Target tracking method, device, equipment and storage medium

Abstract

Embodiments of the present application provide a target tracking method, device and equipment and a storage medium. The method comprises: obtaining a target detection frame region from a video stream; the target detection frame region comprises a candidate target object that is occluded and / or a candidate target object that has strong light; determining, based on feature information of the candidate target object contained in the target detection frame region, an original target object that matches the candidate target object from the video stream; and merging the candidate target object and the original target object in the target detection frame region to obtain a tracking result corresponding to a tracking number of the original target object. The candidate target object that is occluded and / or the candidate target object that has strong light has the same tracking number as the matched original target object, long-term stable tracking of the original target object is ensured in the case of high occlusion and day-night change, the accuracy of multi-target tracking is improved, the tracking continuity of the target when the target is occluded and at night is ensured, and long-term stable tracking of the target is achieved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data tracking and processing, in particular to a target tracking method, a target tracking device, a corresponding electronic device and a corresponding computer storage medium. BACKGROUND

[0002] In recent years, with the increasing improvement of people's living standards in China, the number of private cars has also increased. However, due to the fact that most of the old urban areas in the city were not planned to reserve a large number of parking spaces, the supply and demand contradiction of parking resources has become increasingly acute, and the situation of "one spot difficult to find" has appeared in many areas. Compared with the traditional closed parking lot construction, the intelligent design and planning of curb parking can more flexibly solve the problem of parking difficulty, solve the phenomenon of random parking of vehicles, and also enable the rapid realization of urban assets, which is a new outlet for the urban parking industry.

[0003] At present, the charging service of the curb parking can be realized by manual participation or video monitoring of the parking and parking of vehicles by high-position cameras. Among them, the scheme using video analysis needs to track the vehicle for a long time and stably, and does not allow the situation of number stringing and interruption. However, the curb parking is usually severely blocked, and the vehicle is parked for a long time, so it is necessary to track the vehicle for a long time and stably with strong anti-interference ability. SUMMARY

[0004] In view of the above problems, the present application is proposed to provide a target tracking method, a target tracking device, a corresponding electronic device and a corresponding computer storage medium which overcome the above problems or at least partially solve the above problems.

[0005] The present application discloses a target tracking method, which comprises:

[0006] obtaining a target detection frame region from a video stream; the target detection frame region includes a blocked candidate target object and / or a candidate target object with strong light;

[0007] determining an original target object matched with the candidate target object from the video stream based on the feature information of the candidate target object contained in the target detection frame region;

[0008] merging the candidate target object in the target detection frame region and the original target object to obtain a tracking result corresponding to the tracking number of the original target object.

[0009] Optionally, the video stream includes a plurality of tracking sequences corresponding to a plurality of different tracking numbers respectively, and the merging of the candidate target object in the target detection frame region and the original target object to obtain the tracking result corresponding to the tracking number of the original target object comprises:

[0010] obtaining a raw target tracking sequence of a raw target object and a candidate tracking sequence of the candidate target object, associating the candidate tracking sequence with the raw target tracking sequence;

[0011] determining that the tracking number of the candidate target object is the same as the tracking number of the raw target object, to determine a tracking result of the same tracking number as the tracking number of the raw target object.

[0012] Optionally, the determining, from the video stream, the raw target object matching the candidate target object based on the feature information of the candidate target object contained in the target detection frame region comprises:

[0013] obtaining feature information of a blocked candidate target object in the target detection frame region, and determining, from the video stream, the raw target object matching the blocked candidate target object based on the feature information of the blocked candidate target object, to find the raw target object before the blocked candidate target object is blocked.

[0014] And / or, obtaining feature information of a candidate target object with strong light in the target detection frame region, and determining, from the video stream, the raw target object matching the candidate target object with strong light based on the feature information of the candidate target object with strong light, to determine the raw target object in the case of light interference.

[0015] Optionally, the video stream comprises at least one blocked object in the target detection frame region; and the determining, from the video stream, the raw target object matching the blocked candidate target object based on the feature information of the blocked candidate target object comprises:

[0016] obtaining a blocked candidate target object with stable features from the at least one blocked object in the target detection frame region; the blocked candidate target object with stable features is displayed in the target detection region for more than a preset time length;

[0017] obtaining a target object disappearing for more than the preset time length in a historical video stream before the blocked candidate target object with stable features appears;

[0018] determining, based on average feature information of the blocked candidate target object with stable features and average feature information of the target object disappearing for more than the preset time length in the historical video stream, the raw target object matching the blocked candidate target object.

[0019] Optionally, the average feature information comprises appearance features; and determining the original target object matched with the occluded candidate target object based on the average feature information of the feature-stable occluded candidate target object and the average feature information of the target object disappeared in the historical video stream for more than the preset time length comprises:

[0020] If the appearance features of the feature-stable occluded candidate target object are similar to the appearance features of the target object disappeared in the historical video stream for more than the preset time length, and the feature-stable occluded candidate target object satisfies the preset speed and distance constraint condition, it is determined that the target object disappeared in the historical video stream for more than the preset time length is the original target object matched with the occluded candidate target object.

[0021] Optionally, the feature information of the candidate target object with strong light is used to determine the original target object matched with the candidate target object with strong light from the video stream, comprising:

[0022] A tracking sequence with light in the last frame image is obtained from the video stream.

[0023] The feature information of the candidate target object with strong light is used to determine the original target object matched with the candidate target object with strong light based on the feature information of the last frame image in the tracking sequence with light in the last frame image.

[0024] Optionally, the feature information comprises similarity; and the feature information of the candidate target object with strong light is used to determine the original target object matched with the candidate target object with strong light based on the feature information of the last frame image in the tracking sequence with light in the last frame image, comprising:

[0025] If the similarity between the candidate target object with strong light and the last frame image in the tracking sequence with light in the last frame image reaches a preset degree, it is determined that the target object in the detection frame region in the tracking sequence with light in the last frame image is the original target object matched with the candidate target object with strong light.

[0026] Optionally, the feature information comprises similarity, and the similarity comprises appearance similarity, position similarity and shape similarity.

[0027] The feature information of the candidate target object with strong light is used to determine the original target object matched with the candidate target object with strong light based on the feature information of the last frame image in the tracking sequence with light in the last frame image, further comprising:

[0028] If the appearance similarity between the candidate target object with strong light and the last frame image in the tracking sequence with light of the last frame image is lower than a preset degree, but the position similarity and the shape similarity between the candidate target object with strong light and the last frame image in the tracking sequence with light of the last frame image reach the preset degree, it is determined that the target object in the detection frame region in the tracking sequence with light of the last frame image is the original target object matched with the candidate target object with strong light.

