Target detection method, device and electronic device

Abstract

The application provides a target detection method and device and electronic equipment. In the embodiment, a target group matching method is used to associate and match a plurality of rounds of collected target objects, such as road diseases / road assets, and the like, to automatically realize abnormal detection of the target objects, such as road diseases / road assets, and the like. Compared with manual auditing, the efficiency of road maintenance can be greatly improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of artificial intelligence, in particular to a target detection method and device and electronic equipment. BACKGROUND

[0002] At present, in the road detection application scenario, more manual auditing is usually used to locate the target objects on the road. Here, the target objects are, for example, road abnormalities (such as road cracks, pits, etc., collectively referred to as road diseases), road assets (such as barriers, speed bumps, road signboards such as speed limit signboards, etc., collectively referred to as road assets), etc., which are not specifically limited by the present application. However, the use of manual auditing greatly reduces the efficiency of road maintenance. SUMMARY

[0003] The present application provides a target detection method, device and electronic equipment to improve the efficiency of road maintenance.

[0004] The technical scheme provided by the present application includes:

[0005] A target detection method, applied to an electronic device, comprising:

[0006] determining a first target object group and a second target object group to be matched; the target objects contained in the first target object group are obtained by analyzing road images collected in a current inspection; the target objects contained in the second target object group are obtained by analyzing road images collected in a last historical inspection; the current inspection and the last historical inspection are performed on the same at least one road section;

[0007] For each target object in the first target object group, a matching object having a matching relationship with the target object is determined from the second target object group;

[0008] According to the road key data corresponding to each target object in the first target object group and the road key data corresponding to the matching object having a matching relationship with the target object in the second target object group, an abnormality detection of the target object is performed.

[0009] A target detection device, applied to an electronic device, comprising:

[0010] A determining unit configured to determine a first target object group and a second target object group to be matched;

[0011] A matching unit configured to, for each target object in the first target object group, determine a matching object having a matching relationship with the target object from the second target object group;

[0012] The detection unit is configured to perform abnormality detection on the target object according to road key data corresponding to each target object in the first target object group and road key data corresponding to a matching object having a matching relationship with the target object in the second target object group.

[0013] An electronic device includes a processor and a machine-readable storage medium;

[0014] The machine-readable storage medium stores machine executable instructions executable by the processor;

[0015] The processor is configured to execute the machine executable instructions to implement the steps in the above method.

[0016] As can be seen from the above technical solutions, in the embodiment, the target group matching method is used to associate and match the target objects such as road diseases / road assets collected multiple times, to automatically realize abnormality detection of the target objects such as road diseases / road assets, etc. Compared with manual review, the efficiency of road maintenance can be greatly improved. BRIEF DESCRIPTION OF DRAWINGS

[0017] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present disclosure and serve to explain the principles of the present disclosure together with the specification.

[0018] Figure 1 A method flowchart provided for the embodiment of the present application;

[0019] Figure 2 A system structure diagram provided for the embodiment of the present application;

[0020] Figure 3 An implementation flowchart of step 102 provided for the embodiment of the present application;

[0021] Figure 4 A corresponding relationship diagram provided for the embodiment of the present application;

[0022] Figure 5 An apparatus structure diagram provided for the embodiment of the present application;

[0023] Figure 6 An apparatus hardware structure diagram provided for the embodiment of the present application. DETAILED DESCRIPTION

[0024] In the embodiment, the target group matching method is used to detect the target objects on the road, effectively improve the matching accuracy, reduce the missed matching, optimize the road maintenance resources, and reduce the road maintenance cost.

[0025] To solve the above problems, the embodiment provides a target detection method. To make the method provided by the application easier to understand, the method provided by the application is described in detail below in combination with the drawings and the embodiments:

[0026] Referring to Figure 1 , Figure 1 The method flowchart provided by the embodiment of the application is shown. The flowchart can be applied to an electronic device. In an example, the electronic device here can be a back-end server, and the embodiment is not specifically limited.

[0027] For ease of description, the system involved in the flowchart shown in Figure 1 is described as an example:

[0028] Referring to Figure 2 , Figure 2 The system structure diagram provided by the embodiment of the application is shown. As shown in Figure 2 , the system mainly includes a front-end device and a back-end server (Back-end Serve). As an example, the front-end device and the back-end server described above can be connected through a wireless network, Figure 2 The wireless network is exemplified.

[0029] Optionally, the front-end device at least includes a smart camera (Smart Camera) and a positioning device. The smart camera is used to collect road images (such as image frames in a road inspection video stream) of a road scene in real time when inspecting a road, and to analyze and process the road images in real time by using a processor embedded therein, such as detecting and tracking target objects (road disease targets / road asset targets, etc.) in the field of view. The positioning device (such as GPS or Beidou, etc.) obtains the latitude and longitude of the inspection vehicle and associates it to the target object collected.

[0030] The front-end device transmits target object information corresponding to the target object to the back-end server described above. Here, the road data can include structured information, such as the inspection time when the target object is inspected, the category of the target object such as a gap, a railing, etc., the image area position of the target object in the image, the physical size of the target object, the position of the target object such as latitude and longitude information, etc., and the embodiment is not specifically limited.

[0031] When the back-end server receives the target object information, it will associate and match the target objects on the road, such as disease / asset targets, based on the received target object information, and track the target objects, and perform abnormal detection on the target objects based on the associated matching information, as shown in the flowchart shown in Figure 1 .

