An image recognition-based supervision method and system

Abstract

The application is suitable for the field of computers and provides a supervision method and system based on image recognition. The method comprises the following steps: acquiring an acquisition image of a critical region in a target region, wherein the target region comprises a target water area, and the critical region comprises a feature sub-region from a side bank to the target water area; when an activity target appears in the feature sub-region according to the acquisition image, judging a trend of the activity target in the feature sub-region; when the trend of the activity target is close to the target water area, acquiring a change of height information of the activity target, wherein the height information comprises height information in a non-crouching state; and if it is judged that a height information drop value of the activity target reaches a preset ratio according to the change of the height of the activity target, reporting a prompt information to a supervision end and alarming in the feature sub-region. The application has the beneficial effect that dangerous behaviors entering the target water area can be effectively supervised.

Description

TECHNICAL FIELD

[0001] The present application belongs to the field of computer, and particularly relates to a supervision method and system based on image recognition. BACKGROUND

[0002] An image is composed of many basic pixel points with color types and brightness level information, and is represented by limited digital pixel values. A digital image is also called a digital image, and image recognition mainly experiences three stages: recognition of text information, recognition of digital information, and recognition of objects. Image recognition has been continuously developed through these three stages, fully exerting its own characteristics and advantages, and gradually expanding to various fields and combining with various industry technologies. The main application and development direction of image recognition technology include character recognition, machine vision recognition, graphic image recognition, and biomedical technology.

[0003] At present, water areas include natural water areas and non-natural water areas. The natural water areas are, for example, some reservoirs, ponds, and lakes, and the non-natural water areas are, for example, artificial lakes and artificial ponds. There are some dangerous behaviors in these water areas, and supervision needs to be conducted to prevent accidents such as drowning. However, in the prior art, artificial patrol or installation of monitoring is used for prevention, but these methods consume manpower or have low supervision efficiency. SUMMARY

[0004] The purpose of the embodiment of the present application is to provide a supervision method and system based on image recognition, which aims to solve the problems in the background technology.

[0005] The embodiment of the present application is implemented in the following manner. On one hand, a supervision method based on image recognition includes the following steps:

[0006] An acquisition image of a critical region in a target region is obtained, the target region includes a target water area, and the critical region includes a feature sub-region via a bank to the target water area;

[0007] When it is identified from the acquisition image that an active target appears in the feature sub-region, a trend of the active target in the feature sub-region is judged;

[0008] When it is judged that the trend of the active target approaches the target water area, a change in height information of the active target is obtained, and the height information includes height information in a non-crouching state;

[0009] If it is judged from the change in the height of the active target that a height information drop value of the active target reaches a preset ratio, a prompt information is reported to a supervision end and an alarm is given in the feature sub-region, and the prompt information is used to represent that the active target may be in a dangerous state.

[0010] As a further scheme of the present application, after the acquisition image of the critical region in the target region is acquired, the method further comprises:

[0011] comparing the acquisition image with the image of the feature sub-region in the non-foreign object state, and identifying the first target object in the acquisition image;

[0012] According to the position of the first target object in the acquisition image, it is judged whether the position change of the first target object within the first preset time length exceeds the set threshold;

[0013] When the position change of the first target object within the first preset time length exceeds the set threshold, it is determined that the corresponding first target object is an active target.

[0014] As a further scheme of the present application, after the acquisition image of the critical region in the target region is acquired, the method further comprises:

[0015] extracting feature information of the first target object in the acquisition image, the feature information including external features and / or action features;

[0016] When the external features satisfy the human body contour features, it is determined that the feature information satisfies the first condition;

[0017] When the action features satisfy the human upper limb action features and / or lower limb action features, it is determined that the feature information satisfies the second condition;

[0018] When the feature information satisfies at least one of the first condition and the second condition, the first target is determined to be a first active target, and the active target includes the first active target.

[0019] As a further scheme of the present application, the determination of the trend of the active target in the feature sub-region specifically comprises:

[0020] According to a preset reference position, a first position of the active target is recorded, the preset reference position being located at the side of the feature sub-region facing the bank, and the first position being located on the inner side of the preset reference position towards the target water area;

[0021] If a second position of the active target is continuously captured, and it is determined that the line connecting the second position, the first position and the preset reference position on the same reference plane is an obtuse triangle, it is determined that the trend of the active target approaches the target water area, and the second position is located within the acquisition range of the acquisition terminal.