[0029] The embodiment of the present application further discloses a target tracking device, which comprises:

[0030] A target detection frame region acquisition module is configured to acquire a target detection frame region from a video stream, wherein the target detection frame region comprises a candidate target object with occlusion and / or a candidate target object with strong light.

[0031] An original target object determination module is configured to determine an original target object matched with the candidate target object based on feature information of the candidate target object contained in the target detection frame region from the video stream.

[0032] A tracking result output module is configured to merge the candidate target object in the target detection frame region and the original target object to obtain a tracking result corresponding to the tracking number of the original target object.

[0033] Optionally, the video stream comprises a plurality of tracking sequences corresponding to a plurality of different tracking numbers respectively, and the tracking result output module comprises:

[0034] A tracking sequence association sub-module is configured to acquire an original target tracking sequence of the original target object and a candidate tracking sequence of the candidate target object, associate the candidate tracking sequence with the original target tracking sequence, determine the tracking number of the candidate target object as the same as the tracking number of the original target object, and determine a tracking result with the same tracking number as the tracking number of the original target object.

[0035] Optionally, the original target object determination module comprises:

[0036] A first original target object matching sub-module is configured to acquire feature information of the candidate target object with occlusion in the target detection frame region, determine an original target object matched with the candidate target object with occlusion from the video stream based on the feature information of the candidate target object with occlusion, and find the original target object before the occlusion of the candidate target object with occlusion.

[0037] The second original target object matching submodule is configured to acquire feature information of a candidate target object with strong light in the target detection frame region, and determine an original target object matched with the candidate target object with strong light from the video stream based on the feature information of the candidate target object with strong light, so as to determine the original target object in the case of light interference.

[0038] Optionally, the video stream includes at least one occluded object in the target detection frame region; and the first original target object matching submodule includes:

[0039] The candidate target object acquisition unit is configured to acquire a feature-stable occluded candidate target object from the at least one occluded object in the target detection frame region; and the feature-stable occluded candidate target object is displayed in the target detection region for more than a preset time length.

[0040] The target object determination unit is configured to acquire a target object that disappears for more than the preset time length in a historical video stream before the appearance of the feature-stable occluded candidate target object.

[0041] The first original target object determination unit is configured to determine an original target object matched with the occluded candidate target object based on average feature information of the feature-stable occluded candidate target object and average feature information of the target object that disappears for more than the preset time length in the historical video stream.

[0042] Optionally, the average feature information includes appearance features; and the first original target object determination unit includes:

[0043] The first original target object determination subunit is configured to determine the target object that disappears for more than the preset time length in the historical video stream as the original target object matched with the occluded candidate target object when the appearance features of the feature-stable occluded candidate target object are similar to the appearance features of the target object that disappears for more than the preset time length in the historical video stream, and the feature-stable occluded candidate target object satisfies a preset speed and distance constraint condition.

[0044] Optionally, the second original target object matching submodule includes:

[0045] The tracking sequence acquisition unit is configured to acquire a tracking sequence with light in the last frame image from the video stream.

[0046] The second original object determination unit is configured to determine an original target object matched with the candidate target object with strong light based on the feature information of the candidate target object with strong light and feature information of the last frame image with light in the tracking sequence of the last frame image.

[0047] Optionally, the feature information comprises a similarity degree; and the second original object determining unit comprises:

[0048] a second original target object determining subunit, configured to determine, when a similarity degree between the candidate target object with strong light and a last frame image in the tracking sequence with the last frame image having light reaches a preset degree, a target object in a detection frame region in the tracking sequence with the last frame image having light as an original target object matched with the candidate target object with strong light.

[0049] Optionally, the feature information comprises a similarity degree, and the similarity degree comprises an appearance similarity degree, a position similarity degree and a shape similarity degree; and the second original object determining unit further comprises:

[0050] the second original target object determining subunit is further configured to determine, when an appearance similarity degree between the candidate target object with strong light and a last frame image in the tracking sequence with the last frame image having light is lower than a preset degree, but a position similarity degree and a shape similarity degree between the candidate target object with strong light and the last frame image in the tracking sequence with the last frame image having light reach a preset degree, a target object in a detection frame region in the tracking sequence with the last frame image having light as an original target object matched with the candidate target object with strong light.

[0051] Embodiments of the present application further disclose an electronic device, comprising a processor, a memory, and a computer program stored in the memory and capable of running on the processor, and the computer program, when executed by the processor, implements steps of any of the target tracking methods.

[0052] Embodiments of the present application further disclose a computer readable storage medium, which stores a computer program, and the computer program, when executed by a processor, implements steps of any of the target tracking methods.

[0053] Embodiments of the present application have the following advantages:

[0054] In the embodiments of the present application, when multi-target tracking is performed, the candidate target objects that are occluded and / or the candidate target objects that have strong light in the target detection frame region can be tracked. The original target object matched with the candidate target object can be determined from the video stream based on the feature information of the candidate target object contained in the target detection frame region, so as to merge the candidate target object in the target detection frame region and the original target object, obtain the tracking result of the tracking number corresponding to the original target object, ensure that the candidate target object that is occluded and / or the candidate target object that has strong light has the same tracking number as the matched original target object, and perform long-term stable tracking on the original target object in the case of high occlusion and day-night change, thereby improving the accuracy of multi-target tracking, ensuring the tracking continuity of the target in the case of occlusion and at night, and realizing long-term stable tracking of the target. BRIEF DESCRIPTION OF DRAWINGS

[0055] Figure 1 is a flowchart of multi-target tracking in the related art;

[0056] Figure 2 is a step flowchart of an embodiment of a target tracking method of the present application;

[0057] Figure 3 is a step flowchart of another embodiment of a target tracking method of the present application;

[0058] Figure 4 is a flowchart of target tracking after target recovery according to an embodiment of the present application;

[0059] Figure 5 is a scene diagram of a vehicle being occluded according to an embodiment of the present application;

[0060] Figure 6 is a process diagram of vehicle light detection according to an embodiment of the present application;

[0061] Figure 7 is a structural block diagram of an embodiment of a target tracking device of the present application. DETAILED DESCRIPTION

[0062] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the present application will be further described in detail below with reference to the drawings and specific embodiments.

[0063] The intelligent design and planning of curb parking spaces can more flexibly solve the problem of parking difficulty. The curb parking space can be a parking space designed in the road, which refers to a place for temporary parking of motor vehicles set in the range of urban roads according to law, which includes a carriageway parking space and a sidewalk parking space.

