Contour extraction method, search library establishment method, contour matching method, and system thereof

Abstract

The present application relates to the technical field of image processing, and discloses a contour extraction method, a retrieval library establishment method, a contour matching method and a system thereof. The contour extraction method comprises: for each preset contour feature in a plurality of preset contour features, determining a contour feature quantity of the each preset contour feature within each preset step length on the contour; and for each preset contour feature in the plurality of preset contour features, recording a corresponding relationship between the preset step length and the contour feature quantity. The present application can realize high recall of similar contours through high-precision house type contour matching.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of image processing, in particular to a contour extraction method, a search library establishment method, a contour matching method and a system thereof. BACKGROUND

[0002] The contour of a house / room is a very key feature in a house type. For a house / room, the peripheral contour generally belongs to an unchangeable attribute. If the house types with similar contours of the same type can be aggregated through matching, the house types of the same type can share many valuable contents including decoration schemes and reconstruction designs. However, the current contour matching technologies cannot achieve high-precision matching and recommendation, which is a very painful repetitive work for designers and users.

[0003] At present, the contour features are generally judged based on the overall area or contour missing angle. However, this method is difficult to accurately describe the contour of a house type, and when a large number of house type contours are matched, the above method is also insufficient, and the calculation complexity is as high as O(N 2 ). SUMMARY

[0004] The present application aims to provide a contour extraction method, a search library establishment method, a contour matching method and a system thereof, which can realize high recall of similar contours through high-precision house type contour matching, so as to effectively share valuable contents such as similar decoration schemes.

[0005] In order to achieve the above-mentioned purpose, the present application provides a contour extraction method, which comprises: determining, for each preset contour feature in a plurality of preset contour features, a contour feature quantity of the each preset contour feature within each preset step length on the contour; and recording, for each preset contour feature in the plurality of preset contour features, a corresponding relationship between the preset step length and the contour feature quantity, so as to extract the contour.

[0006] Preferably, the plurality of preset contour features comprises at least two of a contour area feature, an accessory feature, a contour line orientation feature and a wall partition type feature.

[0007] Preferably, in the case that the preset contour feature is a contour area feature, the determining of the contour feature quantity of each preset contour feature in each preset step length on the contour comprises: determining a sampling area formed by a first acquisition point corresponding to the preset step length, a second acquisition point and a preset point in the range enclosed by the contour, wherein the acquisition time of the first acquisition point is earlier than the acquisition time of the second acquisition point; in the case that the preset contour feature is an accessory feature, the determining of the contour feature quantity of each preset contour feature in each preset step length on the contour comprises: determining the result of whether the first acquisition point or the second acquisition point falls in the partial contour corresponding to any accessory; in the case that the preset contour feature is a contour line direction feature, the determining of the contour feature quantity of each preset contour feature in each preset step length on the contour comprises: determining the sampling angle of the first acquisition point or the second acquisition point relative to the preset point; or in the case that the preset contour feature is a wall partition type feature, the determining of the contour feature quantity of each preset contour feature in each preset step length on the contour comprises: determining the partition type in which the first acquisition point or the second acquisition point is located.

[0008] Preferably, the preset point is the gravity point of the contour.

[0009] Preferably, the first acquisition point corresponding to the first preset step length in the preset step lengths is the entrance door of the contour.

[0010] Preferably, the recording of the correspondence between the preset step length and the contour feature quantity comprises: for each preset contour feature in the plurality of preset contour features, in the case that the contour feature quantities in a plurality of adjacent preset step lengths are the same, recording the correspondence between the interval in which the plurality of adjacent preset step lengths are located and the contour feature quantity.

[0011] Through the above technical solution, the contour feature quantity of each preset contour feature in each preset step length on the contour is determined creatively, and then for each preset contour feature, the correspondence between the preset step length and the contour feature quantity is recorded. Therefore, the present application can model (i.e. establish the correspondence between the preset step length and each contour feature quantity) at least two of the contour features, such as contour area feature, accessory feature, contour line direction feature and wall partition type, so as to realize high-precision house type contour matching, thereby realizing high-recall of similar contours, and further effectively sharing valuable contents such as similar decoration schemes.

[0012] The second aspect of the present application provides a search library establishing method, which comprises: determining, for each preset profile feature in a plurality of preset profile features, a corresponding relationship between a preset step length of each preset profile in a plurality of preset profiles and a preset profile feature quantity of the each preset profile feature, wherein the plurality of preset profiles correspond to a plurality of different preset profile types, and the corresponding relationship is determined based on the profile extraction method; and storing the corresponding relationship between the preset step length on the each preset profile and the preset profile feature quantity of the each preset profile feature into the search library.

[0013] Through the above technical solution, the present application creatively determines the corresponding relationship between the preset step length of each of a plurality of preset profiles belonging to different preset profile types and the preset profile feature quantity of each preset profile feature, and then stores the corresponding relationship between the preset step length on each preset profile and the preset profile feature quantity of each preset profile feature into the search library. Thus, the present application can realize high-precision house type profile matching based on the established search library, thereby realizing high recall of similar profiles and effectively sharing valuable contents such as similar decoration schemes.

[0014] The third aspect of the present application provides a profile matching method, which comprises: determining, for each of a plurality of to-be-matched profile features of a to-be-matched profile, a corresponding relationship between a preset step length of the to-be-matched profile and a profile feature quantity of the each to-be-matched profile feature, based on the profile extraction method; screening, from a search library stored according to the search library establishing method, a plurality of target profile features of each preset profile in a plurality of preset profiles that match the plurality of to-be-matched profile features; and determining, based on the corresponding relationship between the preset step length in the search library and the profile feature quantity of each target profile feature in the plurality of target profile features and the profile feature quantity of the each to-be-matched profile feature within each preset step length, the top K preset profiles that match the to-be-matched profile by using a weighted algorithm.

[0015] Preferably, the determining the top K preset profiles matching the profile to be matched comprises: determining a similarity between each target profile feature and a corresponding profile feature to be matched based on the correspondence between the preset step length in the search library and the profile feature quantity of each target profile feature and the profile feature quantity of each profile feature to be matched within the respective preset step length; screening K preset profiles for each profile feature to be matched based on the determined similarity, wherein the similarity between each target profile feature and a corresponding profile feature to be matched of the K preset profiles is ranked within the top K; and re-determining the top K preset profiles matching the profile to be matched according to the ranking of the K preset profiles for each profile feature to be matched and the weight of the ranking.