[0022] As a further scheme of the present application, the method further comprises:

[0023] According to at least two different acquisition positions, a first coordinate of the active target at the first time is acquired;

[0024] acquiring second coordinates of the moving target at a second time, the second time being a time subsequent to the first time, from at least two different acquisition positions;

[0025] judging whether the line connecting the first coordinates and the second coordinates extends towards the target water area;

[0026] when it is determined that the line extends towards the target water area, determining that the moving target approaches the target water area.

[0027] As a further scheme of the present application, the method further comprises:

[0028] acquiring third coordinates of the moving target at a third time, the third time being a time subsequent to the second time, from at least two different acquisition positions;

[0029] judging whether the line connecting the second coordinates and the third coordinates extends away from the target water area;

[0030] if yes, continuing to determine the moving target's trend.

[0031] As a further scheme of the present application, the method further comprises:

[0032] if it is determined that the moving target's trend approaches the target water area at least once more, determining that the moving target has a dangerous behavior tendency;

[0033] reporting early warning information to a supervision end, the early warning information being used to represent that the moving target has a dangerous behavior tendency.

[0034] As a further scheme of the present application, the method further comprises:

[0035] when it is determined that the moving target approaches the target water area, starting to acquire height information of the moving target;

[0036] recording first height information of the moving target in the height information, the first height information being height information when the moving target is at a distance not greater than a first preset distance from the edge of the target water area;

[0037] recording second height information of the moving target in the height information, the second height information being height information when the moving target is above the water surface;

[0038] judging a difference value of the first height information and the second height information;

[0039] if it is determined that the height value of the moving target decreases by a preset ratio according to the difference value, determining that a condition for prompting and alarming is reached, the condition being used to represent that prompt information is reported to a supervision end and alarming is performed in a characteristic sub-region.

[0040] As a further scheme of the present application, in another aspect, an image recognition-based supervision system, the system comprising:

[0041] An image acquisition module, configured to acquire an acquisition image of a critical region in a target region, the target region comprising a target water area, and the critical region comprising a feature sub-region via a shore to the target water area;

[0042] An active target identification module, configured to, when identifying, according to the acquisition image, that an active target appears in the feature sub-region, judge a trend of the active target in the feature sub-region;

[0043] An approaching height acquisition module, configured to, when judging that the trend of the active target approaches the target water area, acquire a change in height information of the active target, the height information comprising height information in a non-crouching state;

[0044] A judgment and early warning module, configured to, when judging, according to the change in the height of the active target, that a height information drop value of the active target reaches a preset ratio, report a prompt information to a supervision end and perform an alarm in the feature sub-region, the prompt information being used to represent that the active target is possibly in a dangerous state.

[0045] The image recognition-based supervision method and system provided by the embodiments of the present application can, when identifying, according to an acquisition image, that an active target appears in a feature sub-region, judge a trend of the active target in the feature sub-region, acquire a change in height information of the active target when judging that the trend of the active target approaches a target water area, the height information comprising height information in a non-crouching state, and report a prompt information to a supervision end and perform an alarm in the feature sub-region when judging, according to the change in the height of the active target, that a height information drop value of the active target reaches a preset ratio, so as to be capable of prompting and alarming a situation in which the active target enters the water area, capable of distinguishing a normal scene such as approaching the water area, ensuring the accuracy of identification, and capable of effectively supervising dangerous behaviors of entering the target water area compared to pure artificial supervision and patrol. BRIEF DESCRIPTION OF DRAWINGS

[0046] Figure 1 FIG. 1 is a main flowchart of an image recognition-based supervision method.

[0047] Figure 2 FIG. 2 is a flowchart of extracting feature information of a first target object in an acquisition image in the image recognition-based supervision method.

[0048] Figure 3 FIG. 3 is a schematic diagram of judging a trend of an active target in a feature sub-region in the image recognition-based supervision method.