[0064] Currently, the charging service for the on-street parking space can be realized by manual participation or video monitoring of vehicle parking and parking out by high-position camera. Specifically, the manual participation mode refers to that the charging personnel is responsible for patrolling the parking spaces in several blocks, and can take pictures when a vehicle is found to enter, and paste a charging two-dimensional code; the video monitoring mode is an intelligent unattended mode, which mainly uses the video captured by the high-position camera to automatically monitor the state of vehicle parking and driving out of the parking space, and charges according to the recognized license plate, at this time the video and image evidence can be retained. The video monitoring mode can improve the efficiency of urban management based on intelligent mode, and the charging service is more accurate, and the video and image evidence can be used for tracing.

[0065] Among them, the video monitoring mode needs to use video analysis, and in the scheme using video analysis, the tracking of the vehicle is mainly realized based on multi-target tracking. Referring to Figure 1 , a flowchart of multi-target tracking in the related art is shown, and multi-target tracking can usually be realized by a tracking by detection scheme based on a detection box (tracking by detection is a multi-target tracking algorithm, which is a method for long-term tracking of multiple targets using video, the input is video and detection box, and the output is the unique ID of each target object, i.e. tracking number).

[0066] It mainly associates the target objects in the front and back frames of the video through data association under the condition of existing detection boxes, to determine the same target objects. Specifically, data association can usually be calculated by calculating a similarity matrix, and the similarity usually uses appearance features, position relationships and shape size relationships as a measure, and a weighted average value of the three similarities is calculated; then KM algorithm (Kuhn-Munkres) is used for matching, and the tracking sequence is updated according to the matching result to maintain the life cycle of the sequence, and the tracking result is output.

[0067] However, the scheme using video analysis needs to track the vehicle for a long time and stably, and since the number of strings will cause incorrect charges and interruptions will cause resource loss, the long-term stable tracking does not allow the occurrence of string number and interruption, but the related technology of multi-target tracking is not suitable for tracking with low occlusion, short time, and needs to cross day and night, especially for tracking with high occlusion (i.e. severe occlusion), long time (e.g. long parking time of the vehicle), and needs to propose higher requirements for the implementation of tracking; and the algorithm is usually deployed in edge devices, and must use a lightweight scheme to improve the computing efficiency and reduce the cost, so the tracking algorithm for on-street parking is a lightweight, efficient and strong anti-interference long-term stable tracking algorithm.

[0068] Referring to Figure 2, shows a step flowchart of an embodiment of a target tracking method of the present application, and can specifically include the following steps:

[0069] Step 201, obtaining a target detection frame region from a video stream;

[0070] The embodiments of the present application are targeted at the characteristics of on-street parking applications, including the characteristics of high occlusion and long-time tracking of on-street parking, to effectively improve the accuracy of high occlusion and long-time tracking. The long-time stable tracking required to be implemented is mainly based on ensuring the invariability of the tracking number of the tracked target object, ensuring the continuity of tracking when the vehicle is occluded, and ensuring the tracking continuity at night.

[0071] In target tracking, the unique ID of each target object, i.e., the tracking number of the target object, can be tracked based on the video stream and the detection frame output to achieve tracking of the corresponding target object. When multiple target tracking is performed, the tracking number of the same target is the same in consecutive video frames, and each target will maintain a tracking sequence inside the algorithm. A tracking sequence can include the position, appearance feature, and other key information of the target in history.

[0072] For a new video frame, the appearance feature of the object in the video frame will be extracted according to the object detection frame, and the position and shape information of the object will be referred to, and similarity matching will be performed with all tracking sequences maintained by the algorithm, i.e., matching the target object that reappears in the new video frame with the target object in the historical tracking sequence. The overall similarity used in the matching is obtained by weighted average of the appearance feature, position similarity, and shape similarity. The object with high similarity will be considered as the same target and will be assigned the same tracking number. If the similarity of the object detected in the object detection frame with all tracking sequences is not high, the target will be assigned a new tracking number.

[0073] In order to ensure that the tracking number of the target object is unchanged, the object contained in the target detection region in the video stream needs to be processed. Therefore, in the case of high occlusion and / or at night, the target detection region to be processed can be the detection frame related to the two cases, i.e., the obtained target detection frame region can include the occluded candidate target object and / or the candidate target object with strong light, so as to process the occluded candidate target object and / or the candidate target object with strong light contained in the detection frame.

[0074] In actual application, the candidate target object with strong light can refer to a strong light object originally existing in the candidate target object itself, for example, a car light turned on by a vehicle, or can refer to a candidate target object reflected or directly irradiated by other strong light objects, i.e., the candidate target object exists strong light from other factors. For the strong light, it can be light with brightness exceeding a set brightness threshold. The set brightness threshold can be a brightness value too large to hinder the camera from shooting the original appearance (including shape, color, etc.) of the object. The brightness value can be affected by weather, external light, etc., and is mainly based on actual situation.

[0075] It should be noted that the video stream used for target tracking can be a received real-time video stream or a recorded video stream, which is not limited by the embodiments of the present application.

[0076] In step 202, based on the feature information of the candidate target object contained in the target detection frame region, an original target object matched with the candidate target object is determined from the video stream.

[0077] In the case of high occlusion of the vehicle and / or at night, the candidate target object contained in the detection frame and / or the candidate target object with strong light is processed, and the original target object matched with the candidate target object can be determined from the video stream to ensure the tracking number of the candidate target object in the case of high occlusion and / or strong light at night. The tracking number of the candidate target object can be the same as the tracking number of the determined matched original target object, i.e., the tracking number remains unchanged to ensure the stability of long-term tracking of the target object.

[0078] In the embodiments of the present application, based on the characteristics of the application of parking on the road, in order to cope with the short-term tracking number change caused by high occlusion, a recovery strategy for the target object in the occlusion process in the picture message can be added, and for the phenomenon of target tracking number change caused by night light, a light interference suppression algorithm can be added to correct the dramatic change of appearance similarity caused by night light irradiation to ensure the stability of the tracking number of the target object.

[0079] Specifically, based on the feature information of the candidate target object contained in the target detection region, the original target object matched with the candidate target object can be determined from the video stream to associate the tracking number of the matched original target object to the candidate target object, so that even in the case of high occlusion of the vehicle and / or at night, the tracking number of the original target object can remain unchanged.