[0016] Preferably, the determining the similarity between each target profile feature and a corresponding profile feature to be matched comprises: for each target profile feature of the plurality of target profile features of each preset profile, calculating an absolute value of a difference between the profile feature quantity of each target profile feature and the profile feature quantity of a corresponding profile feature to be matched within the respective preset step length based on the correspondence between the preset step length in the search library and the profile feature quantity of each target profile feature and the profile feature quantity of each profile feature to be matched within the respective preset step length; summing the absolute value of the difference between the profile feature quantity of each target profile feature and the profile feature quantity of a corresponding profile feature to be matched within the respective preset step length to obtain a sum of the absolute value of the difference between the profile feature quantity of each target profile feature and the profile feature quantity of a corresponding profile feature to be matched within all preset step lengths; and determining the similarity between each target profile feature and a corresponding profile feature to be matched based on the sum of the absolute value of the difference between the profile feature quantity of each target profile feature and the profile feature quantity of a corresponding profile feature to be matched within all preset step lengths.

[0017] Preferably, the re-determining the top K preset profiles matching the profile to be matched comprises: calculating a total weight of each preset profile on the plurality of profile features to be matched according to the ranking of the K preset profiles for each profile feature to be matched and the weight of the ranking to obtain a similarity between each preset profile and the profile to be matched; and re-determining the top K preset profiles matching the profile to be matched based on the similarity between each preset profile and the profile to be matched.

[0018] Preferably, the weight of the ranking of each profile feature to be matched is configured according to a preset requirement.

[0019] Preferably, before the step of determining the top K preset profiles matching the profile to be matched by using the weighted algorithm, the profile matching method further comprises: in the case that a corresponding relationship exists between the interval and the feature profile quantity, decompressing the corresponding relationship between the interval and the feature profile quantity into a corresponding relationship between each of a plurality of adjacent preset steps within the interval and the feature profile quantity.

[0020] By the above technical solution, the application creatively firstly determines the corresponding relationship between the preset step of the profile to be matched and the profile feature quantity of each profile feature to be matched; then, from the retrieval library, filters a plurality of target profile features of each of a plurality of preset profiles matching a plurality of profile features to be matched; finally, based on the corresponding relationship between the preset step in the retrieval library and the profile feature quantity of each of the plurality of target profile features and the profile feature quantity of each profile feature to be matched within each preset step, determines the top K preset profiles matching the profile to be matched by using the weighted algorithm. Thus, the application can realize high-precision house type profile matching based on the established retrieval library, thereby realizing high recall of similar profiles and effectively sharing valuable contents such as similar decoration schemes.

[0021] The fourth aspect of the application provides a profile extraction system, comprising: a feature quantity determination device configured to determine, for each preset profile feature of a plurality of preset profile features, a profile feature quantity of the each preset profile feature within each preset step on the profile; and a recording device configured to record, for each preset profile feature of the plurality of preset profile features, a corresponding relationship between the preset step and the profile feature quantity, so as to extract the profile.

[0022] The specific details and benefits of the profile extraction system provided by the application can be referred to the description of the profile extraction method above, which will not be repeated here.

[0023] The fifth aspect of the application provides a retrieval library establishment system, comprising: the profile extraction system, configured to determine, for each preset profile feature of a plurality of preset profile features, a corresponding relationship between a preset step of each preset profile of a plurality of preset profiles and a preset profile feature quantity of the each preset profile feature, wherein the plurality of preset profiles correspond to a plurality of different preset profile types; and a storage device configured to store the corresponding relationship between the preset step on the each preset profile and the preset profile feature quantity of the each preset profile feature into the retrieval library.

[0024] The specific details and benefits of the retrieval library establishment system provided by the application can be referred to the description of the retrieval library establishment method above, which will not be repeated here.

[0025] The sixth aspect of the present application provides a contour matching system, comprising: the contour extraction system, configured to determine a correspondence between a preset step length of a contour to be matched and a contour feature quantity of each contour feature of a plurality of contour features to be matched; a screening device, configured to screen a plurality of target contour features of each preset contour of a plurality of preset contours matching the plurality of contour features to be matched from a search library stored by the search library establishing system; and a contour determining device, configured to determine the top K preset contours matching the contour to be matched based on the correspondence between the preset step length of the search library and the contour feature quantity of each target contour feature of the plurality of target contour features and the contour feature quantity of each contour feature to be matched within each preset step length by using a weighted algorithm.

[0026] Preferably, the contour determining device comprises: a similarity determining module, configured to determine a similarity between each target contour feature and a corresponding contour feature to be matched based on the correspondence between the preset step length of the search library and the contour feature quantity of each target contour feature of the plurality of target contour features and the contour feature quantity of each contour feature to be matched within each preset step length; a screening module, configured to screen K preset contours for each contour feature to be matched based on the determined similarity, wherein the similarity between each target contour feature of the K preset contours and a corresponding contour feature to be matched is ranked within the top K; and a contour determining module, configured to redetermine the top K preset contours matching the contour to be matched according to the ranking of the K preset contours for each contour feature to be matched and the weight of the ranking of the K preset contours.

[0027] Preferably, the similarity determining module comprises: a first calculating unit configured to calculate, for each target profile feature of the plurality of target profile features of each preset profile, an absolute value of a difference between the profile feature value of each target profile feature and the profile feature value of the corresponding to-be-matched profile feature in each preset step based on the correspondence between the preset step in the search library and the profile feature value of each target profile feature and the profile feature value of the corresponding to-be-matched profile feature in each preset step; a summing unit configured to sum the absolute values of the differences between the profile feature values of each target profile feature and the profile feature values of the corresponding to-be-matched profile features in each preset step to obtain a sum of the absolute values of the differences between the profile feature values of each target profile feature and the profile feature values of the corresponding to-be-matched profile features in all preset steps, for each target profile feature of the plurality of target profile features of each preset profile; and a similarity determining unit configured to determine the similarity between each target profile feature and the corresponding to-be-matched profile feature based on the sum of the absolute values of the differences between the profile feature values of each target profile feature and the profile feature values of the corresponding to-be-matched profile features in all preset steps.

[0028] Preferably, the profile determining module comprises: a second calculating unit configured to calculate a total weight of each preset profile on the plurality of to-be-matched profile features according to the ranking of the K preset profiles for each to-be-matched profile feature and the weight of the ranking for each to-be-matched profile feature to obtain the similarity between each preset profile and the to-be-matched profile; and a profile determining unit configured to re-determine the top K preset profiles matching the to-be-matched profile based on the similarity between each preset profile and the to-be-matched profile.

[0029] Preferably, the weight of the ranking for each to-be-matched profile feature is configured according to a preset requirement.