[0049] Figure 4is a flow chart of an embodiment of a trend of the moving target approaching the target water area in a supervision method based on image recognition.

[0050] Figure 5 is a flow chart related to a height value of the moving target falling to a preset ratio according to the difference value in a supervision method based on image recognition.

[0051] Figure 6 is a main structure diagram of a supervision system based on image recognition. DETAILED DESCRIPTION

[0052] In order to make the objects, technical solutions and advantages of the present application clearer, the present application is further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and do not limit the present application.

[0053] The specific implementation of the present application is described in detail below with reference to specific embodiments.

[0054] The present application provides a supervision method and system based on image recognition, which solves the technical problems in the background art.

[0055] As shown in Figure 1 is a main flow chart of a supervision method based on image recognition provided by an embodiment of the present application, the supervision method based on image recognition comprises:

[0056] Step S10: acquiring a collection image of a critical region in a target region, the target region comprising a target water area, and the critical region comprising a feature sub-region via a shore to the target water area; generally, if a moving target enters the target water area, it is generally through a normal route or road of the target water area close to the shore to enter if it is in an intentional case such as playing or swimming, of course, the feature sub-region of the critical region can also be some non-human crowd and the like frequently accessed region, specifically, in actual application, some high-risk areas of the target region can be selected as the feature sub-region by human;

[0057] Step S11: when a moving target appears in the feature sub-region according to the collection image, judging a trend of the moving target in the feature sub-region; the moving target is an object or organism that can move autonomously or under the control of a person; the trend of the moving target is a moving trend of the moving target, such as gradually approaching, gradually moving away, etc.

[0058] Step S12: when judging that the trend of the activity target approaches the target water area, the change of the height information of the activity target is acquired, the height information including height information in a non-crouching state; that is, the activity target has a trend of approaching the target water area; considering that in real life, there are not only activity targets in the water area that are relatively dangerous, but also some activity targets that can independently move and are relatively safe, behaviors generated by these activity targets include people approaching a washing hand, animals approaching a drinking water, etc.; the crouching state refers to a state of bending the body or squatting, because if an activity target enters the target water area, it generally directly enters the water area and will not appear a crouching state; the non-crouching state includes a situation of people approaching a water and not entering the water area; of course, the height information can also include height information in other states, such as a side-lying state and a back-lying state, which indicates that there may be a situation of stumbling and drowning;

[0059] Step S13: if it is judged according to the change of the height of the activity target that the height information of the activity target decreases by a preset ratio, a prompt information is reported to a supervision end and an alarm is performed in the feature sub-area, the prompt information being used to represent that the activity target may be in a dangerous state. Generally, if an activity target enters the target water area, the height above the water surface will change, or the activity target enters the silt, and the height will also change;

[0060] In application of the embodiment, when the activity target in the feature sub-area is recognized according to the collected image, the trend of the activity target in the feature sub-area is judged; when it is judged that the trend of the activity target approaches the target water area, the change of the height information of the activity target is acquired, the height information including height information in a non-crouching state, if it is judged according to the change of the height of the activity target that the height information of the activity target decreases by a preset ratio, a prompt information is reported to a supervision end and an alarm is performed in the feature sub-area, the activity target entering the water area can be prompted and alarmed, and a normal scene of approaching the water area can be distinguished, the accuracy of identification is ensured, and compared with pure artificial supervision and patrol, the dangerous behavior of entering the target water area can be effectively supervised.

[0061] As a preferred embodiment of the application, after the collected image of the critical region in the target region is acquired, the method further includes:

[0062] Step S20: comparing the collected image with an image of the feature sub-area in a state without foreign matters, and identifying a first target object in the collected image;

[0063] Step S21: judging whether a position change of the first target object in a first preset time length exceeds a set threshold value according to a position of the first target object in the collected image;

[0064] Step S22: When the position change of the first target object within the first preset time length exceeds the set threshold, it is determined that the corresponding first target object is an active target.

[0065] It can be understood that the active target herein includes but is not limited to people (groups), and can also be animals, machines with walking functions such as vehicles, etc. The image of the feature sub-region in the state of no external object indicates that it is in a state of no entry of people or other external objects. When identifying the specific active target, the determination condition is to compare the first target object in the collected image with the preset external object image. When the comparison result meets the similarity condition (similar threshold), the type of the first target object is determined.