[0080] The determination of the original target object can comprise: based on the feature information of the occluded candidate target object in the target detection frame region, searching for the original target object of the occluded candidate target object before being occluded from the video stream, and / or based on the feature information of the candidate target object with strong light in the target detection frame region, determining the original target object in the case of light interference from the video stream. It should be noted that if the on-street parking space of the vehicle is in the case of being occluded at night and strong light, the original target object determined based on the feature information of the occluded candidate target object and the feature information of the candidate target object with strong light in the target detection frame region can be the same target object, and the embodiments of the present application do not limit this.

[0081] In step 203, the candidate target object in the target detection frame region and the original target object are merged to obtain a tracking result corresponding to the tracking number of the original target object.

[0082] In order to keep the tracking number of the original target object unchanged, after determining the original target object matched with the candidate target object in the case of high occlusion and / or at night, the candidate target object in the target detection frame region and the original target object can be merged, which can mainly be represented as associating the tracking sequence of the candidate target object with the tracking sequence of the original target object, so that the tracking number of the candidate target object is the same as the tracking number of the original target object, so as to obtain a tracking result corresponding to the tracking number of the original target object.

[0083] The obtained tracking result is determined based on the tracking sequence after merging the tracking sequence of the candidate target object and the tracking sequence of the original target object, which can be the result obtained based on the tracking number corresponding to the original target object in the merged tracking sequence.

[0084] In the embodiments of the present application, when performing multi-target tracking, the occluded candidate target object and / or the candidate target object with strong light in the target detection frame region can be tracked, which can mainly be based on the feature information of the candidate target object contained in the target detection frame region to determine the original target object matched with the candidate target object from the video stream, to merge the candidate target object in the target detection frame region and the original target object, obtain a tracking result corresponding to the tracking number of the original target object, ensure that the occluded candidate target object and / or the candidate target object with strong light have the same tracking number as the matched original target object, and perform long-term stable tracking on the original target object in the case of high occlusion and day-night change, thereby improving the accuracy of multi-target tracking, ensuring the tracking continuity of the target being occluded and at night, and realizing long-term stable tracking of the target.

[0085] Reference Figure 3FIG. 3 shows a flowchart of steps of another embodiment of the target tracking method of the present application, which can specifically include the following steps:

[0086] In step 301, based on the feature information of the occluded candidate target object in the target detection frame region, the original target object before occlusion of the occluded candidate target object is searched from the video stream.

[0087] The embodiments of the present application are targeted at the characteristics of on-street parking applications, including the characteristics of high occlusion and long-time tracking of on-street parking, to effectively improve the accuracy of high occlusion and long-time tracking. The long-time stable tracking required to be achieved is mainly based on ensuring the invariability of the tracking number of the tracked target object, ensuring the continuity of tracking when the vehicle is occluded, and ensuring the tracking continuity at night.

[0088] In an embodiment of the present application, in order to cope with the short-term tracking number change caused by high occlusion, a recovery strategy for the target object in the occlusion process in the screen message can be added, which can be mainly based on the feature information of the occluded candidate target object in the target detection frame region, to search the original target object before occlusion of the occluded candidate target object from the video stream.

[0089] Specifically, the feature information of the occluded candidate target object in the target detection frame region can be obtained, and based on the feature information of the occluded candidate target object, the original target object matching the occluded candidate target object is determined from the video stream, to search the original target object before occlusion of the occluded candidate target object.

[0090] The video stream can include at least one occluded object in the target detection frame region. When the original target object matching the occluded candidate target object is determined from the video stream, the feature-stable occluded candidate target object can be obtained from the at least one occluded object in the target detection frame region. The feature-stable occluded candidate target object can refer to a target object that is displayed for more than a preset time in the target detection region. At this time, the target object that disappears for more than a preset time in the historical video stream before the appearance of the feature-stable occluded candidate target object can be obtained. Based on the average feature information of the feature-stable occluded candidate target object and the average feature information of the target object that disappears for more than a preset time in the historical video stream, the original target object matching the occluded candidate target object is determined.

[0091] Specifically, refer to Figure 4Fig. 6 shows a flowchart of a method for joining target tracking after target recovery according to an embodiment of the present application. The method mainly includes the following steps: S601, determining whether a candidate target object is blocked; S602, if the candidate target object is blocked, determining whether the candidate target object is a stable candidate target object; S603, if the candidate target object is a stable candidate target object, recovering a lost target; and S604, if the lost target is the same as the target before being blocked, merging the target and outputting a tracking result.

[0092] As an example, refer to Fig. 1, which shows a scene diagram of a vehicle being blocked according to an embodiment of the present application. In this scene, a vehicle A is parked on the roadside and is gradually blocked by a bus B. The detection frame of the vehicle A in the image will gradually become smaller until the detection frame disappears, and then the detection frame will gradually appear completely in the detection field of view. However, because the vehicle A is blocked by the bus B, the appearance feature of the target object (in this example, the vehicle A) will be incomplete, that is, contaminated. Therefore, when the tracked target object is blocked, the appearance feature of the target is unreliable, and the detection frame of the target after being blocked and reappearing is incomplete, that is, when the tracked target object is blocked, the position similarity and the shape similarity of the target are also unreliable. After the target object is blocked, the tracking number will usually be changed. Figure 5 As an example, refer to Fig. 1, which shows a scene diagram of a vehicle being blocked according to an embodiment of the present application. In this scene, a vehicle A is parked on the roadside and is gradually blocked by a bus B. The detection frame of the vehicle A in the image will gradually become smaller until the detection frame disappears, and then the detection frame will gradually appear completely in the detection field of view. However, because the vehicle A is blocked by the bus B, the appearance feature of the target object (in this example, the vehicle A) will be incomplete, that is, contaminated. Therefore, when the tracked target object is blocked, the appearance feature of the target is unreliable, and the detection frame of the target after being blocked and reappearing is incomplete, that is, when the tracked target object is blocked, the position similarity and the shape similarity of the target are also unreliable. After the target object is blocked, the tracking number will usually be changed. Figure 5 In actual applications, when the target object is blocked and then reappears, the shape of the reappearing target object may be squeezed, and at this time, the appearance feature of the corresponding reappearing target object will also change, that is, the multi-target tracking algorithm will determine that the current reappearing target object is not the same target as the target object before being blocked, and a new tracking number will be assigned to the reappearing target object. However, when the object is not blocked for a certain time length, the shape and appearance feature of the object will return to be similar to those before being blocked. At this time, the object similar to that before being blocked can be used to recover the original tracking number. Specifically, as shown in Fig. 2, a target recovery strategy is added to the original algorithm. When a new target appears for a stable time length, for example, reaches a preset time length, it can be considered that the features of the new target appearing after the preset time length are gradually stable compared with the original target object, that is, it can be determined that it is a stable candidate target object that is blocked, so as to process the stable candidate target object. It should be noted that the preset time length can be an empirical threshold value set based on the video frame rate and the average motion speed of the object, that is, it is mainly determined based on the experience value, for example, 20 frames. The present application is not limited in this regard.