[0032] As shown in Figure 1 , the flowchart can include the following steps:

[0033] Step 101, determining a first target object group and a second target object group to be matched.

[0034] Optionally, in the embodiment, the information of the target objects such as the disease targets or the road asset targets obtained by analyzing the road images collected in the current inspection is used to select a certain number of target objects such as 20 target objects satisfying the preset grouping condition as the first target object group. It should be noted that there can be multiple target object groups obtained by analyzing the road images collected in the current inspection, and the first target object group is one of the target object groups.

[0035] Optionally, in the embodiment, the second target object group is obtained by analyzing the road images collected in the last historical inspection, and is one of the target object groups (also referred to as historical object groups) determined in the last historical inspection. The target objects in any historical object group satisfy the preset grouping condition.

[0036] Therefore, as an embodiment, the determination of the first target object group and the second target object group to be matched can include: obtaining the first target object group by analyzing the road images collected in the current inspection; and determining the second target object group satisfying the grouping matching condition with the first target object group from the historical object groups corresponding to the last historical inspection. Here, the grouping matching condition can be set according to actual needs, such as: the target objects contained therein are objects under the same road configuration, etc., and the embodiment is not specifically limited.

[0037] In the above description, the road configuration is the road post number or the mileage post number. The road post number is a number set according to the change of topographic features, which is convenient for the reference positioning of the slope of various pipelines buried under the roadbed and inspection wells. The difference between the post numbers represents the distance between two points, which is usually in units of “meters”. The mileage post number is a post number set by the traffic management department after the completion of the road, which belongs to traffic signs. It is usually in units of “kilometers”, and one post is set every kilometer according to the length of the road center line, and the mileage post number increases sequentially from the starting point to the ending point of the road. The difference between the mileage post numbers represents the distance between two points.

[0038] As an embodiment, the target objects in the first target object group or the second target object group satisfy the following preset grouping condition:

[0039] The target objects in the same group are objects under the same road configuration; here, the road configuration is as described above; and / or,

[0040] The position deviation of the first target object in the same group from other target objects satisfies the preset distance, wherein the first target object in any group is the object that is first inspected in the group; and / or,

[0041] The target objects in the same group are objects whose positions in the same road segment satisfy preset proximity conditions.

[0042] In step 102, for each target object in the first target object group, a matching object having a matching relationship with the target object is determined from the second target object group.

[0043] Optionally, to facilitate determination of the corresponding matching object for the target object in the first target object group, different scoring dimensions such as longitude and latitude, inspection direction, scene (whether elevated), physical size, lane in the image, target category, and the like are preset, and the matching relationship between each target object in the first target object group and each target object in the second target object group is considered from the perspective of each dimension to determine the matching object having a matching relationship with the target object in the first target object group from the second target object group. Figure 3 By way of example, further description is omitted here.

[0044] In step 103, according to the road key data corresponding to each target object in the first target object group and the road key data corresponding to the matching object having a matching relationship with the target object in the second target object group, abnormality detection of the target object is performed.

[0045] Optionally, in this embodiment, step 103 has many implementation manners when specifically implemented. For example, as an embodiment, for each target object in the first target object group, the road key data corresponding to the target object and the road key data corresponding to the matching object are compared, if the two do not satisfy the preset proximity condition such as being different, the road key data corresponding to the target object is recorded, and according to the road key data corresponding to the target object and the road key data corresponding to the matching object, it is detected whether the abnormality repair condition such as whether the crack is larger is satisfied, if so, an abnormality alarm is triggered.

[0046] As another embodiment, for each target object in the first target object group, the target object can be associated to the recorded global target sequence. For example, for each target object in the first target object group, if there is no matching object of the target object in the global target sequence (for example, the target object has no matching object), a road key data corresponding to the target object is newly created in the global target sequence; and if there is a matching object of the target object in the global target sequence, when the road key data corresponding to the target object does not satisfy a preset similar condition (for example, changes including but not limited to area expansion, severity deepening, category deterioration, disease repair, etc.) compared with the road key data corresponding to the matching object, the road key data corresponding to the target object can be updated to the recorded global target sequence. Of course, if the above satisfies the above preset similar condition, a deduplication mark is set for the road key data corresponding to the target object, so as to subsequently deduplicate / delete the data with the deduplication mark.

[0047] In the embodiment, the global refers to all road sections that need to be maintained by the maintenance unit / user. Alternatively, in the embodiment, the above global can be divided into different sub-domains according to road sections or stake numbers, and the system allocates a certain amount of storage units for each sub-domain to store the road key data of the corresponding target object. Finally, the target new creation, data update and target deletion of all targets of the corresponding global can be realized according to the sub-domain number and the storage unit number, and it is also convenient for the user to realize the query and access of the corresponding global target.

[0048] Based on the description of the global target sequence as above, the embodiment based on the global target sequence can realize the closed-loop management of the target object from the occurrence (first detection), evolution (category, area, severity change, etc.) to maintenance (repair) and other links, which is beneficial to the maximization of road maintenance efficiency.

[0049] Alternatively, in the embodiment, based on the global target sequence, when it is found that a target object starts to evolve (category, area, severity change, etc.) after the occurrence (first detection), the target object reaching a certain severity is automatically processed by a single dispatch to perform related maintenance (repair).