[0030] The specific details and benefits of the profile matching system provided by the present application can be referred to the description of the profile matching method above, which will not be repeated here.

[0031] The seventh aspect of the present application provides a machine readable storage medium, the machine readable storage medium has stored instructions for causing a machine to execute the profile extraction method, the search library establishing method and / or the profile matching method.

[0032] Other features and advantages of the present application will be described in detail in the following detailed description. BRIEF DESCRIPTION OF DRAWINGS

[0033] The accompanying drawings are included to provide a further understanding of embodiments of the application, and are incorporated in and constitute a part of this specification, illustrate embodiments of the application, and together with the description serve to explain embodiments of the application, but do not limit the present application. In the drawings:

[0034] Figure 1 is a flow chart of the profile extraction method provided by an embodiment of the present application;

[0035] Figure 2 is a flow chart of determining the feature quantity of the plurality of preset profile features in each preset step length provided by an embodiment of the present application;

[0036] Figure 3 is a schematic diagram of performing the sampling process by using the Cartesian coordinate system provided by an embodiment of the present application;

[0037] Figure 4 is a schematic diagram of the house type profile region provided by an embodiment of the present application;

[0038] Figure 5 is a flow chart of the profile matching method provided by an embodiment of the present application;

[0039] Figure 6 is a flow chart of determining the top K preset profiles matched with the to-be-matched profile by using the weighted algorithm provided by an embodiment of the present application;

[0040] Figure 7 is a flow chart of determining the similarity between each target profile feature and the corresponding to-be-matched profile feature provided by an embodiment of the present application;

[0041] Figure 8 is a structure diagram of the profile matching system provided by an embodiment of the present application;

[0042] Figure 9 is a structure diagram of the profile extraction system provided by an embodiment of the present application; and

[0043] Figure 10 is a structure diagram of the profile matching system provided by an embodiment of the present application. DETAILED DESCRIPTION

[0044] The specific embodiments of the present application will be described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are merely intended to illustrate and explain the present application, and are not intended to limit the present application.

[0045] Figure 1is a flowchart of a contour extraction method provided by an embodiment of the present application. The contour extraction method can include: step S101, determining, for each preset contour feature in a plurality of preset contour features, a contour feature quantity of the each preset contour feature within each preset step length on the contour; and step S102, recording, for each preset contour feature in the plurality of preset contour features, a correspondence between the preset step length and the contour feature quantity, to extract the contour. The preset step length refers to a step length between any two adjacent collection points on the contour.

[0046] The plurality of preset contour features can include at least two of a contour area feature, an accessory feature, a contour line orientation feature, and a wall partition type feature.

[0047] In a case where the preset contour feature is the contour area feature, for step S101, the determining of the contour feature quantity of the each preset contour feature within each preset step length on the contour can include determining a sampling area of a preset point in a range enclosed by the contour and a first collection point and a second collection point corresponding to the each preset step length. The collection time of the first collection point is earlier than the collection time of the second collection point.

[0048] In a case where the preset contour feature is the accessory feature, for step S101, the determining of the contour feature quantity of the each preset contour feature within each preset step length on the contour can include determining a result of whether the first collection point or the second collection point falls within a partial contour corresponding to any accessory.

[0049] In a case where the preset contour feature is the contour line orientation feature, for step S101, the determining of the contour feature quantity of the each preset contour feature within each preset step length on the contour can include determining a sampling included angle of the first collection point or the second collection point relative to the preset point.

[0050] In a case where the preset contour feature is the wall partition type feature, for step S101, the determining of the contour feature quantity of the each preset contour feature within each preset step length on the contour can include determining a partition type in which the first collection point or the second collection point is located.

[0051] In order to maintain the rotational invariance of the contour, in a preferred embodiment of the present application, the preset point can be set as a barycentric point of the contour capable of embodying the contour feature. Specifically, the preset point can be a barycentric point of the contour.

[0052] In order to keep the translation and symmetry invariance of the contour, in the preferred embodiment of the present application, the acquisition point (for example, the first acquisition point) can also be arranged at the entrance door of the common house type. Specifically, the first acquisition point corresponding to the first preset step length in the respective preset step length can be at the entrance door of the contour.

[0053] Specifically, the above process of determining the feature quantity of the plurality of preset contour features in each preset step length can be performed by Figure 10 the contour extraction system 10 (which can also be referred to as a contour extractor) shown in the figure.

[0054] As Figure 2 shown, the process performed by the contour extraction system 10 mainly includes steps S201-S206.

[0055] Step S201, selecting the center of gravity point of the house type contour.

[0056] Step S202, setting the sampling step length of the house type.

[0057] For example, the sampling step length can include: step 1, step 2, step 3, …, step i, …, step 360. The specific step length can be dynamically adjusted according to the search accuracy / speed, and 360 step lengths can be used by default; for high response scenarios, 180 step lengths can be used by default.

[0058] Step S203, setting the sampling starting point of the contour as the entrance door of the contour.

[0059] For example, in the Cartesian coordinate system, the origin of the Cartesian coordinate system is set at the center of gravity point of the contour; the sampling direction is selected as the counterclockwise direction. Setting the sampling starting point of the contour as the entrance door of the contour is equivalent to point A in the Cartesian coordinate system, as Figure 3 shown.

[0060] The advantage of this sampling method is that it has nothing to do with the rotation and translation of the house type. In essence, this is because the entrance door area in the common house type is certain.

[0061] Step S204, calculating each step length on the contour under the angle equidistribution condition.

[0062] Step S205, walking in the counterclockwise direction and calculating the feature quantity of the plurality of preset contour features in each step length.

[0063] Specifically, if the plurality of contour features include: contour area feature, accessory feature, contour line orientation feature, and wall partition type feature, the process of calculating the feature quantity of each preset contour feature in each step length is as follows.

[0064] (1)Contour area feature: calculate the area of the triangle formed by the current collection point, the last collection point and the origin of the coordinate system. This method has high calculation efficiency. In the case of high-precision matching, the area of the more complex polygon formed by the current collection point, the last collection point and the origin of the coordinate system, and the actual intersection area of the current contour can also be calculated.

