Artificial intelligence garbage classification logistics sorting method

Abstract

The invention relates to an artificial intelligent garbage classification logistics sorting method, which comprises the following steps of: respectively acquiring a multi-attitude feature set of each type of garbage in advance, respectively constructing multi-attitude garbage attribute recognition models of different types of garbage based on the multi-attitude feature sets, and respectively acquiring multi-attitude features of each actual garbage to be classified; the obtained multi-attitude features of the actual garbage are respectively matched with all the constructed multi-attitude garbage attribute recognition models; and then according to whether the multi-attitude feature sets corresponding to all the multi-attitude garbage attribute recognition models of the multi-attitude features of the actual garbage are consistent and matched or not, the sorting operation of the corresponding types of garbage is executed, so that the effect of recognizing the actual garbage types based on the multi-attitude features of the garbage is achieved. And it is ensured that the garbage can be accurately classified, recognized and sorted in different postures.

Description

Technical Field

[0001] The present invention relates to the field of logistics, and in particular to an artificial intelligence garbage classification logistics sorting method. Background Art

[0002] With the development of economic life, various types of garbage have continued to emerge. In order to maintain environmental hygiene and achieve resource recycling, it is necessary to use logistics to sort all garbage and classify it, and then implement different disposal operations for different types of garbage.

[0003] However, the existing garbage classification and sorting methods have shortcomings: during the garbage sorting process, once the posture of the garbage does not match the pre-stored posture, or even when a piece of garbage is blocked by other garbage, it is impossible to accurately identify the type of garbage, and it is also impossible to classify and sort the garbage to be sorted more accurately, thereby reducing the efficiency of garbage classification and sorting. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide an artificial intelligence garbage classification and logistics sorting method based on the above-mentioned existing technology.

[0005] The technical solution adopted by the present invention to solve the above technical problems is: an artificial intelligence garbage classification and logistics sorting method, which is characterized by comprising the following steps:

[0006] Step 1: Obtain a multi-posture feature set for each type of garbage in advance;

[0007] Step 2: Based on the obtained multi-posture feature sets corresponding to each type of garbage, multi-posture garbage attribute recognition models are constructed for each type of garbage. Each type of garbage corresponds to a multi-posture garbage attribute recognition model, and each multi-posture garbage attribute recognition model has a multi-posture feature set corresponding to the type of garbage it corresponds to.

[0008] Step 3: Obtain multi-posture features of each actual garbage to be classified;

[0009] Step 4: Match the obtained multi-posture features of each actual garbage with all the constructed multi-posture garbage attribute recognition models:

[0010] When the multi-posture features of any actual garbage are consistent with the multi-posture feature set corresponding to any multi-posture garbage attribute recognition model, it is determined that the actual garbage is the type of garbage corresponding to any multi-posture garbage attribute recognition model, and the sorting operation of the type of garbage corresponding to any multi-posture garbage attribute recognition model is performed; otherwise, it is determined that the actual garbage is garbage of a temporarily unknown type, and the sorting operation of garbage of a temporarily unknown type is performed.

[0011] Improved, in the artificial intelligence garbage classification and logistics sorting method, in step 1, the process of pre-acquiring a multi-posture feature set for each type of garbage includes the following steps:

[0012] Step a1: pre-collecting posture feature groups corresponding to each type of garbage under multiple shooting angles; wherein each shooting angle corresponds to a group of posture feature groups of the garbage, and each type of garbage corresponds to multiple posture feature groups;

[0013] Step a2: Calculate the contribution weight of each posture feature in identifying the corresponding type of garbage based on all posture feature groups collected for each type of garbage; wherein the posture feature corresponds one-to-one to the contribution weight in identifying the corresponding type of garbage;

[0014] Step a3: Arrange the contribution weights corresponding to the posture features in all posture feature groups of each type of garbage in descending order to obtain a descending sequence of contribution weights of the corresponding types of garbage;

[0015] Step a4: Check for duplicate posture features on the obtained multi-posture features of all types of garbage:

[0016] When any posture feature is in the posture feature group of at least two types of garbage, go to step a5; otherwise, go to step a6;

[0017] Step a5: Mark any posture feature as a coincident posture feature, select the posture feature group in which the coincident posture feature has the maximum contribution weight value, and use the selected posture feature group as the multi-posture feature set for the corresponding garbage type;

[0018] In step a6, each multi-gesture feature is used as a multi-gesture feature set of the corresponding type of garbage.