[0093] Figure 1

[0094] ​​Specifically, the average feature information used to determine the original target object matched with the occluded candidate target object can include appearance features, and specifically can be the average value of the appearance features of the tracking sequence maintained by the algorithm. If the appearance features of the occluded candidate target object with stable features are similar to the appearance features of the target object disappeared in the historical video stream for more than a preset time length, and the occluded candidate target object with stable features meets the preset speed and distance constraint condition, it can be determined that the target object disappeared in the historical video stream for more than a preset time length is the original target object matched with the candidate target object, that is, the occluded candidate target object and the target object disappearing for more than a preset time length are identified as the same target.

[0095] The average feature is used to reduce the influence of the pollution of the appearance features of the target object when it is occluded, and to improve the accuracy of target tracking. The target recovery algorithm added by the average feature can effectively deal with the situation that the tracking number changes due to occlusion, that is, even if the target object changes temporarily during occlusion, the target object before occlusion can be recovered to track, thereby realizing long-term stable tracking. It does not cause resource consumption of the service.

[0096] It should be noted that, in the constraint condition of meeting the preset speed and distance, the constraint of speed and distance can mean that the same target object cannot move in a manner that does not conform to reality within a limited time, that is, the speed and distance enjoyed by the target object will be constrained. Specifically, the distance constraint can mean that the target object cannot move too far within a limited time, and the speed constraint can mean that the target object cannot move too fast within a limited time. The threshold of the constraint can be set according to the actual scene, and the embodiments of the present application do not limit this.

[0097] Step 302, determining the original target object in the case of light interference from the video stream based on the feature information of the candidate target object with strong light in the target detection frame region;

[0098] In an embodiment of the present application, in order to deal with the phenomenon of target tracking number change due to night light, a light interference suppression algorithm is added to correct the dramatic change of appearance similarity caused by night light irradiation, so as to ensure the stability of the tracking number of the target object. It can be mainly manifested as determining the original target object in the case of light interference from the video stream based on the feature information of the candidate target object with strong light in the target detection frame region.

[0099] Specifically, the original target object can be determined by obtaining the feature information of the candidate target object with strong light in the target detection frame region, and determining the original target object matching the candidate target object with strong light in the video stream based on the feature information of the candidate target object with strong light, so as to determine the original target object in the case of light interference and suppress the light interference.

[0100] In actual application, the candidate target object with strong light can refer to a strong light object originally existing in the candidate target object itself, such as a vehicle light, and can also refer to a candidate target object reflected or directly irradiated by other strong light objects, i.e., the candidate target object exists strong light from other factors. For strong light, it can be light with brightness exceeding a set brightness threshold. The set brightness threshold can be a brightness value that is too large to hinder the camera from capturing the original appearance (including shape, color, etc.) of the object. This brightness value can be affected by weather, external light, etc., and is mainly determined based on actual situation.

[0101] When determining the original target object matching the candidate target object with strong light in the video stream, a tracking sequence with light in the last frame image can be obtained from the video stream. The tracking sequence can be a tracking sequence before the vehicle light is turned off, i.e., before the existing strong light weakens or even disappears. At this time, the original target object matching the candidate target object with strong light can be determined based on the feature information of the candidate target object with strong light and the feature information of the last frame image in the tracking sequence with light in the last frame image.

[0102] Specifically, the feature information used to determine the original target object matching the candidate target object with strong light can include similarity. At this time, when determining the original target object matching the candidate target object with strong light, for the detection frame with strong light and the tracking sequence with light in the last frame, the original target object matching the candidate target object with strong light can be matched according to the determination of similarity. If the similarity between the candidate target object with strong light and the last frame image in the tracking sequence with light in the last frame image reaches a preset degree, the target object in the detection frame region in the tracking sequence with light in the last frame image is determined as the original target object matching the candidate target object with strong light, i.e., the tracking number of the candidate target object with strong light is determined to be the same as the tracking number of the tracking sequence with light in the last frame image, and the tracking sequence of the candidate target object with strong light is associated with it.

[0103] The similarity can include appearance similarity, position similarity, and shape similarity. The similarity can be based on, for example, Figure 1 or Figure 4The similarity calculation implementation shown, the similarity usually uses appearance features, position relationship and shape size relationship as the measurement, and can be calculated using the weighted average of the three similarities.

[0104] In a preferred embodiment, since the matching of the target object appearing in the new video frame and the target object in the historical tracking sequence is mainly based on the overall similarity obtained by the weighted average of the appearance similarity, position similarity and shape similarity, the on-off of the night vehicle light will cause the appearance features to change dramatically, so that the similarity of the same target will decrease and cause the matching to fail, and thus the tracking number will also frequently change. After the vehicle light detection is added, when the vehicle light state changes, the overall similarity can be calculated using the position similarity and shape similarity, and the influence of the appearance features can be ignored, so that the tracking number is ensured not to change. Then, in addition to matching the original target object based on the similarity, i.e. in addition to the normal matching, for the detection frame that fails to match, assuming that there is a target object in the detection frame that has high position similarity and shape similarity but low appearance similarity, the target object can be recorded at this time, and the target object can be mainly marked as a position close target, i.e. the position of the target object is close to the target (such as the vehicle light).

[0105] Specifically, when the appearance similarity between the candidate target object with strong light and the last frame image of the tracking sequence with light in the last frame image is lower than a preset degree, but the position similarity and the shape similarity between the candidate target object with strong light and the last frame image of the tracking sequence with light in the last frame image reach the preset degree, it can be determined that the target object in the detection frame region in the tracking sequence with light in the last frame image is the original target object matched with the candidate target object with strong light, i.e. it can be determined that the tracking number of the candidate target object with strong light is the same as the tracking number of the tracking sequence with light in the last frame image, and the tracking sequence of the candidate target object with strong light is associated with it.

[0106] As an example, in addition to facing the situation that the vehicle is blocked, the related detection service of the in-road parking also needs to face the phenomenon of interference by the night vehicle light, such as the vehicle light turned on when the vehicle starts at night and the vehicle light turned off when the vehicle stops. The position of the vehicle light and the camera will change during the driving of the vehicle, and sometimes the camera will be directly irradiated, which will cause strong light spots in the image collected by the camera. These light spots will cause the appearance features of the vehicle in the image to change dramatically, so that the tracking number of the vehicle changes. Especially during the process of the vehicle entering and leaving the field, if the tracking number of the vehicle changes, it is very likely that the capture of this event will be lost, which will cause the generation of resource waste.