[0065] (2) Accessory feature: determine whether the current collection point falls in the contour area with accessories. If it does, record 1 (for example, collection point B shown in FIG. 6 falls in the contour area with accessories), otherwise record 0 (for example, collection point C shown in FIG. 6 does not fall in the contour area with accessories). Figure 4 Figure 4 The lower part shows the order of collecting each accessory (for example, doors and windows). Figure 4

[0066] (3) Contour line orientation feature: a ray is emitted from the origin of the coordinate system as the starting point to the current collection point, and the included angle between the wall where the current collection point is located and the ray is calculated. The included angle algorithm uses abs(atan2) to calculate:

[0067]

[0068] where y, x are the longitudinal coordinate and transverse coordinate of the resultant vector of the two vectors of the wall orientation (with the origin O as the reference, the counterclockwise direction is the positive direction) and the ray. Finally, min(180-abs(atan2), abs(atan2)) is taken as the included angle value corresponding to the contour line orientation feature. The advantage of this algorithm is that it can obtain the included angle between the ray and the wall while being compatible with the 90-degree case, ensuring that the same included angle is obtained regardless of whether it is calculated clockwise or counterclockwise. Therefore, if the house type is flipped symmetrically, the entire vector can still be matched by inverting it.

[0069] (4) Wall compartment type feature: the collection point is encoded using compartment type encoding, and finally converted to binary for storage.

[0070] Specifically, different walls may connect two different compartments, and all types of compartments are sorted and encoded using one-hot encoding and addition. For example, if the wall corresponds to compartment types 5 and 4, it can be encoded as "11000" (binary). Since a 11000 string does not need to be saved, it is more efficient in storage. Finally, the binary is converted to 2 4 +2 3 (decimal). This way, it is very fast to read directly into binary for bitwise comparison (byte by byte), and the comparison process is more efficient.

[0071] Step S206, statistics of the feature quantity of a plurality of preset contour features in each step.​​

[0072] After walking around the house type contour for 360 steps, the feature values of the corresponding contour area features in each step can be stored in the form of an area slice histogram; the feature values of the corresponding accessory features in each step can be stored in the form of a one-hot item histogram; the feature values of the corresponding accessory features in each step can be stored in the form of a towards histogram; and the feature values of the corresponding accessory features in each step can be stored in the form of an area type histogram. That is, a 360-dimensional vector is counted. Thus, the embodiment can obtain the final histogram vector of each feature based on the boundary walking idea and using equal-interval sampling; and the histogram vectors of multiple features corresponding to different types of contours can be stored in a retrieval library, and the efficient matching process of the house type contour can be realized by matching the histogram in the future.

[0073] Since there are many repeated structures in the house type, in order to optimize storage, the feature values of each dimension of the sample are compressed. The compression strategy is: if the feature values corresponding to the continuous adjacent steps are the same, the continuous adjacent steps are merged into an interval represented as [(start_index, end_index), value]. Thus, the compression function is to compress the repeated area values

[0074] For step S102, recording the correspondence between the preset step and the contour feature value can include: for each preset contour feature in the plurality of preset contour features, in the case that the contour feature values in the plurality of adjacent preset steps are the same, recording the correspondence between the interval where the plurality of adjacent preset steps are located and the contour feature value, to extract the contour.

[0075] Suppose the contour areas corresponding to steps 1-200 of a house type are all the same value (such as 30), at this time it is not necessary to save 200 30s, only the corresponding step index range needs to be saved, that is, [(1, 200), 30]. That is, the 360-dimensional vector is compressed. Subsequently, when actually comparing, 200 30s will be decompressed (thus becoming an equal-length array), and then compared with the values corresponding to the step indexes of the house type contour to be matched.

[0076] In summary, the present application determines the profile feature quantity of each of a plurality of preset profile features within each preset step length on the profile, and then records the correspondence between the preset step length and the profile feature quantity for each preset profile feature. In this way, the present application can model (i.e. establish the correspondence between the preset step length and each profile feature quantity) at least two of the profile area feature, the accessory feature, the profile line orientation feature, and the wall partition type, thereby achieving high-precision house type profile matching, achieving high recall of similar profiles, and effectively sharing valuable content such as similar decoration schemes.

[0077] An embodiment of the present application also provides a retrieval library establishment method. The method can include: for each preset profile feature in a plurality of preset profile features, determining the correspondence between the preset step length of each preset profile in a plurality of preset profiles and the preset profile feature quantity of each preset profile feature based on the profile extraction method, wherein the plurality of preset profiles correspond to a plurality of different preset profile types; storing the correspondence between the preset step length on each preset profile and the preset profile feature quantity of each preset profile feature in the retrieval library.

[0078] This process is used to save vectors and establish a spatial index of house type vectors. The specific steps can include: (1) stringizing the compressed high-dimensional vectors described above; (2) saving the correspondence between the house type and the high-dimensional vector feature corresponding to each preset profile feature; (3) after all are saved, performing spatial clustering and dividing the index of the high-dimensional vector corresponding to each preset profile feature. For example, the index of the high-dimensional vector corresponding to the profile area feature of house type f1 is set to 11; the index of the high-dimensional vector corresponding to the profile area feature of house type f2 is set to 12; …… the index of the high-dimensional vector corresponding to the profile area feature of house type f360 is set to 1(360). The index of the high-dimensional vector corresponding to the accessory feature of house type f1 is set to 21; the index of the high-dimensional vector corresponding to the accessory feature of house type f2 is set to 22; …… the index of the high-dimensional vector corresponding to the accessory feature of house type f360 is set to 2(360). Other features can be indexed in a similar manner. (4) Complete the index library writing. The spatial index established in this process can facilitate subsequent high-performance queries.

[0079] In summary, this invention creatively determines the correspondence between the preset step size of each of multiple preset contours belonging to different preset contour types and the preset contour feature quantity of each preset contour feature. Then, the correspondence between the preset step size of each preset contour and the preset contour feature quantity of each preset contour feature is stored in a retrieval database. Therefore, this invention can achieve high-precision apartment contour matching based on the established retrieval database, thereby achieving high recall of similar contours and effectively sharing valuable content such as similar decoration schemes.

[0080] Figure 5 This is a flowchart of a contour matching method provided in an embodiment of the present invention. Figure 5 As shown, the contour matching method may include steps S501-S503.

[0081] Step S501: For each of the multiple contour features to be matched in the contour to be matched, based on the contour extraction method, determine the correspondence between the preset step size of the contour to be matched and the contour feature quantity of each contour feature to be matched.

[0082] This step S501 can be performed by... Figure 10 The contour extraction system 10 in the above text executes the process, and the specific execution process can be found in the relevant description of the contour extraction method above. It should be noted that the preset step size in this step S501 should be consistent with the preset step size in the contour extraction method above, that is, the dimension of the feature vector to be matched (or retrieved) (which can be referred to as the retrieval dimension) needs to be consistent with the dimension of each feature vector in the retrieval library (which can be referred to as the vector dimension in the library).