[0050] Alternatively, in the embodiment, for the target object that has been marked for repair or maintenance processing (after the dispatch processing), the target object in the global target sequence is set with a deletion mark after a certain number of rounds, such as N rounds of inspection (N is set according to actual needs), so as to timely delete the target object set with the deletion mark, realize the retention of necessary useful information on the basis of the above closed-loop management, maximize the saving of storage resources, reduce the artificial audit cost, and improve the intelligent management level of road maintenance.

[0051] Thus, the above is completedFigure 1 The flowchart is shown.

[0052] By Figure 1 As can be seen from the flowchart, the target clustering matching method is used to associate and match the target objects such as road diseases / road assets collected multiple times to automatically realize the abnormal detection of the target objects such as road diseases / road assets, which can greatly improve the efficiency of road maintenance compared with manual review.

[0053] The above step 102 is described as follows:

[0054] Referring to Figure 3 , Figure 3 The step 102 provided by the embodiment of the present application realizes a flowchart. As shown in the flowchart, the flowchart can include the following steps: Figure 3

[0055] Step 301: For each target object in the first target object cluster, determine the reference matching score between the target object and each target object in the second target object cluster in each preset scoring dimension.

[0056] In the embodiment, the preset scoring dimension can be set according to the implementation requirements, such as latitude and longitude, inspection direction, scene (whether elevated), physical size, lane in the image, target category, etc.

[0057] Optionally, in the embodiment, the full score in each dimension can be 100 points, and the matching score is calculated separately for each dimension.

[0058] ​As one embodiment, the aforementioned dimension can be a dimension represented by a numerical parameter. For example, if the dimension is latitude and longitude location information, the latitude and longitude location information deviation between each target object in the first target object group and each target object in the second target object group can be calculated. Based on the latitude and longitude location information deviation between each target object in the first target object group and each target object in the second target object group, a reference matching score can be determined between each target object in the first target object group and each target object in the second target object group. For example, the smaller the latitude and longitude location information deviation, the higher the reference matching score; if the latitude and longitude location information deviation exceeds a certain threshold, the reference matching score is directly assigned a specified score, such as 0. As another example, if the dimension is physical size, the physical size deviation between each target object in the first target object group and each target object in the second target object group can be calculated. Based on the physical size deviation between each target object in the first target object group and each target object in the second target object group, a reference matching score can be determined between each target object in the first target object group and each target object in the second target object group. For example, the smaller the physical size deviation, the higher the reference matching score. If the physical size deviation exceeds a certain threshold, the reference matching score is directly assigned a specified score, such as 0. Pixel size is similar and will not be elaborated further.

[0059] As an example, the above dimension can also be a lane dimension. For each target object in the first target object group, the reference matching score between the target object and each target object in the second target object group is determined based on the lane to which the target object belongs and the lane to which each target object in the second target object group belongs. Among them, when the lanes are the same or adjacent, the higher the reference matching score, and when the lanes are more than a set requirement, such as being separated by one lane, the reference matching score is a specified score, such as 0.

[0060] As an example, the scoring dimension mentioned above is a category dimension. For each target object in the first target object group, a reference matching score is determined between the target object and each target object in the second target object group under this dimension, based on the object category to which the target object belongs and the object category to which each target object in the second target object group belongs; wherein, the reference matching score is higher for the same category or related categories. For example, the reference matching score is higher for the same category and the categories that are specified as allowed to be associated, while in other cases, the reference matching score can be directly defaulted to a specified score, such as 0.

[0061] As an embodiment, the scoring dimension described above can also be a patrol direction dimension, which can determine the reference matching score of each target object in the first target object group and each target object in the second target object group in the dimension according to the patrol direction of the current patrol corresponding to the first target object group and the patrol direction of the last historical patrol corresponding to the second target object group; wherein the higher the reference matching score, the more the same patrol direction. Optionally, in the present embodiment, if the patrol directions are different, the road key data of the collected target objects can have a large deviation, and in the specific implementation, the patrol direction dimension has a veto right, for example, if the patrol directions are different, the reference matching score is directly assigned a specified score such as 0.

[0062] As an embodiment, the scoring dimension is a scene dimension. For each target object in the first target object group, the reference matching score of the target object and each target object in the second target object group in the dimension is determined according to the road scene to which the target object belongs and the road scene to which each target object in the second target object group belongs; wherein the higher the reference matching score, the more the same road scene, and the road scene includes elevated road and non-elevated road. Corresponding to the scene, the latitude and longitude data on the upper and lower surfaces of the elevated road is the same, but it is indeed a different scene and cannot be matched and associated, so the scene category has a veto right, for example, if the scene categories are different, the reference matching score is directly assigned a specified score such as 0.

[0063] The above examples describe how to determine the reference matching score between each target object in the first target object group and each target object in the second target object group in each predetermined scoring dimension.

[0064] Step 302, determining the target matching score between the target object and each target object in the second target object group according to the reference matching score of the target object and each target object in the second target object group in each scoring dimension, and the weight corresponding to each scoring dimension.

[0065] In the present embodiment, the weight corresponding to each scoring dimension can be adjusted according to the calculation accuracy of the target latitude and longitude, target physical size and other data, can be adjusted according to the detection rate and accuracy of target detection, and can also be allocated by different weight allocation schemes for different target categories, and the present embodiment is not specifically limited. Taking the scoring dimensions as the above dimensions as an example, the weight corresponding to each scoring dimension can be: latitude and longitude weight 50%, physical size 20%, position in lane 10%, pixel size 10%, and target category 10%.

