions to the prior art, and the computer software products are stored in a storage medium (such as ROM/RAM, magnetic disk, CD-ROM), including several instructions to make a terminal device (which may be a mobile phone, a computer, a server, or a network device, etc.) execute the methods described in the various embodiments of the present invention.

This embodiment also provides an apparatus for determining a target object, the apparatus is used to implement the above-mentioned embodiments and preferred implementations, and what has been described will not be repeated. As used below, the term "module" may be a combination of software and/or hardware that implements a predetermined function. Although the apparatus described in the following embodiments is preferably implemented in software, implementations in hardware, or a combination of software and hardware, are also possible and contemplated.

Fig. 3 is a structural block diagram of the apparatus for removing a target object according to an embodiment of the present invention. As shown in Fig. 3, the apparatus includes:

The image acquisition module 32 is used for acquiring target images of consecutive frames;

The object type filtering module 34 is used to perform object type filtering processing on the target image to obtain the target processing area;

The regional feature filtering module 36 is configured to perform regional feature filtering processing on the target processing region to obtain a first recognition region, wherein the target processing region includes the first recognition region;

The motion feature detection module 38 is configured to perform object motion feature detection processing on the first recognition area based on preset motion feature conditions, and use the initial object that meets the motion feature conditions contained in the first recognition area as the first object;

a pixel difference identification module 310, configured to perform pixel difference identification processing on the first object;

The target object determination module 312 is configured to determine the first object as the target object when the processing result is that the first object satisfies the pixel difference condition.

In an optional embodiment, the object type filtering module 34 includes:

a type segmentation unit 342, configured to perform environment type segmentation processing on the target image to obtain a first processing area;

The feature recognition unit 344 is used to perform object feature recognition processing on the target image to obtain the second processing area;

The coincidence filtering unit 346 is configured to perform area coincidence filtering processing on the first processing area based on the second processing area, so as to obtain the target processing area.

In an optional embodiment, the coincidence filtering unit 346 includes:

The coincidence detection subunit 3462 is used to perform the coincidence detection process on the first processing area and the second processing area to obtain the area coincidence value;

The area determination subunit 3464 is used to determine the first processing area as the target processing area under the condition that the determined area coincidence value is less than or equal to the preset value;

The first filtering subunit 3466 is configured to perform a first filtering process on the first processing region when the determined region coincidence value is greater than a preset value.

In an optional embodiment, the motion feature detection module 38 includes:

A motion state detection unit 382, configured to perform motion state detection processing on the initial object;

A duration detection unit 384, configured to perform stationary duration detection processing on the initial object when it is determined that the initial object is in a stationary state;

The first object determining unit 386 is configured to use the initial object as the first object when it is determined that the stationary duration of the initial object satisfies the threshold.

In an optional embodiment, the regional feature filtering module 36 includes:

The area detection unit 362 is used to perform area feature detection processing on the target processing area to obtain the area feature information of the target processing area;

The feature comparison unit 364 is configured to perform area feature comparison processing on the area feature information, and when it is determined that the area feature information satisfies the area threshold, the target processing area is used as the first identification area; when it is determined that the area feature information does not meet the area threshold Next, a second filtering process is performed on the target processing area.

In an optional embodiment, the device further includes:

The image acquisition unit 302 is configured to acquire target images of consecutive frames before performing regional feature filtering processing on the target processing area to obtain the first recognition area;

The target detection module 340 is configured to perform image target detection processing on the target image to obtain the target processing area.

In an optional embodiment, the device further includes:

The storage module 316 is configured to store the target object in the target storage area after determining the first object as the target object when the processing result is that the first object satisfies the pixel difference condition.

It should be noted that the above modules can be implemented by software or hardware, and the latter can be implemented in the following ways, but not limited to this: the above modules are all located in the same processor; or, the above modules can be combined in any combination The forms are located in different processors.

The present invention will be described below with reference to specific embodiments.