[0083] If consistency cannot be guaranteed, it can be solved by interpolation, which has two possible scenarios.

[0084] First, the dimension of the library vector > the dimension of the retrieval: the dimension of the feature vector to be retrieved needs to be upsampled to the vector in the library at intervals.

[0085] Second, if the dimension of the vector in the library is less than the dimension of the retrieval vector, the dimension of the feature vector to be retrieved needs to be downsampled at equal intervals.

[0086] Step S502: From the retrieval library stored according to the retrieval library establishment method, select multiple target contour features of each preset contour that match the multiple preset contour features to be matched.

[0087] This step S502 can be performed by... Figure 10 The filtering device 20 performs the operation. Specifically, when the contour features to be matched include contour area features, attachment features, contour line orientation features, and wall partition type features, the filtering device 20 filters out the corresponding four features from the search library.

[0088] Before performing the following step S503, the profile matching method can further comprise: in the case that the correspondence relationship between the interval and the feature profile quantity exists, decompressing the correspondence relationship between the interval and the feature profile quantity into a correspondence relationship between each of a plurality of adjacent preset step lengths within the interval and the feature profile quantity.

[0089] Specifically, the decompressed correspondence relationship [(1, 200), 30] involved in the detailed description of step S102 can be decompressed to decompress 200 30s (i.e. [1, 30], [2, 30] … [1, 30], [200, 30], thereby becoming a 300-dimensional array), which together with the array corresponding to step lengths 201-360, forms a 360-dimensional array. The decompressed correspondence relationship can be used for comparison with the values corresponding to each step length index of the house type profile to be matched.

[0090] Step S503, based on the correspondence relationship between the preset step lengths in the search library and the profile feature quantities of each target profile feature in the plurality of target profile features and the profile feature quantity of each to-be-matched profile feature within each preset step length, determining the top K preset profiles matched with the to-be-matched profile by using a weighted algorithm.

[0091] For step S503, as shown in Figure 6 , the determination of the top K preset profiles matched with the to-be-matched profile by using a weighted algorithm can include steps S601-S603, which can be performed by the profile determination device 30 (i.e. the profile feature matcher) in Figure 10 .

[0092] Step S601, based on the correspondence relationship between the preset step lengths in the search library and the profile feature quantities of each target profile feature in the plurality of target profile features and the profile feature quantity of each to-be-matched profile feature within each preset step length, determining the similarity between each target profile feature and the corresponding to-be-matched profile feature.

[0093] For step S601, as shown in Figure 7 , the determination of the similarity between each target profile feature and the corresponding to-be-matched profile feature can include steps S701-S703.

[0094] Step S701, for each target contour feature of the plurality of target contour features of each preset contour, based on the corresponding relationship between each target contour feature and preset step length in the search library and the contour feature quantity of each target contour feature within each preset step length, calculate the absolute value of the difference between the contour feature quantity of each target contour feature and the contour feature quantity of the corresponding to-be-matched contour feature within each preset step length.

[0095] For example, for the contour area feature of the house type f1 in the search library, the corresponding relationship between the preset step length and the feature quantity of the contour area feature is [1, 10], [2, 10], [3, 12], … [360, 13] (this corresponding relationship can also be expressed as (step length 1, step length 2, step length 3 … step length 360):(10, 10, 12 … 13)); and for the to-be-searched house type, the feature quantity of the contour area feature within each preset step length is (10, 10, 11 … 13). Then, the absolute value of the difference between the feature quantity of the contour area feature of the to-be-searched house type and the feature quantity of the contour area feature of the house type f1 in the corresponding search library within each preset step length can be calculated as [1, 0], [2, 0], [3, 1], … [360, 0] (this corresponding relationship can also be expressed as (step length 1, step length 2, step length 3 … step length 360):(0, 0, 1 … 0)). For other house types f2, f3, … in the search library, the corresponding calculation process is similar to the above, and will not be described here.

[0096] Step S702, for each target contour feature of the plurality of target contour features of each preset contour, sum the absolute value of the difference between the contour feature quantity of each target contour feature and the contour feature quantity of the corresponding to-be-matched contour feature within each preset step length, to obtain the sum of the absolute value of the difference between the contour feature quantity of each target contour feature and the contour feature quantity of the corresponding to-be-matched contour feature within all preset step lengths.

[0097] For example, for the contour area feature of the house type f1 in the search library, the absolute values of the differences corresponding to each preset step length calculated by step S701 can be summed to obtain 1 (i.e., the distance of the high-dimensional vector).

[0098] Step S703, based on the sum of the absolute value of the difference between the contour feature quantity of each target contour feature and the contour feature quantity of the corresponding to-be-matched contour feature within all preset step lengths, determine the similarity between each target contour feature and the corresponding to-be-matched contour feature.

[0099] The relationship between the similarity S and the sum Sum of the absolute values of the difference values can be expressed as follows: S = f(Sum), wherein S and Sum have an inverse function relationship, that is, the greater the sum Sum of the absolute values of the difference values, the smaller the similarity S of each target profile feature and the corresponding profile feature to be matched. The specific form of the relationship can be set according to actual conditions.

[0100] If the sum Sum1 and Sum2 of the absolute values of the difference values corresponding to the profile area features of the house types f1 and f2 obtained through steps S701-S702 are 1 and 2 respectively, it can be determined that the similarity S1 of the profile area features of the house type f1 and the profile area features of the house type to be matched is greater than the similarity S2 of the profile area features of the house type f2 and the profile area features of the house type to be matched.

[0101] In step S602, based on the determined similarity, K preset profiles for each profile feature to be matched are screened.

[0102] The similarity of each target profile feature of the K preset profiles and the corresponding profile feature to be matched is ranked in the top K positions. K is usually set according to the business query range, or can be set by specifying the maximum tolerable distance difference (i.e., the absolute value of the difference value).

[0103] Steps S601-S602 can be performed by a vector retrieval engine group (which includes vector retrieval engines 2, 3, 4, 5, and the vector retrieval engine group can specifically include a similarity determination module (not shown) and a screening module (not shown)). Figure 8 Specifically, for the profile area feature, the similarity between each house type and the house type to be retrieved determined through step S601 can be arranged in descending order, and then the top K (i.e., Top-K) house types are screened. For example, for the profile area feature, the top 3 house types are f1, f2, and f3. For other features, the top K (i.e., Top-K) house types can be similarly screened. Specifically, for the accessory feature, the top 3 house types are f3, f4, and f1; for the profile line orientation feature, the top 3 house types are f2, f5, and f1; and for the wall partition type feature, the top 3 house types are f1, f2, and f6, as shown in Figure 8 .