[0019] Furthermore, in the artificial intelligence garbage classification and logistics sorting method, in step 2, the process of constructing the multi-posture garbage attribute recognition model includes the following steps:

[0020] Step b1, pre-collecting a number of garbage images corresponding to the same type of garbage;

[0021] Step b2: extracting edge contours of the garbage recorded in each garbage image, and forming an edge contour set corresponding to the same type of garbage from all the extracted edge contours;

[0022] Step b3, mapping each edge contour in the edge contour set of the same type of garbage into the same plane rectangular coordinate system to obtain corresponding edge contour curves;

[0023] Step b4: extracting all inflection points and all singular points on each edge contour curve to form an inflection point set and a singular point set corresponding to each edge contour curve;

[0024] Step b5, counting the number of repetitions of each inflection point in all inflection point sets and the number of repetitions of each singular point in all singular point sets;

[0025] Step b6, calculating the repetition frequency of each inflection point in the set of all inflection points of the edge contour curve of the same type of garbage and the repetition frequency of each singular point in the set of all singular points of the edge contour curve of the same type of garbage;

[0026] Step b7: taking the inflection points whose repetition frequency is higher than the first preset frequency value as the first posture feature, and forming a first posture feature set from all the first posture features; and taking the singular points whose repetition frequency is higher than the second preset frequency value as the second posture feature, and forming a second posture feature set from all the second posture features;

[0027] Step b8: forming a multi-posture garbage attribute recognition model for the same type of garbage based on the obtained first posture feature set and second posture feature set of the same type of garbage.

[0028] Further improvement, in the artificial intelligence garbage classification and logistics sorting method, the multi-posture garbage attribute recognition model for this type of garbage is constructed as follows:

[0029]

[0030] Among them, u represents a type of garbage to be identified, X u Represents the first posture feature set on the contour curve of the garbage u, Y u Represents the second posture feature set on the contour curve of the type of garbage u, P X,u Represents the first posture feature set X on the contour curve of this type of garbage u u The total number of inflection points, P X,th Represents the first posture feature set X on the preset contour curve of the type of garbage u u The total number of inflection points threshold, P Y,u Represents the second posture feature set Y on the contour curve of this type of garbage u u The total number of internal singular points, P Y,th Represents the second posture feature set Y on the preset contour curve of the type of garbage u u The total number of internal singular points threshold, u m Indicates the mth type of garbage.

[0031] Improved, in the artificial intelligence garbage classification and logistics sorting method, the first preset frequency value is greater than the second preset frequency value, and the total number threshold of inflection points is greater than the total number threshold of singular points.

[0032] As a further improvement, in the artificial intelligence garbage classification and logistics sorting method, in step 3, the multi-posture features of each actual garbage to be classified are collected respectively by a camera device in a 360-degree rotation manner.

[0033] Compared with the prior art, the advantages of the present invention are: the artificial intelligence garbage classification and logistics sorting method of the invention obtains the multi-posture feature set of each type of garbage in advance, and constructs multi-posture garbage attribute recognition models of different types of garbage based on the multi-posture feature set, and then obtains the multi-posture features of each actual garbage to be classified, and matches the obtained multi-posture features of each actual garbage with all the constructed multi-posture garbage attribute recognition models, and then performs the sorting operation of the corresponding type of garbage according to whether the multi-posture feature set corresponding to all the multi-posture garbage attribute recognition models of the multi-posture features of the actual garbage is consistent or not, thereby achieving the effect of identifying the actual type of garbage based on the multi-posture features of the garbage, ensuring that the garbage can be accurately classified, identified and sorted in different postures.

[0034] Description of the attached figure

[0035] Figure 1 This is a flow chart of an artificial intelligence garbage classification and logistics sorting method in an embodiment of the present invention;

[0036] Figure 2 Schematic diagram of the construction process of the multi-posture garbage attribute recognition model in an embodiment of the present invention. DETAILED DESCRIPTION

[0037] The present invention will be described in further detail below with reference to the accompanying drawings and embodiments.