[0107] When the car light is on and directly shines on the camera, there will be a particularly high brightness area at the position of the car light, i.e. the image RGB values of the area are particularly high, usually greater than a certain threshold (e.g. 240), while in the image, the white color corresponds to the RGB pixel of 255, but the white color seen in nature is usually not the pure white in the image sense, and usually does not appear particularly high in three-channel pixel value. Only when the strong light directly shines on the camera, the pixel value will be particularly high.

[0108] For the detection frame that is not matched successfully, in order to mark the position close to the target based on the high position similarity and shape similarity, the car light can be detected in the manner shown in Figure 6 . Specifically, it is determined whether the position similarity and shape similarity of the strong light object are car lights, and if they are car lights, it is determined that the target object with light in the last frame is the same target as the target object with light. In this case, the detection frame can be scaled to a uniform size, and the pixel value in the image can be detected. When the pixel value is greater than a certain threshold, a mask is generated. The connected region of the mask is calculated, and whether it is a car light is determined based on the size information and position information of the connected region.

[0109] In a specific implementation, the detection frame can be scaled to a uniform size, specifically, the image block is resized to a uniform size, so that the size of all target objects can remain consistent and is not affected by the actual size of the object. At this time, the pixel value in the image can be detected based on the scaled detection frame. According to whether the pixel value is greater than a certain threshold (usually 240), a mask is generated. Then, morphological operations (including a morphological operation or a b morphological operation, such as erosion and dilation) can be performed on the generated mask, and the connected region of the mask is calculated. Whether it is a car light is determined by the size, aspect ratio and relative position of the connected region in the image. In addition to normal matching, for the detection frame that is not matched, it can be checked whether there is a position close to the target based on the manner shown in Figure 6 . If there is, it can be considered as a target affected by the light, and it is still associated with the related tracking sequence.

[0110] In this example, this way can effectively deal with night light interference and realize long-term day and night tracking. For example, in a certain video, if it does not perform night light suppression, it can be seen that when the vehicle with tracking number 1 leaves, the color changes obviously due to the car light directly shining on the camera, and thus the tracking number also changes. However, assuming that the light interference detection algorithm is added in another video, the tracking number of the vehicle with tracking number 1 does not change during the leaving process, so that the leaving time can be effectively captured and the loss of assets can be avoided.

[0111] It should be noted that the light interference suppression algorithm is not limited by the time setting, and is not limited to night, but will also be used to suppress light in rainy weather during the day. The embodiments of the present application do not limit this.

[0112] Step 303, associate the candidate tracking sequence with the original target tracking sequence, and determine the tracking result with the same tracking number as the tracking number of the original target object.

[0113] In an embodiment of the present application, in order to keep the tracking number of the original target object unchanged, after determining the original target object matched with the candidate target object under the condition of high occlusion and / or night, the candidate target object and the original target object within the target detection frame region can be merged.

[0114] Specifically, the video stream can include a plurality of tracking sequences corresponding to a plurality of different tracking numbers respectively. The merging of the candidate target object and the original target object can be performed by obtaining the original target tracking sequence of the original target object and the candidate tracking sequence of the candidate target object, associating the candidate tracking sequence with the original target tracking sequence, so that the tracking number of the candidate target object is the same as the tracking number of the original target object, to determine the tracking result with the same tracking number as the tracking number of the original target object.

[0115] The obtained tracking result is determined based on the merged tracking sequence of the tracking sequence with the candidate target object and the tracking sequence of the original target object, which can be the result obtained by tracking based on the corresponding tracking number of the original target object in the merged tracking sequence.

[0116] In the embodiments of the present application, according to the characteristics of the on-street parking application scenario, by adding the recovery strategy and the night light interference suppression algorithm in the multi-target tracking algorithm, long-term stable tracking of the vehicle under high occlusion in the roadside parking can be effectively realized. The above-mentioned recovery strategy and night light interference suppression scheme can be used simultaneously or any one of them based on actual needs, to ensure the accuracy of the roadside parking service and avoid loss, thereby improving the intelligent level of the overall service, reducing personnel participation, and improving the efficiency of city operation; and long-term stable tracking of the original target object under high occlusion and day-night change, improves the accuracy of multi-target tracking while ensuring the tracking continuity of the target under occlusion and at night, and realizes long-term stable tracking of the target.

[0117] It should be noted that, for the method embodiments, the series of acts combined is described for simplicity, but those skilled in the art should know that the application embodiments are not limited to the order of the acts described, because according to the application embodiments, certain steps can be performed in other orders or at the same time. Secondly, those skilled in the art should know that the embodiments described in the specification are all preferred embodiments, and the acts involved are not necessarily the application embodiments.

[0118] Referring to Figure 7 , a structural block diagram of an object tracking device embodiment of the application is shown, which can specifically include the following modules:

[0119] The target detection frame area acquisition module 701 is configured to acquire a target detection frame area from a video stream; the target detection frame area includes a candidate target object that is occluded and / or a candidate target object that exists in strong light;

[0120] The original target object determination module 702 is configured to determine an original target object that matches a candidate target object in the target detection frame area from the video stream based on feature information of the candidate target object;

[0121] The tracking result output module 703 is configured to merge the candidate target object in the target detection frame area and the original target object to obtain a tracking result corresponding to a tracking number of the original target object.

[0122] In an embodiment of the application, the video stream includes a plurality of tracking sequences corresponding to a plurality of different tracking numbers respectively, and the tracking result output module 703 can include the following sub-modules:

[0123] The tracking sequence association sub-module is configured to acquire an original target tracking sequence of an original target object and a candidate tracking sequence of a candidate target object, associate the candidate tracking sequence with the original target tracking sequence, determine a tracking number of the candidate target object to be the same as a tracking number of the original target object, and determine a tracking result corresponding to the same tracking number as the tracking number of the original target object.

[0124] In an embodiment of the application, the original target object determination module 702 can include the following sub-modules:

[0125] The first original target object matching sub-module is configured to acquire feature information of an occluded candidate target object in the target detection frame area, determine an original target object that matches the occluded candidate target object from the video stream based on the feature information of the occluded candidate target object, and find an original target object before the occlusion of the occluded candidate target object.