[0066] As shown in Figure 2 As a preferred embodiment of the present application, after obtaining the collected image of the critical region in the target region, the method further includes:

[0067] Step S30: Extracting feature information of the first target object in the collected image, the feature information including external features and / or action features;

[0068] Step S31: When the external features meet the human body contour feature, it is determined that the feature information meets the first condition; the human body contour feature includes the human face contour feature; the face contour feature does not have to be specific face information;

[0069] Step S32: When the action features meet the human upper limb action features and / or lower limb action features, it is determined that the feature information meets the second condition; the limb action features include the arm swinging gesture feature and the foot lifting gesture feature. Without obtaining specific biological information of the person, the action features can be used to determine whether the active target is a person.

[0070] Step S33: When the feature information meets at least one of the first condition and the second condition, it is determined that the first target is a first active target, and the active target includes the first active target.

[0071] In the implementation of the present embodiment, considering the application situation in practice, when the active target is a person, it meets the human activity feature. In this condition, as long as the external features meet the human body contour feature and the action features meet at least one of the human upper limb action features and / or the lower limb action features, it is determined that the first target is a first active target. Since the first active target is the object of key supervision, it needs to be identified and focused.

[0072] As shown in Figure 3 As a preferred embodiment of the present application, the determination of the trend of the active target in the feature sub-region specifically includes:

[0073] Step S111: record a first position of the activity target according to a preset reference position, the preset reference position is located at a side edge of the characteristic sub-region facing the bank, and the first position is located at an inner side of the preset reference position deviating toward the target water area; in actual application, a ground mark or a marker can be arranged in the characteristic sub-region to facilitate distinguishing or identifying the position of the activity target.

[0074] Step S112: if a second position of the activity target is continuously captured, and it is determined that a line connecting the second position, the first position and the preset reference position is an obtuse triangle, it is determined that the activity target approaches the target water area, and the second position is located in the collection range of the collection terminal.

[0075] It should be understood that the preset reference position is preferably the position of the collection terminal, that is, the preset reference position is located at an edge position of the side of the characteristic sub-region facing the bank, that is, an edge position outside the side of the target water area, and the edge position can facilitate covering the characteristic sub-region when collecting images; the first position is located at an inner side of the preset reference position deviating toward the target water area, and can be understood as a non-edge area of the characteristic sub-region, that is, the activity position of most interactive targets, and the embodiment can reduce the number of collection or capture terminals; for example, when the preset reference position is preferably the position of the collection terminal, at least one collection terminal can be used to complete the embodiment.

[0076] As shown in FIG. 1, as a preferred embodiment of the present application, the method further comprises: Figure 4

[0077] Step S40: acquire a first coordinate of the activity target at a first time according to at least two different collection positions; the collection position generally refers to the position of the collection terminal, and in actual application, if the number of collection terminals is greater than or equal to 2, the collection coverage ranges of the collection terminals overlap, virtual block information of the overlapping region is established, each virtual block is a part of the overlapping region, and when the coordinate of the activity target is acquired, the position of the virtual block can be used as the coordinate of the activity target; the at least two collection terminals can be relatively arranged and distributed at two edge positions of one side of the characteristic sub-region.

[0078] Step S41: acquire a second coordinate of the activity target at a second time according to at least two different collection positions, and the second time is a subsequent time of the first time.

[0079] Step S42: determine whether the line connecting the first coordinate and the second coordinate extends toward the target water area.

[0080] Step S43: when it is determined that the line extends toward the target water area, it is determined that the activity target approaches the target water area.

[0081] ​The embodiment actually provides a method different from the method of the previous embodiment, and whether the trend of the moving target approaches the target water area can be determined. The main advantage of the embodiment is that the coordinates of the moving target are determined according to the coordinates of at least two collection positions. In actual application, since the subsequent judgment is about the extension of the line, the requirement for the accuracy of the coordinates is generally low.