[0038] This embodiment provides an artificial intelligence garbage classification and logistics sorting method. Specifically, the artificial intelligence garbage classification and logistics sorting method in this embodiment includes the following steps 1 to 4:

[0039] Step 1: Pre-acquire a multi-posture feature set for each type of garbage; wherein each multi-posture feature set contains a number of posture features of the corresponding type of garbage;

[0040] Step 2: Based on the obtained multi-pose feature sets corresponding to each type of garbage, multi-pose garbage attribute recognition models for different types of garbage are constructed respectively. Each type of garbage corresponds to a multi-pose garbage attribute recognition model, and each multi-pose garbage attribute recognition model has a multi-pose feature set corresponding to the type of garbage it corresponds to. Of course, according to actual needs, the currently mature YOLOv8 algorithm can be used for model construction and training. For example, when building and training a model based on the YOLOv8 algorithm, the reference code is as follows:

[0041]

[0042] Step 3: Acquire multi-posture features of each piece of actual garbage to be classified. In this embodiment, preferably, the multi-posture features of each piece of actual garbage to be classified are acquired by a camera in a 360-degree rotation manner.

[0043] Step 4: Match the obtained multi-posture features of each actual garbage with all the constructed multi-posture garbage attribute recognition models:

[0044] When the multi-posture features of any actual garbage are consistent with the multi-posture feature set corresponding to any multi-posture garbage attribute recognition model, the actual garbage is determined to be the type of garbage corresponding to the multi-posture garbage attribute recognition model, and the sorting operation of the type of garbage corresponding to the multi-posture garbage attribute recognition model is performed; otherwise, the actual garbage is determined to be garbage of temporarily unknown type, and the sorting operation of garbage of temporarily unknown type is performed. For example, when it is determined that the type of the actual garbage is temporarily unknown, the actual garbage is sorted out and placed in a garbage storage device of unknown type to facilitate subsequent manual sorting and disposal by staff;

[0045] Specifically in this embodiment, in step 1, the process of pre-acquiring a multi-posture feature set for each type of garbage includes the following steps a1 to a6:

[0046] Step a1: pre-collecting posture feature groups corresponding to each type of garbage under multiple shooting angles; wherein each shooting angle corresponds to a group of posture feature groups of the garbage, and each type of garbage corresponds to multiple posture feature groups;

[0047] For example, for type A garbage, the shooting angles are θ1, θ2, ..., θ n The posture feature groups corresponding to Class A garbage are collected in advance, thereby obtaining multiple posture feature groups corresponding to different shooting angles, namely:

[0048] The posture feature group of type A garbage corresponding to the shooting angle θ1 is marked as S1, S1 = {s 11 ,s 12 ,…、s1m};s 1m refers to the mth posture feature in the posture feature group S1;

[0049] The posture feature group of type A garbage corresponding to the shooting angle θ2 is marked as S2, S2 = {s 21 ,s 22 ,…、s 2m};s 2m refers to the mth posture feature in the posture feature group S2;

[0050] …;

[0051] Shooting angle θ n The posture feature group corresponding to type A garbage is marked as S n , S n ={s n1 ,s n2 ,…、s nm};s nm Refers to the posture feature group S n The m-th posture feature in;

[0052] Step a2: Based on all the posture feature groups collected for each type of garbage, calculate the contribution weight of each posture feature in the process of identifying the corresponding garbage type; wherein the posture feature corresponds to the contribution weight in the process of identifying the corresponding garbage type one by one; for example:

[0053] For the posture feature group S1 of type A garbage, after calculation, its posture feature s is obtained 1m The contribution weight in the process of identifying Class A garbage is μ 1m ; Of course, μ 11 +μ 12 +…+μ 1m =1;

[0054] Similarly, for the posture feature group S2 of type A garbage, after calculation, its posture feature s 2m The contribution weight in the process of identifying Class A garbage is μ 2m ; Of course, μ 21 +μ 22 +…+μ 2m =1;