[0104] In step S603, according to the ranking of the K preset profiles for each profile feature to be matched and the weight occupied by the ranking of the K preset profiles, the top K preset profiles ranked in the ranking and matched with the profile to be matched are re-determined.

[0105] For step S603, the re-determination of the top-K preset profiles matching the to-be-matched profile can comprise: calculating a total weight of each preset profile on the plurality of to-be-matched profile features according to the ranking of the K preset profiles for each to-be-matched profile feature and the weight of the ranking for each to-be-matched profile feature, to obtain a similarity between each preset profile and the to-be-matched profile; and re-determining the top-K preset profiles matching the to-be-matched profile based on the similarity between each preset profile and the to-be-matched profile.

[0106] The weight of the ranking of the K preset profiles is configured according to preset requirements. Specifically, according to different preferences, the ranking corresponding to each profile feature can be set to different weights. For example, if a designer pays great attention to the accessory feature of a house type, the corresponding weights can be set as follows: for the profile area feature, the weights of the top 3 house types f1, f2, and f3 are 1, 0.8, and 0.6, respectively; for the accessory feature, the weights of the top 3 house types f3, f4, and f1 are 10, 8, and 6, respectively; for the profile line orientation feature, the weights of the top 3 house types f2, f5, and f1 are 1, 0.8, and 0.6, respectively; and for the wall partition type feature, the weights of the top 3 house types f1, f2, and f6 are 1, 0.8, and 0.6, respectively.

[0107] First, the total weights of the house types f1, f2, f3, f4, f5, and f6 on the four to-be-matched profile features are calculated as 8.6 (1+6+0.6+1), 2.6 (0.8+1+0.8), 10.6 (0.6+10), 8, 0.8, and 0.6, respectively, that is, the similarities between the house types f1, f2, f3, f4, f5, and f6 and the to-be-matched profile (which is positively correlated with the total weight, for example, the total weight can be directly used as the similarity, or an integer multiple of the total weight can be used as the similarity, etc.) can be 8.6, 2.6, 10.6, 8, 0.8, and 0.6, respectively; then, according to the descending order of the similarities, the top-3 house types matching the to-be-matched house type are re-determined as f3, f1, and f4 (the top-3 house type profiles are most similar to the profile of the to-be-retrieved house type).

[0108] Specifically, as Figure 8As shown, first, the contour feature of the real-time retrieved house type is extracted by the contour extraction system 10; second, the corresponding contour feature (i.e. vector feature) is filtered from the retrieval library, and the nearest neighbor Top-K house type is queried from the vector feature in a single dimension by each vector retrieval engine in the vector retrieval engine group (vector retrieval engines 2-5); and finally, the Top-K house type results returned by the vector retrieval engines in various dimensions are secondarily sorted by the contour determination module 6 (which has a weighted calculation function) to output the final Top-K house type. Thus, the embodiment of the present application can solve the problem of low efficiency of two-by-two comparison of contour features, thereby realizing a high-precision and low-delay matching process of massive data.

[0109] To sum up, the present application creatively first determines the corresponding relationship between the preset step length of the contour to be matched and the contour feature quantity of each contour feature to be matched; then, from the retrieval library, filters the multiple target contour features of each of the multiple preset contours matched with the multiple contour features to be matched; and finally, based on the corresponding relationship between the preset step length in the retrieval library and the contour feature quantity of each of the multiple target contour features and the contour feature quantity of each contour feature to be matched in each preset step length, determines the top K preset contours matched with the contour to be matched by using a weighted algorithm. Thus, the present application can realize high-precision house contour matching based on the established retrieval library, thereby realizing high recall of similar contours and effectively sharing valuable contents such as similar decoration schemes.

[0110] Figure 9 is a structural diagram of the contour extraction system provided by an embodiment of the present application. As shown in Figure 9 The contour extraction system 10 can include: a feature quantity determination device 100 for determining, for each preset contour feature in multiple preset contour features, the contour feature quantity of the preset contour feature in each preset step length on the contour; and a recording device 200 for recording, for each preset contour feature in the multiple preset contour features, the corresponding relationship between the preset step length and the contour feature quantity, to extract the contour.

[0111] Preferably, the multiple preset contour features include at least two of the following: contour area feature, accessory feature, contour line orientation feature, and wall partition type feature.

[0112] Preferably, when the preset profile feature is a profile area feature, the feature quantity determining device comprises an area determining module configured to determine a sampling area surrounded by a preset point in a range enclosed by the first acquisition point, the second acquisition point and the profile, wherein the acquisition time of the first acquisition point is earlier than the acquisition time of the second acquisition point; when the preset profile feature is an accessory feature, the feature quantity determining device comprises an accessory determining module configured to determine a result of whether the first acquisition point or the second acquisition point falls in a partial profile corresponding to any accessory; when the preset profile feature is a profile line orientation feature, the feature quantity determining device comprises an angle determining module configured to determine a sampling angle of the first acquisition point or the second acquisition point relative to the preset point; or when the preset profile feature is a wall partition type feature, the feature quantity determining device comprises a partition determining module configured to determine a partition type in which the first acquisition point or the second acquisition point is located.

[0113] Preferably, the preset point is a gravity point of the profile.

[0114] Preferably, the first acquisition point corresponding to a first preset step length in the preset step lengths is a front door of the profile.

[0115] Preferably, the recording device configured to record the corresponding relationship between the preset step lengths and the profile feature quantities comprises, for each preset profile feature in the plurality of preset profile features, recording a corresponding relationship between an interval in which a plurality of adjacent preset step lengths are located and the profile feature quantity in a case where the profile feature quantities in the plurality of adjacent preset step lengths are the same.

[0116] The specific details and benefits of the profile extraction system provided by the present application can be referred to the description of the profile extraction method above, which will not be repeated here.

[0117] The fifth aspect of the present application provides a retrieval library establishment system, which comprises: the profile extraction system configured to determine, for each preset profile feature in a plurality of preset profile features, a corresponding relationship between a preset step length of each preset profile in a plurality of preset profiles and a preset profile feature quantity of the each preset profile feature, wherein the plurality of preset profiles correspond to a plurality of different preset profile types; and a storage device configured to store the corresponding relationship between the preset step length on the each preset profile and the preset profile feature quantity of the each preset profile feature to the retrieval library.

[0118] The specific details and benefits of the retrieval library establishment system provided by the present application can be referred to the description of the retrieval library establishment method above, which will not be repeated here.