[0082] As a preferred embodiment of the present application, the method further comprises:

[0083] Step S50: obtaining a third coordinate of the moving target at a third time according to at least two different collection positions, the third time being a subsequent time of the second time;

[0084] Step S51: judging whether a line connecting the second coordinate and the third coordinate extends to a direction away from the target water area; the line is from the second coordinate to the third coordinate;

[0085] Step S52: if yes, the trend of the moving target is continuously determined.

[0086] It can be understood that the embodiment provides a situation that whether the moving target approaches the target water area and then moves away. For such a moving target, there may be a dangerous behavior tendency, or it may be mistakenly entered into the feature sub-area. The embodiment considers the comprehensiveness of the moving target approaching or moving away.

[0087] As a preferred embodiment of the present application, the method further comprises:

[0088] Step S60: if the trend of the moving target is determined to approach the target water area at least once again, it is determined that the moving target may have a dangerous behavior tendency;

[0089] Step S61: reporting early warning information to a supervision end, the early warning information being used to represent that the moving target may have a dangerous behavior tendency.

[0090] It can be understood that the embodiment provides a method for early discovery of a moving target with a dangerous tendency. Such a person may have a tendency to hesitate to enter the target water area due to some reasons, and thus may linger in the feature sub-area and repeatedly approach or move away from the feature sub-area. Therefore, for such a moving target, early warning information can be directly reported to the supervision end to take measures such as discouragement as early as possible.

[0091] As shown in Figure 5 As a preferred embodiment of the present application, the method further comprises:

[0092] Step S70: starting to obtain height information of the moving target when it is determined that the trend of the moving target approaches the target water area;

[0093] Step S71: record the first height information of the moving target in the height information, the first height information being the height information when the distance between the moving target and the edge of the target water area is not greater than a first preset distance;

[0094] Step S72: record the second height information of the moving target in the height information, the second height information being the height information when the moving target is above the water surface; here, the height information can be obtained by setting a reference object at the edge of the target water area for comparison, for example, a warning sign or a natural reference object.

[0095] Step S73: determine the difference value of the first height information and the second height information;

[0096] Step S74: if it is determined according to the difference value that the height value of the moving target decreases by a preset ratio, it is determined that the condition for prompting and alarming is reached, the condition being used to represent the reporting of the prompt information to the monitoring end and the alarming in the feature sub-area. It should be noted that the preset ratio here can be different values for different height groups, because the differences between adults and children, or different height groups are considered, especially for those with smaller height, the corresponding preset ratio should be smaller, for example, for the height value between 150-160mm, the preset ratio is 1 / 18, and for the height value between 165-170mm, the preset ratio is 1 / 16.

[0097] It should be understood that in general cases, in the process of detecting the moving target, if only the collected images are used for identification, the collection frequency needs to be improved to meet the requirements of height information comparison, otherwise it may lead to late prompting and early warning. For some dangerous water areas, a monitoring center needs to be equipped nearby, the monitoring end can be set in the monitoring center and can be looked after by a special person, and when necessary, the monitoring end can be connected with the alarm device to establish a rapid alarm response channel.

[0098] As shown in Figure 6 As another preferred embodiment of the present application, on the other hand, a monitoring system based on image recognition, the system comprises:

[0099] The image acquisition module 100 is used to acquire the collected images of the critical area in the target area, the target area including a target water area, and the critical area including a feature sub-area from the edge of the bank to the target water area.

[0100] The moving target identification module 200 is used to determine the trend of the moving target in the feature sub-area when the moving target appears in the feature sub-area according to the collected images.

[0101] The approximation height acquisition module 300 is configured to: when it is judged that the trend of the activity target approaches the target water area, acquire the change of the height information of the activity target, and the height information includes the height information in the non-crouching state.

[0102] The judgment and warning module 400 is configured to: if it is judged that the height information of the activity target decreases by a preset ratio according to the change of the height of the activity target, report a prompt information to a supervision end and perform warning in the feature sub-area, and the prompt information is used to represent that the activity target is in a dangerous state.