[0066] Optionally, for each target object in the first target object group, first, for each dimension, the reference matching score of the target object and a target object in the second target object group in the dimension is multiplied by the weight of the dimension to obtain an operation result; the operation results of the target object and the target object in the second target object group in each dimension are added, and the final result can be the target matching score between the target object and the target object in the second target object group.

[0067] Step 303, according to the target matching scores between the target object and each target object in the second target object group, a matching object having a matching relationship with the target object is determined from the second target object group.

[0068] Optionally, in this step 303, the correspondence relationship between the target objects in the first target object group and the second object group can be determined based on the target matching scores between each target object in the first target object group and each target object in the second target object group. Wherein, when the target matching score between the target object in the first target object group and the target object in the second target object group is greater than a set score threshold, it indicates that the target object in the first target object group and the target object in the second target object group have a corresponding relationship. Under this premise, the correspondence relationship between the target objects in the first target object group and the second object group can exist in one-to-one, one-to-many, many-to-one, many-to-many and no corresponding target, such as Figure 4 the correspondence relationship between the target objects in the first target object group and the second object group in the above embodiment.

[0069] After that, the matching object having a matching relationship for each target object in the first target object group can be determined from the second target object group according to the correspondence relationship. For example, the target objects in the first target object group are traversed according to the inspection order, and the following steps are executed for each target object traversed as a current object: determining a candidate object corresponding to the current object in the second target object group according to the correspondence relationship; determining the matching object of the current object from all candidate objects corresponding to the current object.

[0070] Here, determining the matching object of the current object from all candidate objects corresponding to the current object can include: checking whether there is an unmarked reference object in all candidate objects corresponding to the current object,

[0071] If yes, if the number of reference objects is 1, when the reference object only corresponds to the current object, the reference object is determined as the matching object of the target object; when the reference object also corresponds to other target objects in the first target object group, if the target matching score between the current object and the reference object is greater than the target matching score between other target objects and the reference object, the reference object is determined as the matching object of the current object; if the number of reference objects is greater than 1, according to the target matching scores between the current object and each reference object, the reference object with the highest target matching score is selected as the matching object of the current object;

[0072] If no, the first-in-first-out fish queue rule is used to determine the matching object of the current object, wherein the matching object of the current object is the same as the matching object of at least one target object adjacent to the current object in the first target object group, and the at least one target object adjacent to the current object refers to the target object adjacent to the current object in the inspection order. Here, after the above marking, most target objects in the first target group have found the corresponding matching objects, and there may be a few target objects that are marked in all candidate objects. In this case, the above first-in-first-out fish queue rule can be used.

[0073] Optionally, in the embodiment, after the matching object of the current object is determined, the matching object is marked.

[0074] At this point, the method shown in the flowchart is completed. Figure 3 The flowchart is shown.

[0075] The flowchart shown in Figure 3 The flowchart shown in

[0076] The above describes the method provided by the embodiment. The device provided by the application is described below:

[0077] Referring to Figure 5 , Figure 5 The device structure diagram provided by the embodiment of the application. The device is applied to the electronic device described above, and includes:

[0078] A determination unit is configured to determine a first target object group and a second target object group to be matched;

[0079] A matching unit is configured to determine, for each target object in the first target object group, a matching object having a matching relationship with the target object from the second target object group;

[0080] The detection unit is configured to perform abnormality detection on the target objects according to road key data corresponding to each target object in the first target object group and road key data corresponding to a matching object having a matching relationship with the target object in the second target object group.

[0081] Optionally, the first target object group and the second target object group to be matched by huddling are determined as follows:

[0082] The first target object group is obtained by analyzing road images collected in the current inspection, and the first target object group includes target objects satisfying a preset grouping condition.

[0083] The second target object group satisfying a huddling matching condition with the first target object group is determined from each historical object group corresponding to the last historical inspection, and each historical object group is obtained by analyzing road images collected in the last historical inspection, and target objects in each historical object group satisfy a preset grouping condition.

[0084] Optionally, the target objects in the first target object group or the second target object group satisfy a preset grouping condition as follows: the target objects in the same group are objects under the same road configuration; the road configuration is a road post number or a mileage post number; and / or, the position deviation of the first target object and other target objects in the same group satisfies a preset distance; the first target object in any group is an object that is first inspected in the group; and / or, the target objects in the same group are objects whose positions in the same road section satisfy a preset proximity condition.

[0085] Optionally, the matching object having a matching relationship with each target object in the first target object group is determined from the second target object group as follows: for each target object in the first target object group, a reference matching score between the target object and each target object in the second target object group is determined in each preset scoring dimension; a target matching score between the target object and each target object in the second target object group is determined according to the reference matching scores of the target object and each target object in the second target object group in each scoring dimension and the weight corresponding to each scoring dimension; and the matching object having a matching relationship with the target object is determined from the second target object group according to the target matching scores between the target object and each target object in the second target object group.