[0119] Figure 10 is a structural diagram of a profile matching system provided by an embodiment of the present application. As shown in the figure, the profile matching system can include the profile extraction system 10, the screening device 20, and the profile determination device 30. Figure 10 The profile determination device 30 can also be referred to as a profile feature matcher. The profile determination device 30 is configured to determine the top K preset profiles matching the to-be-matched profile based on the correspondence between the preset step length and the profile feature quantity of each target profile feature in the preset profiles matching the to-be-matched profile and the profile feature quantity of each to-be-matched profile feature in each preset step length, by using a weighted algorithm.

[0120] The specific details and benefits of the profile matching system provided by the present application can be referred to the description of the profile matching method above, and will not be described here again.

[0121] An embodiment of the present application further provides a machine readable storage medium, and the machine readable storage medium stores instructions for causing a machine to execute the profile extraction method, the search library establishment method, and / or the profile matching method.

[0122] The above describes optional embodiments of the embodiments of the present application in detail in combination with the drawings, but the embodiments of the present application are not limited to the specific details in the above embodiments, and within the technical concept range of the embodiments of the present application, the technical solutions of the embodiments of the present application can be variously modified, and these simple modifications all belong to the protection range of the embodiments of the present application.

[0123] In addition, it should be noted that each specific technical feature described in the above specific embodiments can be combined in any appropriate manner without contradiction. In order to avoid unnecessary repetition, the embodiments of the present application will not describe various possible combinations again.

[0124] Those skilled in the art can understand that all or part of the steps of the methods in the above embodiments can be completed by instructing the relevant hardware through a program stored in a storage medium, including a plurality of instructions for enabling a single-chip microcomputer, a chip or a processor to execute all or part of the steps of the methods described in various embodiments of the present application. The foregoing storage medium includes a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and various media capable of storing program codes.

[0125] In addition, various different embodiments of the embodiments of the present application can also be combined arbitrarily, as long as they do not deviate from the idea of the embodiments of the present application, and should be considered as disclosed by the embodiments of the present application.

Claims

1. A method for extracting apartment layout outlines, characterized in that, The method for extracting the apartment layout outline includes: For each of a plurality of preset contour features, determine the contour feature quantity of each preset contour feature within each preset step size on the contour; and For each of the plurality of preset contour features, the correspondence between the preset step size and the contour feature quantity is recorded to extract the apartment layout contour. The plurality of preset contour features include at least two of the following: contour area features, attachment features, contour line orientation features, and wall partition type features. When the preset contour feature is a contour area feature, determining the contour feature quantity of each preset contour feature within each preset step length on the apartment contour includes: determining the sampling area enclosed by the first sampling point and the second sampling point corresponding to each preset step length and the preset points within the range enclosed by the apartment contour, wherein the sampling time of the first sampling point is earlier than the sampling time of the second sampling point. When the preset contour feature is an attachment feature, determining the contour feature quantity of each preset contour feature within each preset step size on the apartment layout contour includes: determining whether the first collection point or the second collection point falls within the part of the apartment layout contour corresponding to any attachment, wherein any attachment is a door or a window; When the preset contour feature is a contour line orientation feature, determining the contour feature quantity of each preset contour feature within each preset step size on the apartment contour includes: determining the sampling angle between the first sampling point or the second sampling point and the preset point; or When the preset contour feature is a wall partition type feature, determining the contour feature quantity of each preset contour feature within each preset step length on the apartment contour includes: determining the partition type where the first collection point or the second collection point is located.

2. The method for extracting apartment layout outlines according to claim 1, characterized in that, The preset point is the center of gravity of the apartment layout.

3. The method for extracting apartment layout outlines according to claim 1, characterized in that, The first collection point corresponding to the first preset step size among the preset step sizes is the entrance door of the apartment layout.

4. The method for extracting apartment layout outlines according to claim 1, characterized in that, The recording of the correspondence between the preset step size and the contour feature quantity includes: For each of the plurality of preset contour features, if the contour feature quantity is the same within a plurality of adjacent preset step sizes, the correspondence between the interval where the plurality of adjacent preset step sizes are located and the feature contour quantity is recorded.

5. A method for establishing a retrieval database, characterized in that, The method includes: For each of a plurality of preset contour features, based on the apartment layout contour extraction method according to any one of claims 1-4, a correspondence is determined between a preset step size of each preset contour and a preset contour feature quantity of each preset contour feature, wherein the plurality of preset contours correspond to a plurality of different preset contour types; and The correspondence between the preset step size on each preset contour and the preset contour feature quantity of each preset contour feature is stored in the retrieval library.

6. A contour matching method, characterized in that, The contour matching method includes: For each of the multiple contour features to be matched in the contour to be matched, based on the apartment contour extraction method according to any one of claims 1-4, the correspondence between the preset step size of the contour to be matched and the contour feature quantity of each contour feature to be matched is determined. From the retrieval library stored in the retrieval library establishment method according to claim 5, select multiple target contour features of each of multiple preset contours that match the multiple unmatched contour features; and Based on the correspondence between the preset step size in the retrieval library and the contour feature quantity of each target contour feature among the plurality of target contour features, and the contour feature quantity of each contour feature to be matched within each preset step size, a weighted algorithm is used to determine the top K preset contours that match the contour to be matched.

7. The contour matching method according to claim 6, characterized in that, The step of using a weighted algorithm to determine the top K preset contours that match the contour to be matched includes: Based on the correspondence between the preset step size in the retrieval library and the contour feature quantity of each target contour feature among the multiple target contour features, and the contour feature quantity of each contour feature to be matched within each preset step size, the similarity between each target contour feature and the corresponding contour feature to be matched is determined. Based on the determined similarity, K preset contours are selected for each contour feature to be matched, wherein the similarity ranking of each target contour feature of the K preset contours with the corresponding contour feature to be matched is among the top K; and Based on the sorting of the K preset contours for each contour feature to be matched and the weight of the sorting of the K preset contours, the top K preset contours that match the contour to be matched are re-determined.

8. The contour matching method according to claim 7, characterized in that, Determining the similarity between each target contour feature and the corresponding contour feature to be matched includes: For each of the plurality of target contour features of each preset contour, based on the correspondence between the preset step size in the retrieval library and the contour feature quantity of each target contour feature, and the contour feature quantity of each contour feature to be matched within each preset step size, the absolute value of the difference between the contour feature quantity of each target contour feature within each preset step size and the contour feature quantity of the corresponding contour feature to be matched is calculated. For each of the plurality of target contour features of each preset contour, the absolute values ​​of the differences between the contour feature quantity of each target contour feature and the contour feature quantity of the corresponding target contour feature within each preset step size are summed to obtain the sum of the absolute values ​​of the differences between the contour feature quantity of each target contour feature and the contour feature quantity of the corresponding target contour feature within all preset step sizes; and The similarity between each target contour feature and the corresponding target contour feature is determined by summing the absolute values ​​of the differences between the contour feature quantity of each target contour feature and the contour feature quantity of the corresponding target contour feature within all preset step sizes.