[0126] The second original target object matching submodule is configured to acquire feature information of a candidate target object with strong light in the target detection frame region, and determine an original target object matched with the candidate target object with strong light from the video stream based on the feature information of the candidate target object with strong light, so as to determine the original target object in the case of light interference.

[0127] In an embodiment of the present application, the video stream includes at least one occluded object in the target detection frame region; the first original target object matching submodule can include the following units:

[0128] The candidate target object acquisition unit is configured to acquire a feature-stable occluded candidate target object from the at least one occluded object in the target detection frame region; the feature-stable occluded candidate target object is displayed in the target detection region for more than a preset time length;

[0129] The target object determination unit is configured to acquire a target object that disappears for more than the preset time length in a historical video stream before the appearance of the feature-stable occluded candidate target object;

[0130] The first original target object determination unit is configured to determine an original target object matched with the occluded candidate target object based on average feature information of the feature-stable occluded candidate target object and average feature information of the target object that disappears for more than the preset time length in the historical video stream.

[0131] In an embodiment of the present application, the average feature information includes appearance features; the first original target object determination unit can include the following subunit:

[0132] The first original target object determination subunit is configured to determine the target object that disappears for more than the preset time length in the historical video stream as the original target object matched with the occluded candidate target object when the appearance features of the feature-stable occluded candidate target object are similar to the appearance features of the target object that disappears for more than the preset time length in the historical video stream, and the feature-stable occluded candidate target object satisfies a preset speed and distance constraint condition.

[0133] In an embodiment of the present application, the second original target object matching submodule can include the following units:

[0134] The tracking sequence acquisition unit is configured to acquire a tracking sequence with light from the video stream, the last frame image of the tracking sequence having light;

[0135] The second original object determination unit is configured to determine an original target object matched with the candidate target object with strong light based on feature information of the candidate target object with strong light and feature information of the last frame image in the tracking sequence with light of the last frame image.

[0136] In an embodiment of the present application, the feature information comprises a similarity degree; and the second original object determination unit can comprise the following sub-units:

[0137] The second original target object determination sub-unit is configured to determine a target object in a detection frame region in the tracking sequence with light of the last frame image as the original target object matched with the candidate target object with strong light when the similarity degree between the candidate target object with strong light and the last frame image in the tracking sequence with light of the last frame image reaches a preset degree.

[0138] In an embodiment of the present application, the feature information comprises a similarity degree, and the similarity degree comprises an appearance similarity degree, a position similarity degree and a shape similarity degree; and the second original object determination unit can further comprise the following sub-units:

[0139] The second original target object determination sub-unit is further configured to determine a target object in a detection frame region in the tracking sequence with light of the last frame image as the original target object matched with the candidate target object with strong light when the appearance similarity degree between the candidate target object with strong light and the last frame image in the tracking sequence with light of the last frame image is lower than a preset degree, but the position similarity degree and the shape similarity degree between the candidate target object with strong light and the last frame image in the tracking sequence with light of the last frame image reach the preset degree.

[0140] For the device embodiment, it is basically similar to the method embodiment, so the description is relatively simple, and the related parts refer to the part of the method embodiment.

[0141] The present application further provides an electronic device, comprising:

[0142] The electronic device comprises a processor, a memory and a computer program stored in the memory and capable of running on the processor. When the computer program is executed by the processor, each process of the target tracking method embodiment is realized, and the same technical effect is achieved. To avoid repetition, no further description is given here.

[0143] The embodiment of the present application further provides a computer readable storage medium, and the computer readable storage medium stores a computer program. The computer program is executed by a processor to implement each process of the target tracking method embodiment and achieve the same technical effects. To avoid repetition, details are not described herein.

[0144] Each embodiment in the specification is described in a progressive manner, and each embodiment focuses on the difference from other embodiments. The same or similar parts between each embodiment are referred to each other.

[0145] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, device or computer program product. Therefore, the embodiments of the present application can be in the form of a complete hardware embodiment, a complete software embodiment or an embodiment combining software and hardware aspects. Moreover, the embodiments of the present application can be in the form of a computer program product implemented on one or more computer usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer usable program code.

[0146] The embodiments of the present application are described with reference to flowcharts and / or block diagrams according to the method, terminal device (system) and computer program product of the embodiments of the present application. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, and the combination of the flows and / or blocks in the flowcharts and / or block diagrams can be implemented by computer program instructions. These computer program instructions can be provided to a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing terminal device to produce a machine, so that the instructions executed by the computer or other programmable data processing terminal device produce a device for implementing the functions specified in the flowcharts and / or block diagrams. Figure 1 one or more flows and / or blocks Figure 1 an apparatus that carries out the functions specified in one or more flows and / or blocks.

[0147] These computer program instructions can also be stored in a computer readable memory that can guide the computer or other programmable data processing terminal device to work in a specific way, so that the instructions stored in the computer readable memory produce a product including instruction apparatus, which implements the functions specified in the flowcharts and / or block diagrams. Figure 1 one or more flows and / or blocks Figure 1 an apparatus that carries out the functions specified in one or more flows and / or blocks.

[0148] These computer program instructions can also be loaded into a computer or other programmable data processing terminal device, so that a series of operation steps are performed on the computer or other programmable terminal device to produce a computer implemented process, so that the instructions executed on the computer or other programmable terminal device provide a process for implementing the functions specified in the flowcharts and / or block diagrams. Figure 1one or more processes and / or blocks Figure 1 the steps of a function specified in one or more blocks.

[0149] While preferred embodiments of the application have been described, those skilled in the art will appreciate that other modifications and changes can be made thereto without departing from the basic inventive concepts as recited in the claims. Accordingly, the claims are intended to cover all such modifications and changes as fall within the scope of the application.

[0150] Finally, it is to be understood that the phraseology or terminology employed herein, such as "first" and "second", etc., are for descriptive purposes only and should not be construed to be limiting unless otherwise indicated. It is to be understood that the terms "including", "comprising", or "having" contain for the purposes of disclosure an open term such that the methods or compositions described can include some other elements or steps not expressly named or inherent to such methods or compositions. Thus, it is intended that various modifications and changes can be made by those skilled in the art to the particular embodiments disclosed without departing from the scope of this application. Accordingly, reference should be made to the appended claims as indicating the scope of the application.

[0151] The above provides a target tracking method, a target tracking device, a corresponding electronic device and a corresponding computer storage medium, the principles and implementation modes of the application are described by applying specific examples in the present text, the above example is only used to help understand the method of the application and its core idea; at the same time, for those skilled in the art, according to the idea of the application, the specific implementation mode and the application range will be changed, and the above description should not be understood as the limitation of the application.