[0086] Optionally, the reference matching score between each target object in the first target object group and each target object in the second target object group in each preset scoring dimension is determined as follows:

[0087] If the scoring dimension is a numerical parameter dimension, for each target object in the first target object group, a deviation between the numerical parameter of the target object in the dimension and the numerical parameter of each target object in the second target object group in the dimension is calculated, and a reference matching score of the target object and each target object in the second target object group in the dimension is determined according to the deviation; wherein the smaller the deviation, the higher the reference matching score, and if the deviation exceeds a set deviation threshold, the reference matching score is a specified score; the numerical parameter is at least one of longitude and latitude position information, physical size, and pixel size; and / or,

[0088] If the scoring dimension is a lane dimension, for each target object in the first target object group, a reference matching score of the target object and each target object in the second target object group in the dimension is determined according to the lane to which the target object belongs and the lane to which each target object in the second target object group belongs; wherein the same or adjacent lanes, the higher the reference matching score, and if the distance between the lanes exceeds a set requirement, the reference matching score is a specified score; and / or,

[0089] If the scoring dimension is a category dimension, for each target object in the first target object group, a reference matching score of the target object and each target object in the second target object group in the dimension is determined according to the object category to which the target object belongs and the object category to which each target object in the second target object group belongs; wherein the same category or associated category, the higher the reference matching score; and / or,

[0090] If the scoring dimension is an inspection direction dimension, a reference matching score of each target object in the first target object group and each target object in the second target object group in the dimension is determined according to the inspection direction of the current inspection corresponding to the first target object group and the inspection direction of the above-mentioned last historical inspection corresponding to the second target object group; wherein the same inspection direction, the higher the reference matching score; and / or,

[0091] If the scoring dimension is a scene dimension, for each target object in the first target object group, a reference matching score of the target object and each target object in the second target object group in the dimension is determined according to the road scene to which the target object belongs and the road scene to which each target object in the second target object group belongs; wherein the same road scene, the higher the reference matching score, and the road scene includes elevated road and non-elevated road.

[0092] Optionally, the determining the matching object having the matching relationship with the target object from the second target object group according to the target matching score between the target object and each target object in the second target object group comprises: determining the correspondence between the target objects in the first target object group and the second target object group; wherein, when the target matching score between a target object in the first target object group and a target object in the second target object group is greater than a set score threshold, it indicates that the two target object groups have the correspondence; and determining the matching object having the matching relationship with each target object in the first target object group from the second target object group according to the correspondence.

[0093] Optionally, the determining the matching object having the matching relationship with the target object from the second target object group according to the correspondence comprises: traversing the target objects in the first target object group according to the inspection sequence, and performing the following steps for each target object as a current object: determining the candidate objects corresponding to the current object in the second target object group according to the correspondence; and determining the matching object of the current object from all the candidate objects corresponding to the current object.

[0094] Optionally, the determining the matching object of the current object from all the candidate objects corresponding to the current object comprises: checking whether there is an unmarked reference object in all the candidate objects corresponding to the current object,

[0095] If yes, when the number of the reference objects is 1, if the reference object only corresponds to the current object, determining the reference object as the matching object of the target object; if the reference object also corresponds to other target objects in the first target object group, if the target matching score between the current object and the reference object is greater than the target matching scores between the other target objects and the reference object, determining the reference object as the matching object of the current object; if the number of the reference objects is greater than 1, selecting the reference object with the highest target matching score as the matching object of the current object according to the target matching scores between the current object and each reference object.

[0096] If no, using the first-in first-out rule to determine the matching object of the current object, wherein the matching object of the current object is the same as the matching object of at least one target object adjacent to the current object in the first target object group, and the at least one target object adjacent to the current object refers to the target object adjacent to the current object in the inspection sequence.

[0097] Optionally, the detection unit further marks the matching object after determining the matching object of the current object.

[0098] Optionally, based on the road key data corresponding to each target object in the first target object group and the road key data corresponding to the matching objects in the second target object group that have a matching relationship with the target object, the anomaly detection of the target object includes: for each target object in the first target object group, if there is a matching object matching the target object in the recorded global target sequence, then compare the road key data corresponding to the target object with the road key data corresponding to the matching object in the global target sequence. If the two do not meet the preset approximation conditions, then associate the road key data corresponding to the target object with the global target sequence, and detect whether the anomaly repair conditions are met based on the road key data corresponding to the target object and the road key data corresponding to the matching object. If so, trigger an anomaly alarm; if the two meet the preset approximation conditions, then set a deduplication mark for the road key data corresponding to the target object.

[0099] If no matching object exists for the target object, or if no matching object exists in the recorded global target sequence, then the road key data corresponding to the target object is created in the global target sequence.

[0100] Optionally, the detection unit further tracks the target object from each subsequent process according to the global target sequence, so as to automatically dispatch a work order to the target object for repair based on the tracking results.

[0101] This concludes the process. Figure 5 Structural description of the device shown.

[0102] Correspondingly, this application also provides Figure 5 The hardware structure of the device shown. See also Figure 6 The hardware structure may include: a processor and a machine-readable storage medium, the machine-readable storage medium storing machine-executable instructions that can be executed by the processor; the processor is used to execute the machine-executable instructions to implement the method disclosed in the above example of this application.

[0103] Based on the same application concept as the above method, this application embodiment also provides a machine-readable storage medium storing a plurality of computer instructions, which, when executed by a processor, can implement the method disclosed in the above examples of this application.

[0104] Exemplarily, the machine-readable storage medium described above can be any electronic, magnetic, optical, or other physical storage device that contains or stores information such as executable instructions, data, etc. For example, the machine-readable storage medium can be a RAM (Random Access Memory), a volatile memory, a non-volatile memory, a flash memory, a storage drive (such as a hard drive), a solid-state drive, any type of storage disk (such as a compact disk, a DVD, etc.), or similar storage media, or a combination thereof.