9. The contour matching method according to claim 7, characterized in that, The process of re-determining the top K preset contours that match the contour to be matched includes: Based on the ranking of the K preset contours for each feature to be matched and the weight assigned to each feature, the total weight of each preset contour on the plurality of features to be matched is calculated to obtain the similarity between each preset contour and the feature to be matched; and Based on the similarity between each preset contour and the contour to be matched, the top K preset contours that match the contour to be matched are re-determined.

10. The contour matching method according to claim 9, characterized in that, The weight assigned to each feature to be matched is configured according to preset requirements.

11. The contour matching method according to claim 6, characterized in that, Before performing the step of determining the top K preset contours that match the contour to be matched using a weighted algorithm, the contour matching method further includes: In the case where there is a correspondence between the interval and the feature contour quantity, the correspondence between the interval and the feature contour quantity is decompressed into a correspondence between each of the multiple adjacent preset step sizes within the interval and the feature contour quantity.

12. A floor plan outline extraction system, characterized in that, The apartment layout extraction system includes: A feature quantity determination device is used to determine, for each preset contour feature among a plurality of preset contour features, the contour feature quantity of each preset contour feature within a preset step size on the apartment layout contour; and A recording device is used to record the correspondence between the preset step size and the contour feature quantity for each of the plurality of preset contour features, in order to extract the apartment layout contour. The plurality of preset contour features include at least two of the following: contour area features, attachment features, contour line orientation features, and wall partition type features. When the preset contour feature is a contour area feature, determining the contour feature quantity of each preset contour feature within each preset step length on the apartment contour includes: determining the sampling area enclosed by the first sampling point and the second sampling point corresponding to each preset step length and the preset points within the range enclosed by the apartment contour, wherein the sampling time of the first sampling point is earlier than the sampling time of the second sampling point. When the preset contour feature is an attachment feature, determining the contour feature quantity of each preset contour feature within each preset step size on the apartment layout contour includes: determining whether the first collection point or the second collection point falls within the part of the apartment layout contour corresponding to any attachment, wherein any attachment is a door or a window; When the preset contour feature is a contour line orientation feature, determining the contour feature quantity of each preset contour feature within each preset step size on the apartment contour includes: determining the sampling angle between the first sampling point or the second sampling point and the preset point; or When the preset contour feature is a wall partition type feature, determining the contour feature quantity of each preset contour feature within each preset step length on the apartment contour includes: determining the partition type where the first collection point or the second collection point is located.

13. A retrieval database establishment system, characterized in that, The system includes: The apartment layout outline extraction system according to claim 12 is used to determine, for each of a plurality of preset outline features, the correspondence between a preset step size of each preset outline and a preset outline feature quantity of each preset outline feature, wherein the plurality of preset outlines correspond to a plurality of different preset outline types; and A storage device is used to store the correspondence between the preset step size on each preset contour and the preset contour feature quantity of each preset contour feature into the retrieval database.

14. A contour matching system, characterized in that, The contour matching system includes: The apartment layout outline extraction system according to claim 12 is used to determine the correspondence between the preset step size of the outline to be matched and the outline feature quantity of each outline feature to be matched for each outline feature among a plurality of outline features to be matched. A filtering device for filtering, from a retrieval library stored by the retrieval library establishment system according to claim 13, a plurality of target contour features of each of a plurality of preset contours that match the plurality of to-be-matched contour features; and A contour determination device is used to determine the top K preset contours that match the contours to be matched, based on the correspondence between the preset step size in the retrieval library and the contour feature quantity of each target contour feature among the plurality of target contour features, and the contour feature quantity of each contour feature to be matched within each preset step size, using a weighted algorithm.

15. The contour matching system according to claim 14, characterized in that, The contour determining device includes: The similarity determination module is used to determine the similarity between each target contour feature and the corresponding target contour feature based on the correspondence between the preset step size in the retrieval library and the contour feature quantity of each target contour feature among the plurality of target contour features, and the contour feature quantity of each target contour feature to be matched within each preset step size. A filtering module is used to filter K preset contours for each contour feature to be matched based on the determined similarity, wherein the similarity ranking of each target contour feature of the K preset contours with the corresponding contour feature to be matched is among the top K; and The contour determination module is used to redetermine the top K preset contours that match the contour to be matched, based on the sorting of the K preset contours for each contour feature to be matched and the weight of the sorting of the K preset contours.

16. The contour matching system according to claim 15, characterized in that, The similarity determination module includes: The first calculation unit is used to calculate, for each of the plurality of target contour features of each preset contour, the absolute value of the difference between the contour feature quantity of each target contour feature and the contour feature quantity of the corresponding target contour feature within each preset step, based on the correspondence between the preset step size in the retrieval library and the contour feature quantity of each target contour feature within each preset step size. The summation unit is used to sum the absolute values ​​of the differences between the contour feature quantity of each target contour feature and the contour feature quantity of the corresponding target contour feature within each preset step size for each target contour feature among the plurality of target contour features of each preset contour, so as to obtain the sum of the absolute values ​​of the differences between the contour feature quantity of each target contour feature and the contour feature quantity of the corresponding target contour feature within all preset step sizes; and The similarity determination unit is used to determine the similarity between each target contour feature and the corresponding target contour feature based on the sum of the absolute values ​​of the differences between the contour feature quantity of each target contour feature within all preset step sizes and the contour feature quantity of the corresponding target contour feature.

17. The contour matching system according to claim 15, characterized in that, The contour determination module includes: The second calculation unit is used to calculate the total weight of each preset contour on the plurality of contour features to be matched, based on the sorting of the K preset contours for each contour feature to be matched and the weight of the sorting for each contour feature to be matched, so as to obtain the similarity between each preset contour and the contour to be matched; and The contour determination unit is used to redetermine the top K preset contours that match the contour to be matched based on the similarity between each preset contour and the contour to be matched.

18. The contour matching system according to claim 17, characterized in that, The weight assigned to each feature to be matched is configured according to preset requirements.

19. A machine-readable storage medium, characterized in that, The machine-readable storage medium stores instructions for causing the machine to perform the floor plan outline extraction method of any one of claims 1-4, the retrieval database establishment method of claim 5, and / or the outline matching method of any one of claims 6-11.

Images