As shown in Figure 4, this embodiment mainly includes the following steps:

Step S401, the image instance segmentation part, including collecting a large number of images containing throwing objects, and using the collected images to train MASK-RCNN or other instance segmentation models, which can divide each pixel of the image into road, sky, people, and vehicles. , throwing objects, other a total of 6 categories.

In step S402, each frame of the image of the test video stream is input into the instance segmentation model, and a plurality of irregular regions containing suspected thrown objects in the image can be obtained.

Step S403, the person and vehicle detection part, including using the trained yolov5 deep learning detection model to detect people and vehicles, detect the image, and obtain the rectangular area of people and vehicles in the image;

Step S404, remove the part of the person and vehicle area, if the throwing object segmentation area obtained in step S402 overlaps with the person and vehicle area detected in step S403, then remove the throwing object segmentation area to obtain the throwing object that does not overlap with people and vehicles. area, which is the pre-trigger area of the projectile;

In step S405, the logic filtering part performs area size filtering on the pre-triggered area of the throwing object obtained in step S404 (oversized and too small targets are filtered out), duration filtering (the throwing object needs to be on the road for a period of time to be considered It is a real throwing object. Only the pre-triggered throwing object always exists in the same position of the multi-frame image is a real throwing object), and the difference with the pixels around the area (the throwing object needs to be significantly different from the surrounding image pixels, which is used to exclude the road) and other logic Rule calculation to determine whether the object in the area is a throwing object;

In step S406, if it is determined that the object is thrown, an alarm process is performed.

Correspondingly, as shown in FIG. 5 , in order to realize the above steps, the present embodiment also provides a high-speed throwing object/remaining object detection system, which at least includes the following units:

Image acquisition unit 51, used for video stream reading;

A high-speed thrown object/remaining object detection unit 52, which is used to realize high-speed thrown object/remaining object detection through an algorithm;

The alarm unit 53 is used for outputting alarm information;

The event retrieval and query unit 54 is used for post-event retrieval and query.

Embodiments of the present invention further provide a computer-readable storage medium, where a computer program is stored in the computer-readable storage medium, wherein the computer program is configured to execute the steps in any of the above method embodiments when running.

In an exemplary embodiment, the above-mentioned computer-readable storage medium may include, but is not limited to, a USB flash drive, a read-only memory (Read-Only Memory, referred to as ROM for short), and a random access memory (Random Access Memory, referred to as RAM for short) , mobile hard disk, magnetic disk or CD-ROM and other media that can store computer programs.

An embodiment of the present invention also provides an electronic device, comprising a memory and a processor, where a computer program is stored in the memory, and the processor is configured to run the computer program to execute the steps in any of the above method embodiments.

In an exemplary embodiment, the above-mentioned electronic device may further include a transmission device and an input-output device, wherein the transmission device is connected to the above-mentioned processor, and the input-output device is connected to the above-mentioned processor.

For specific examples in this embodiment, reference may be made to the examples described in the foregoing embodiments and exemplary implementation manners, and details are not described herein again in this embodiment.

Obviously, those skilled in the art should understand that the above-mentioned modules or steps of the present invention can be implemented by a general-purpose computing device, which can be centralized on a single computing device, or distributed in a network composed of multiple computing devices On the other hand, they can be implemented in program code executable by a computing device, so that they can be stored in a storage device and executed by the computing device, and in some cases, can be performed in a different order than shown here. Or the described steps, or they are respectively made into individual integrated circuit modules, or a plurality of modules or steps in them are made into a single integrated circuit module to realize. As such, the present invention is not limited to any particular combination of hardware and software.

The above descriptions are only preferred embodiments of the present invention, and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and changes. Any modification, equivalent replacement, improvement, etc. made within the principles of the present invention shall be included within the protection scope of the present invention.