[0105] The systems, apparatuses, modules or units illustrated by the above embodiments can be specifically implemented by a computer processor or entity, or by a product with certain functions. A typical implementation device is a computer, and the specific form of the computer can be a personal computer, a laptop computer, a cellular phone, a camera phone, a smart phone, a personal digital assistant, a media player, a navigation device, an e-mail device, a game console, a tablet computer, a wearable device, or a combination of any of these devices.

[0106] For the convenience of description, the above apparatuses are described in various units by function respectively when described. Of course, the functions of the units can be implemented in one or more software and / or hardware when implementing the present application.

[0107] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the present application can take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the embodiments of the present application can take the form of a computer program product implemented on one or more computer-usable storage media (including, but not limited to, a disk storage, a CD-ROM, an optical storage, etc.) containing computer-usable program code.

[0108] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each 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 processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing apparatus to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing apparatus produce the functions described in the flowcharts and / or block diagrams for implementing the flows and / or blocks in the flowcharts and / or block diagrams. Figure 1 The functions specified in a flow or multiple flows and / or blocks Figure 1 The apparatus that implements the functions specified in a flow or multiple flows and / or blocks.

[0109] Moreover, these computer program instructions can also be stored in a computer readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer readable memory produce an article of manufacture including instructions which implement the flow Figure 1 The flow or flows and / or blocks Figure 1 The flow or flows and / or blocks

[0110] The computer program instructions can also be loaded into a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the flow Figure 1 The flow or flows and / or blocks Figure 1 The flow or flows and / or blocks

[0111] The embodiments of the present application described above are merely intended to illustrate the principles of the present application, and should not be used to limit the scope of the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present application should be included in the scope of protection of the present application.

Claims

1. A target detection method, characterized in that, This method is applied to electronic devices, including: A first target object group and a second target object group to be matched are determined. The target objects in the first target object group are obtained by analyzing road images collected in the current inspection. The target objects in the second target object group are obtained by analyzing road images collected in the most recent historical inspection. The current inspection and the most recent historical inspection are conducted on at least one road segment. The target objects include at least road defect targets. The second target object group is a historical object group that meets the matching condition with the first target object group among the historical object groups corresponding to the most recent historical inspection. The matching condition is that the target objects it contains are all objects under the same road configuration. For each target object in the first target object group, identify the matching object with a matching relationship from the second target object group; Based on the road key data corresponding to each target object in the first target object group and the road key data corresponding to the matching objects in the second target object group that have a matching relationship with the target object, anomaly detection of the target object is performed; the anomaly detection includes at least whether the target object meets the anomaly repair conditions.

2. The method according to claim 1, characterized in that, The determination of the first target group and the second target group to be matched includes: Based on the analysis of the road images collected during the current inspection, the first target object cluster is obtained; the first target object cluster contains target objects that meet the preset clustering conditions. From the historical object clusters corresponding to the most recent historical inspection, a second target object cluster that meets the clustering matching condition with the first target object cluster is determined; each historical object cluster is obtained by analyzing the road images collected in the most recent historical inspection, and the target objects in each historical object cluster meet the preset clustering conditions.

3. The method according to claim 1 or 2, characterized in that, The target objects in the first target object group or the second target object group all satisfy the following preset grouping conditions: The target objects within the same group are those under the same road configuration; the road configuration is a road marker or mileage marker; and / or, The positional deviation between the first target object in the same cluster and other target objects meets a preset distance; the first target object in any cluster is the object first inspected in that cluster; and / or, The target objects in the same group are those whose locations meet the preset similarity conditions within the same road segment.

4. The method according to claim 1, characterized in that, The step of determining matching objects from the second target object group that have a matching relationship with each target object in the first target object group includes: For each target object in the first target object group, determine the reference matching score between that target object and each target object in the second target object group under each preset scoring dimension; Based on the reference matching scores of the target object and each target object in the second target object group under each scoring dimension, and the weights corresponding to each scoring dimension, the target matching score between the target object and each target object in the second target object group is determined. Based on the target matching scores between the target object and each target object in the second target object group, the matching objects that have a matching relationship with the target object are determined from the second target object group.

5. The method according to claim 4, characterized in that, The step of determining the reference matching score between each target object in the first target object group and each target object in the second target object group under each preset scoring dimension includes: If the scoring dimension is a numerical parameter dimension, then for each target object in the first target object group, the deviation between the numerical parameter of that target object in that dimension and the numerical parameter of each target object in the second target object group in that dimension is calculated. Based on the deviation, a reference matching score is determined between that target object and each target object in the second target object group in that dimension; wherein, the smaller the deviation, the higher the reference matching score; if the deviation exceeds a set deviation threshold, the reference matching score is a specified score; the numerical parameter is at least one of latitude and longitude location information, physical size, and pixel size; and / or, If the scoring dimension is the lane dimension, then for each target object in the first target object group, based on the lane to which that target object belongs and the lane to which each target object in the second target object group belongs, a reference matching score is determined between that target object and each target object in the second target object group in that dimension; wherein, if the lanes are the same or adjacent, the reference matching score is higher; if the lane distance exceeds the set requirement, the reference matching score is a specified score; and / or, If the scoring dimension is a category dimension, then for each target object in the first target object group, a reference matching score is determined between that target object and each target object in the second target object group based on the object category to which that target object belongs and the object category to which each target object in the second target object group belongs; wherein, the reference matching score is higher for the same category or related categories; and / or, If the scoring dimension is the inspection direction dimension, then based on the inspection direction of the current inspection corresponding to the first target object group and the inspection direction of the most recent historical inspection corresponding to the second target object group, a reference matching score is determined for each target object in the first target object group and each target object in the second target object group under this dimension; wherein, the higher the reference matching score, the higher the inspection direction; and / or, If the scoring dimension is the scene dimension, then for each target object in the first target object group, the reference matching score between the target object and each target object in the second target object group in this dimension is determined based on the road scene to which the target object belongs and the road scene to which each target object in the second target object group belongs; where the road scene is the same, the higher the reference matching score, and the road scene includes: elevated road and non-elevated road.