[0103] The above embodiment of the application provides a supervision method based on image recognition, and the supervision method based on image recognition provides a supervision system based on image recognition, when an activity target is recognized in a feature sub-area according to a collected image, the trend of the activity target in the feature sub-area is judged; when it is judged that the trend of the activity target approaches the target water area, the change of the height information of the activity target is acquired, and the height information includes the height information in the non-crouching state; if it is judged that the height information of the activity target decreases by a preset ratio according to the change of the height of the activity target, a prompt information is reported to a supervision end and warning is performed in the feature sub-area, which can prompt and warn the situation that the activity target enters the water area, and can distinguish the normal scene of approaching the water area, so as to ensure the accuracy of the recognition, and compared with the pure artificial supervision and patrol, the dangerous behavior of entering the target water area can be effectively supervised.

[0104] In order to enable the above method and system to run smoothly, the system can include more or less components than described above, or combine some components, or different components, for example, can include input and output devices, network access devices, buses, processors and memories, etc.

[0105] The processor can be a central processing unit (CPU), and can also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor and the like. The above processor is the control center of the system, and is connected with various parts by various interfaces and lines.

[0106] The above memory can be used to store computer and system programs and / or modules, and the above processor realizes the above various functions by running or executing the computer programs and / or modules stored in the memory and calling the data stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, application programs required by at least one function (such as an information collection template display function, a product information publishing function, etc.), and the like. The data storage area can store data created according to the use of the berth state display system (such as product information collection templates corresponding to different product categories, product information to be published by different product providers, etc.). In addition, the memory can include a high-speed random access memory, and can also include a non-volatile memory, for example, a hard disk, a memory, a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, at least one disk storage device, a flash memory device, or other volatile solid-state storage devices.

[0107] It should be understood that, although each step in the flowchart of each embodiment of the present application is displayed in sequence according to the direction of the arrow, these steps are not necessarily executed in sequence according to the direction of the arrow. Unless otherwise specified herein, the execution of these steps is not strictly limited in sequence, and these steps can be executed in other sequences. Moreover, at least part of the steps in each embodiment can include multiple sub-steps or multiple stages, which are not necessarily executed at the same time, but can be executed at different times, and the execution sequence of these sub-steps or stages is not necessarily sequential, but can be executed alternately or alternately with at least part of other steps or sub-steps or stages of other steps.

[0108] Each technical feature of the above-described embodiments can be combined arbitrarily, and in order to make the description concise, each technical feature in the above-described embodiments is not described in all possible combinations, however, as long as the combination of these technical features does not exist contradictory, it should be considered as the scope of the present application.

[0109] The above-described embodiments only express several implementation manners of the present application, and the description is more specific and detailed, but it should not be understood as a limitation on the patent scope of the present application. It should be noted that for ordinary skilled persons in the art, without departing from the concept of the present application, a number of modifications and improvements can be made, which are within the protection scope of the present application. Therefore, the protection scope of the present application patent should be subject to the appended claims.

[0110] The above merely describes preferred embodiments of the present application, and is not used to limit the present application, any modification, equivalent replacement and improvement within the spirit and principle of the present application should be included in the protection scope of the present application.

Claims

1. An image recognition-based supervision method, characterized by, The method comprises: acquiring a collection image of a critical region in a target region, the target region comprising a target water area, and the critical region comprising a feature sub-region via a bank to the target water area; when an active target appears in the feature sub-region according to the collection image, judging a trend of the active target in the feature sub-region; when the trend of the active target is judged to approach the target water area, acquiring a change in height information of the active target, the height information comprising height information in a non-crouching state; specifically comprising: starting to acquire the height information of the active target when the trend of the active target is judged to approach the target water area; recording first height information of the active target in the height information, the first height information being height information when the distance between the active target and the edge of the target water area is not greater than a first preset distance; and recording second height information of the active target in the height information, the second height information being height information when the active target is above the water surface; if it is judged according to the change in the height of the active target that the height information of the active target decreases by a preset ratio, reporting prompt information to a supervision end and performing an alarm in the feature sub-region, the prompt information being used to represent that the active target may be in a dangerous state; specifically comprising: judging a difference value between the first height information and the second height information; if it is judged according to the difference value that the height of the active target decreases by a preset ratio, it is judged that the condition for prompting and alarming is reached, the condition being used to represent that the prompt information is reported to the supervision end and the alarm is performed in the feature sub-region.