Claims (9)

1. A target object determination method, comprising:

acquiring target images of continuous frames;

carrying out object type filtering processing on the target image to obtain a target processing area;

performing regional characteristic filtering processing on the target processing region to obtain a first identification region, wherein the target processing region comprises the first identification region;

performing object motion feature detection processing on the first identification area based on a preset motion feature condition, and taking an initial object which meets the motion feature condition and is contained in the first identification area as a first object;

performing pixel difference recognition processing on the first object;

determining the first object as the target object in the case that the first object satisfies a pixel difference condition as a result of the processing;

wherein the performing object type filtering processing on the target image to obtain a target processing region includes: performing environment type segmentation processing on the target image to obtain a first processing area;

carrying out object feature recognition processing on the target image to obtain a second processing area;

and performing area superposition filtering processing on the first processing area based on the second processing area to obtain the target processing area.

2. The method of claim 1, wherein said performing a region-coincidence filtering process on the first processing region based on the second processing region to obtain the target processing region comprises:

carrying out coincidence detection processing on the first processing area and the second processing area to obtain an area coincidence value;

determining the first processing area as the target processing area under the condition that the area coincidence value is determined to be smaller than or equal to a preset value;

and under the condition that the area coincidence value is determined to be larger than the preset value, performing first filtering processing on the first processing area.

3. The method according to claim 1, wherein the performing object motion feature detection processing on the first recognition area based on a preset motion feature condition, and the taking an initial object included in the first recognition area and satisfying the motion feature condition as a first object comprises:

carrying out motion state detection processing on the initial object;

under the condition that the initial object is determined to be in a static state, carrying out static duration detection processing on the initial object;

taking the initial object as the first object when the static duration of the initial object is determined to meet a threshold.

4. The method according to claim 1, wherein the performing a region feature filtering process on the target processing region to obtain a first identified region comprises:

carrying out area characteristic detection processing on the target processing region to acquire area characteristic information of the target processing region;

performing area feature comparison processing on the area feature information, and taking the target processing region as the first identification region under the condition that the area feature information is determined to meet an area threshold; and under the condition that the area characteristic information is determined not to meet the area threshold value, performing second filtering processing on the target processing area.

5. The method according to claim 1, wherein before said performing a region feature filtering process on said target processing region to obtain a first identified region, said method further comprises:

acquiring target images of continuous frames;

and carrying out image target detection processing on the target image to obtain the target processing area.

6. The method according to claim 1, wherein after determining the first object as the target object in a case where the processing result is that the first object satisfies a pixel difference condition, the method further comprises:

and storing the target object to a target storage area.

7. A target object determination apparatus, comprising:

the image acquisition module is used for acquiring target images of continuous frames;

the object type filtering module is used for carrying out object type filtering processing on the target image to obtain a target processing area;

the regional characteristic filtering module is used for performing regional characteristic filtering processing on the target processing region to obtain a first identification region, wherein the target processing region comprises the first identification region;

the motion characteristic detection module is used for carrying out object motion characteristic detection processing on the first identification area based on a preset motion characteristic condition and taking an initial object which is contained in the first identification area and meets the motion characteristic condition as a first object;

the pixel difference identification module is used for carrying out pixel difference identification processing on the first object;

a target object determination module, configured to determine the first object as the target object if the processing result is that the first object satisfies a pixel difference condition;

wherein the object type filtering module comprises:

the type segmentation unit is used for carrying out environment type segmentation processing on the target image to obtain a first processing area; the characteristic identification unit is used for carrying out object characteristic identification processing on the target image to obtain a second processing area; and the overlapping filtering unit is used for carrying out area overlapping filtering processing on the first processing area based on the second processing area so as to obtain a target processing area.

8. A computer-readable storage medium, in which a computer program is stored, wherein the computer program is arranged to perform the method of any of claims 1 to 6 when executed.

9. An electronic device comprising a memory and a processor, wherein the memory has stored therein a computer program, and wherein the processor is arranged to execute the computer program to perform the method of any of claims 1 to 6.

Images