6. The method according to claim 4, characterized in that, The step of determining the matching objects that have a matching relationship with the target object from the second target object group based on the target matching score between the target object and each target object in the second target object group includes: Determine the correspondence between target objects in the first target object group and the second target object group; wherein, when the target matching score between the target objects in the first target object group and the target objects in the second target object group is greater than a set score threshold, it indicates that there is a correspondence between the target objects in the first target object group and the target objects in the second target object group; Based on the aforementioned correspondence, matching objects with matching relationships are determined from the second target object group for each target object in the first target object group.

7. The method according to claim 6, characterized in that, The step of determining matching objects with matching relationships from the second target object group for each target object in the first target object group based on the correspondence includes: The target objects in the first target object group are traversed according to the inspection order. For each target object traversed, the following steps are performed as the current object: Based on the correspondence, a candidate object corresponding to the current object is determined in the second target object group; and the matching object of the current object is determined from all the candidate objects corresponding to the current object.

8. The method according to claim 7, characterized in that, The step of determining the matching object of the current object from all candidate objects corresponding to the current object includes: Check if there are any unmarked reference objects among all candidate objects corresponding to the current object. If so, if the number of reference objects is 1, when the reference object corresponds only to the current object, the reference object is determined to be the matching object of the target object; when the reference object also corresponds to other target objects in the first target object group, if the target matching score between the current object and the reference object is greater than the target matching score between other target objects and the reference object, the reference object is determined to be the matching object of the current object; if the number of reference objects is greater than 1, based on the target matching scores between the current object and each reference object, the reference object with the highest target matching score is selected as the matching object of the current object. If not, the first-in-first-out (FIFO) queuing rule is used to determine the matching object of the current object. The matching object of the current object is the same as the matching object of at least one target object adjacent to the current object in the first target object group. At least one target object adjacent to the current object refers to the target object that is adjacent to the inspection sequence of the current object. The method further includes: after determining the matching object of the current object, marking the matching object.

9. The method according to claim 1, characterized in that, Based on the road key data corresponding to each target object in the first target object group, and the road key data corresponding to the matching objects in the second target object group that have a matching relationship with that target object, the anomaly detection of the target objects includes: For each target object in the first target object group, if there is a matching object in the recorded global target sequence, the road key data corresponding to the target object is compared with the road key data corresponding to the matching object in the global target sequence. If the two do not meet the preset approximation conditions, the road key data corresponding to the target object is associated with the global target sequence. Based on the road key data corresponding to the target object and the road key data corresponding to the matching object, it is checked whether the anomaly repair conditions are met. If so, an anomaly alarm is triggered. If the two meet the preset approximation conditions, a deduplication mark is set for the road key data corresponding to the target object. If no matching object exists for the target object, or if no matching object exists in the recorded global target sequence, then the road key data corresponding to the target object is created in the global target sequence.

10. The method according to claim 9, characterized in that, The method further includes: Based on the global target sequence, the target object is tracked from each subsequent process after it occurs, and an automatic work order is dispatched to the target object based on the tracking results for repair.

11. A target detection device, characterized in that, This device is used in electronic devices, including: A determining unit is used to determine a first target object group and a second target object group to be matched; the target objects in the first target object group are obtained by analyzing road images collected in the current inspection; the target objects in the second target object group are obtained by analyzing road images collected in the most recent historical inspection; the current inspection and the most recent historical inspection are conducted on at least one road segment; the target objects include at least road defect targets; the second target object group is a historical object group that meets the grouping matching condition with the first target object group among the historical object groups corresponding to the most recent historical inspection, and the grouping matching condition is that the target objects it contains are all objects under the same road configuration; The matching unit is used to determine, for each target object in the first target object group, a matching object that has a matching relationship with the target object from the second target object group; The detection unit is used to perform anomaly detection on target objects based on the road key data corresponding to each target object in the first target object group and the road key data corresponding to the matching objects in the second target object group that have a matching relationship with the target object; the anomaly detection includes at least whether the target object meets the anomaly repair conditions.