2. The image recognition based supervision method according to claim 1, characterized in that, After acquiring the collection image of the critical region in the target region, the method further comprises: comparing the collection image with an image of the feature sub-region in a state without foreign objects, and identifying a first target object in the collection image; judging whether a position change of the first target object within a first preset time length exceeds a set threshold value according to the position of the first target object in the collection image; when the position change of the first target object within the first preset time length exceeds the set threshold value, it is determined that the corresponding first target object is an active target.

3. The image recognition based monitoring method of claim 1, wherein, After acquiring the collection image of the critical region in the target region, the method further comprises: extracting feature information of the first target object in the collection image, the feature information comprising external features and / or action features; when the external features satisfy human body contour features, it is determined that the feature information satisfies a first condition; when the action features satisfy human body upper limb action features and / or lower limb action features, it is determined that the feature information satisfies a second condition; when the feature information satisfies at least one of the first condition and the second condition, it is determined that the first target is a first active target, and the active target comprises the first active target.

4. The image recognition based monitoring method of claim 3, wherein, The judgment of the trend of the active target in the feature sub-region specifically comprises: recording a first position of the active target according to a preset reference position, the preset reference position being located at a side of the feature sub-region facing the bank, and the first position being located at an inner side of the preset reference position facing the target water area; If a second position of the moving target is continuously captured, and if a line connecting the second position, the first position and a preset reference position is determined as an obtuse triangle, it is determined that the moving target approaches the target water area, and the second position is located in the collection range of the collection terminal.

5. The image recognition based monitoring method of claim 1, wherein, The method further comprises: acquiring a first coordinate of the moving target at a first time according to at least two different collection positions; acquiring a second coordinate of the moving target at a second time according to at least two different collection positions, the second time being a subsequent time of the first time; determining whether a line connecting the first coordinate and the second coordinate extends towards the target water area; if it is determined that the line extends towards the target water area, it is determined that the moving target approaches the target water area.

6. The image recognition based monitoring method of claim 5, wherein, The method further comprises: acquiring a third coordinate of the moving target at a third time according to at least two different collection positions, the third time being a subsequent time of the second time; determining whether a line connecting the second coordinate and the third coordinate extends in a direction away from the target water area; if yes, the determination of the moving trend of the moving target is continuously performed.

7. The image recognition based supervision method according to claim 5 or 6, characterized in that, The method further comprises: if it is determined that the moving trend of the moving target approaches the target water area at least once more, it is determined that the moving target has a dangerous behavior tendency; reporting early warning information to a supervision terminal, the early warning information being used to represent that the moving target has a dangerous behavior tendency.

8. An image recognition based monitoring system, characterized in that, The system comprises: a collection image acquisition module, configured to acquire a collection image of a critical region in a target region, the target region comprising a target water area, and the critical region comprising a feature sub-region from a shore to the target water area; a moving target identification module, configured to, when the moving target appears in the feature sub-region according to the collection image, determine a moving trend of the moving target in the feature sub-region; an approaching height acquisition module, configured to, when it is determined that the moving trend of the moving target approaches the target water area, acquire a change of height information of the moving target, the height information comprising height information in a non-crouching state; specifically, when it is determined that the moving trend of the moving target approaches the target water area, the height information of the moving target is acquired; first height information of the moving target in the height information is recorded, the first height information being height information when a distance between the moving target and an edge of the target water area is not greater than a first preset distance; second height information of the moving target in the height information is recorded, the second height information being height information when the moving target is above the water surface; a determination and early warning module, configured to, if it is determined that a height value of the moving target decreases by a preset ratio according to the change of the height of the moving target, report prompt information to a supervision terminal and perform an alarm in the feature sub-region, the prompt information being used to represent that the moving target is in a dangerous state; specifically, a difference value between the first height information and the second height information is determined; if it is determined that the height value of the moving target decreases by the preset ratio according to the difference value, a condition for the prompt and the alarm is determined, the condition being used to represent that the prompt information is reported to the supervision terminal and the alarm is performed in the feature sub-region.

Images