Claims

1. A target tracking method characterized by, The method comprises: obtaining a target detection frame region from a video stream; the target detection frame region comprises a candidate target object that is blocked and / or a candidate target object that exists in strong light; the video stream comprises at least one blocked object in the target detection frame region; determining a target object that matches the candidate target object from the video stream based on feature information of the candidate target object contained in the target detection frame region; finding the target object that matches the candidate target object from the video stream based on feature information of the candidate target object that is blocked in the target detection frame region, and / or determining the target object that matches the candidate target object from the video stream based on feature information of the candidate target object that exists in strong light in the target detection frame region; the finding the target object that matches the candidate target object from the video stream based on feature information of the candidate target object that is blocked in the target detection frame region comprises: obtaining a candidate target object that is blocked and has stable features from the at least one blocked object in the target detection frame region; the candidate target object that is blocked and has stable features is displayed in the target detection frame region for more than a preset time length; obtaining a target object that disappears for more than the preset time length in a historical video stream before the candidate target object that is blocked and has stable features appears; determining the target object that matches the candidate target object that is blocked based on average feature information of the candidate target object that is blocked and has stable features and average feature information of the target object that disappears for more than the preset time length in the historical video stream; the determining the target object that matches the candidate target object from the video stream based on feature information of the candidate target object that exists in strong light in the target detection frame region comprises: obtaining a tracking sequence in which the last frame image has light from the video stream; determining the target object that matches the candidate target object that exists in strong light based on feature information of the candidate target object that exists in strong light and feature information of the last frame image in the tracking sequence in which the last frame image has light. merging the candidate target object in the target detection frame region and the target object to obtain a tracking result corresponding to a tracking number of the target object.

2. The method of claim 1, wherein, The video stream comprises a plurality of tracking sequences corresponding to a plurality of different tracking numbers respectively; the merging the candidate target object in the target detection frame region and the target object to obtain a tracking result corresponding to a tracking number of the target object comprises: obtaining a target tracking sequence of the target object and a candidate tracking sequence of the candidate target object, associating the candidate tracking sequence with the target tracking sequence, and determining a tracking number of the candidate target object to be the same as a tracking number of the target object to determine a tracking result corresponding to the tracking number of the target object. The determining the target object that matches the candidate target object from the video stream based on feature information of the candidate target object contained in the target detection frame region comprises:

3. The method according to claim 1 or 2, characterized in that, ​ obtain feature information of the occluded candidate target object in the target detection frame region, and determine a target object matched with the occluded candidate target object from the video stream based on the feature information of the occluded candidate target object, so as to find the original target object of the occluded candidate target object before the occlusion; and / or, obtain feature information of the candidate target object with strong light in the target detection frame region, and determine a target object matched with the candidate target object with strong light from the video stream based on the feature information of the candidate target object with strong light, so as to determine the original target object in the case of light interference.

4. The method of claim 1, wherein, The average feature information includes appearance features; and the determination of the original target object matched with the occluded candidate target object based on the average feature information of the feature-stable occluded candidate target object and the average feature information of the target object disappeared for more than the preset time length in the historical video stream includes: If the appearance features of the feature-stable occluded candidate target object are similar to the appearance features of the target object disappeared for more than the preset time length in the historical video stream, and the feature-stable occluded candidate target object satisfies the preset speed and distance constraint condition, the target object disappeared for more than the preset time length in the historical video stream is determined as the original target object matched with the occluded candidate target object.

5. The method of claim 1, wherein, The feature information includes similarity; and the determination of the original target object matched with the candidate target object with strong light based on the feature information of the candidate target object with strong light and the feature information of the last image in the tracking sequence with light in the last image includes: If the similarity of the candidate target object with strong light to the last image in the tracking sequence with light in the last image reaches a preset degree, the target object in the detection frame region in the tracking sequence with light in the last image is determined as the original target object matched with the candidate target object with strong light.

6. The method according to claim 1 or 5, characterized in that, The feature information includes similarity, and the similarity includes appearance similarity, position similarity, and shape similarity; The determination of the original target object matched with the candidate target object with strong light based on the feature information of the candidate target object with strong light and the feature information of the last image in the tracking sequence with light in the last image further includes: If the appearance similarity of the candidate target object with strong light to the last image in the tracking sequence with light in the last image is lower than a preset degree, but the position similarity and the shape similarity of the candidate target object with strong light to the last image in the tracking sequence with light in the last image reach a preset degree, the target object in the detection frame region in the tracking sequence with light in the last image is determined as the original target object matched with the candidate target object with strong light.

7. A target tracking device, characterized by The device includes: The target detection frame region acquisition module is configured to acquire a target detection frame region from a video stream; the target detection frame region includes a blocked candidate target object and / or a candidate target object with strong light; and the video stream includes at least one blocked object in the target detection frame region. The original target object determination module is configured to determine an original target object matching the candidate target object from the video stream based on feature information of the candidate target object included in the target detection frame region; find the original target object before the blocked candidate target object is blocked from the video stream based on feature information of the blocked candidate target object in the target detection frame region; and determine the original target object in a light interference case from the video stream based on feature information of the candidate target object with strong light in the target detection frame region. The finding of the original target object before the blocked candidate target object is blocked from the video stream based on the feature information of the blocked candidate target object in the target detection frame region includes: acquiring a blocked candidate target object with stable features from the at least one blocked object in the target detection frame region; the blocked candidate target object with stable features is displayed in the target detection frame region for more than a preset time length; acquiring a target object disappearing for more than the preset time length in a historical video stream before the blocked candidate target object with stable features appears; and determining the original target object matching the blocked candidate target object based on average feature information of the blocked candidate target object with stable features and average feature information of the target object disappearing for more than the preset time length in the historical video stream. The determination of the original target object in the light interference case from the video stream based on the feature information of the candidate target object with strong light in the target detection frame region includes: acquiring a tracking sequence with light in a last frame image from the video stream; and determining the original target object matching the candidate target object with strong light based on feature information of the candidate target object with strong light and feature information of the last frame image in the tracking sequence with light in the last frame image. The tracking result output module is configured to merge the candidate target object in the target detection frame region and the original target object to obtain a tracking result corresponding to the tracking number of the original target object.

8. An electronic device, comprising: The computer program is stored on the computer readable storage medium and is executed by the processor to implement the steps of the target tracking method according to any one of claims 1-6. The computer program is stored on the computer readable storage medium and is executed by the processor to implement the steps of the target tracking method according to any one of claims 1-6.

9. A computer-readable storage medium, characterized in that, ​

Images