12. The apparatus according to claim 11, characterized in that, The determination of the first target group and the second target group to be matched includes: Based on the analysis of the road images collected during the current inspection, the first target object cluster is obtained; the first target object cluster contains target objects that meet the preset clustering conditions. From the historical object clusters corresponding to the most recent historical inspection, a second target object cluster that meets the clustering matching condition with the first target object cluster is determined; each historical object cluster is obtained by analyzing the road images collected in the most recent historical inspection, and the target objects in each historical object cluster meet the preset clustering conditions; The target objects in the first target object group or the second target object group all meet the following preset grouping conditions: the target objects in the same group are objects under the same road configuration; the road configuration is a road marker or mileage marker; and / or, the positional deviation between the first target object in the same group and other target objects meets a preset distance; the first target object in any group is the object that was inspected first in that group; and / or, the target objects in the same group are objects whose positions meet the preset proximity conditions within the same road segment; The step of determining matching objects from the second target object group for each target object in the first target object group includes: for each target object in the first target object group, determining a reference matching score between the target object and each target object in the second target object group under each preset scoring dimension; determining a target matching score between the target object and each target object in the second target object group based on the reference matching scores of the target object and each target object in the second target object group under each scoring dimension, and the weights corresponding to each scoring dimension; and determining matching objects from the second target object group based on the target matching scores of the target object and each target object in the second target object group. The step of determining the reference matching score between each target object in the first target object group and each target object in the second target object group under each preset scoring dimension includes: If the scoring dimension is a numerical parameter dimension, then for each target object in the first target object group, the deviation between the numerical parameter of that target object in that dimension and the numerical parameter of each target object in the second target object group in that dimension is calculated. Based on the deviation, a reference matching score is determined between that target object and each target object in the second target object group in that dimension; wherein, the smaller the deviation, the higher the reference matching score; if the deviation exceeds a set deviation threshold, the reference matching score is a specified score; the numerical parameter is at least one of latitude and longitude location information, physical size, and pixel size; and / or, If the scoring dimension is the lane dimension, then for each target object in the first target object group, based on the lane to which that target object belongs and the lane to which each target object in the second target object group belongs, a reference matching score is determined between that target object and each target object in the second target object group in that dimension; wherein, if the lanes are the same or adjacent, the reference matching score is higher; if the lane distance exceeds the set requirement, the reference matching score is a specified score; and / or, If the scoring dimension is a category dimension, then for each target object in the first target object group, a reference matching score is determined between that target object and each target object in the second target object group based on the object category to which that target object belongs and the object category to which each target object in the second target object group belongs; wherein, the reference matching score is higher for the same category or related categories; and / or, If the scoring dimension is the inspection direction dimension, then based on the inspection direction of the current inspection corresponding to the first target object group and the inspection direction of the most recent historical inspection corresponding to the second target object group, a reference matching score is determined for each target object in the first target object group and each target object in the second target object group under this dimension; wherein, the higher the reference matching score, the higher the inspection direction; and / or, If the scoring dimension is the scene dimension, then for each target object in the first target object group, the reference matching score between the target object and each target object in the second target object group in this dimension is determined based on the road scene to which the target object belongs and the road scene to which each target object in the second target object group belongs; where the road scene is the same, the higher the reference matching score, and the road scene includes: elevated road and non-elevated road. The step of determining matching objects with a matching relationship with the target object from the second target object group based on the target matching score between the target object and each target object in the second target object group includes: determining the correspondence between the target objects in the first target object group and the second target object group; wherein, when the target matching score between a target object in the first target object group and a target object in the second target object group is greater than a set score threshold, it indicates that there is a correspondence between the two target object groups; and determining matching objects with a matching relationship from the second target object group for each target object in the first target object group based on the correspondence. The step of determining matching objects with matching relationships from the second target object group for each target object in the first target object group based on the correspondence includes: traversing the target objects in the first target object group according to the inspection order, and performing the following steps for each traversed target object as the current object: determining candidate objects corresponding to the current object in the second target object group based on the correspondence; determining matching objects for the current object from all candidate objects corresponding to the current object; The step of determining the matching object of the current object from all candidate objects corresponding to the current object includes: Check if there are any unmarked reference objects among all candidate objects corresponding to the current object. If so, if the number of reference objects is 1, when the reference object corresponds only to the current object, the reference object is determined to be the matching object of the target object; when the reference object also corresponds to other target objects in the first target object group, if the target matching score between the current object and the reference object is greater than the target matching score between other target objects and the reference object, the reference object is determined to be the matching object of the current object; if the number of reference objects is greater than 1, based on the target matching scores between the current object and each reference object, the reference object with the highest target matching score is selected as the matching object of the current object. If not, the first-in-first-out (FIFO) queuing rule is used to determine the matching object of the current object. The matching object of the current object is the same as the matching object of at least one target object adjacent to the current object in the first target object group. At least one target object adjacent to the current object refers to the target object that is adjacent to the inspection sequence of the current object. After determining the matching object of the current object, the detection unit further marks the matching object. Based on the road key data corresponding to each target object in the first target object group and the road key data corresponding to the matching objects in the second target object group that have a matching relationship with the target object, the anomaly detection of the target object includes: for each target object in the first target object group, if there is a matching object matching the target object in the recorded global target sequence, then compare the road key data corresponding to the target object with the road key data corresponding to the matching object in the global target sequence. If the two do not meet the preset approximation conditions, then associate the road key data corresponding to the target object with the global target sequence, and check whether the anomaly repair conditions are met based on the road key data corresponding to the target object and the road key data corresponding to the matching object. If so, trigger an anomaly alarm; if the two meet the preset approximation conditions, then set a deduplication mark for the road key data corresponding to the target object. If there is no matching object for the target object, or if there is no matching object for the target object in the recorded global target sequence, then create the road key data corresponding to the target object in the global target sequence. The detection unit further tracks the target object from each subsequent process based on the global target sequence, and automatically dispatches work orders to the target object for repair based on the tracking results.

13. An electronic device, characterized in that, The electronic device includes: a processor and a machine-readable storage medium; The machine-readable storage medium stores machine-executable instructions that can be executed by the processor; The processor is configured to execute machine-executable instructions to implement the method steps of any one of claims 1-10.

Images