[0055] For the posture feature group S of type A garbage n , after calculation, we get its posture feature s nm The contribution weight in the process of identifying Class A garbage is μ nm ; Of course, μ n1 +μ n2 +…+μ nm =1;

[0056] Step a3: Arrange the contribution weights corresponding to the posture features in all posture feature groups of each type of garbage in descending order to obtain a descending sequence of contribution weights of the corresponding types of garbage;

[0057] Specifically, for the n posture feature groups of Class A garbage, the contribution weights μ corresponding to all the posture features in the n posture feature groups are 11 、μ 12 ,…,μ 1m 、μ 21 、μ 22 ,…,μ 2m ,…,μ n1 、μ n2 ,…,μ nm Arrange in descending order, that is, compare the contribution weights corresponding to all posture features and then arrange them in descending order. The resulting descending sequence of contribution weights corresponding to Class A garbage is marked as Ф;

[0058] Step a4: Check for duplicate posture features on the obtained multi-posture features of all types of garbage:

[0059] When any posture feature is in the posture feature group of at least two types of garbage, go to step a5; otherwise, go to step a6;

[0060] Step a5: Mark any posture feature as a coincident posture feature, select the posture feature group in which the coincident posture feature has the maximum contribution weight value, and use the selected posture feature group as the multi-posture feature set for the corresponding garbage type;

[0061] For example, after checking the posture feature duplication, it is found that the posture feature s 12 It appears in the posture feature group S1 of type A garbage and in the posture feature group S B1 If it appears in 12 Mark as the coincident posture feature, and select the coincident posture feature s 12 The posture feature group S with the maximum contribution weight value B1 , and the posture feature group S will be selected B1 As the multi-posture feature set of the corresponding Class B garbage;

[0062] In step a6, each multi-gesture feature is used as a multi-gesture feature set of the corresponding type of garbage.

[0063] It should be noted that, in this embodiment, although there are multiple posture feature groups for the multi-posture features of a certain type of garbage, after selecting the multi-posture feature set, the posture feature group that best represents the certain type of garbage is selected from the multiple posture feature groups, so as to use it as the multi-posture feature set of the certain type of garbage to prepare for the subsequent characterization of the characteristics of the certain type of garbage and the identification of the type of garbage.

[0064] In addition, in the artificial intelligence garbage classification logistics sorting method of this embodiment, see Figure 2 As shown, in step 2, the construction process of the multi-posture garbage attribute recognition model includes the following steps b1 to b8:

[0065] Step b1, pre-collecting a number of garbage images corresponding to the same type of garbage;

[0066] Step b2: extracting edge contours of the garbage recorded in each garbage image, and forming an edge contour set corresponding to the same type of garbage from all the extracted edge contours;

[0067] Step b3, mapping each edge contour in the edge contour set of the same type of garbage into the same plane rectangular coordinate system to obtain corresponding edge contour curves;

[0068] Step b4: extract all inflection points and all singular points (singular points may also be called "singular points", which are defined in mathematics) on each edge contour curve to form an inflection point set and a singular point set corresponding to each edge contour curve;

[0069] Step b5, counting the number of repetitions of each inflection point in all inflection point sets and the number of repetitions of each singular point in all singular point sets;

[0070] Step b6, calculating the repetition frequency of each inflection point in the set of all inflection points of the edge contour curve of the same type of garbage and the repetition frequency of each singular point in the set of all singular points of the edge contour curve of the same type of garbage;

[0071] For example, suppose that according to statistics, there are M edge contour curves corresponding to type A garbage, namely edge contour curve C1, edge contour curve C2, ..., edge contour curve C M ;

[0072] For the M edge contour curves of Class A garbage, after counting the number of inflection points and singular points on each curve, the repetition frequency of any inflection point Q in the set of all inflection points is obtained, which is marked as p. Q And the repetition frequency of any singular point R in the set of all singular points is marked as p R ; Then, p Q =num Q / NumQ , num Q Num is the number of repetitions of inflection point Q in all inflection point sets. Q is the total number of inflection points in all inflection point sets; p R =num R / Num R , num R is the number of repetitions of the singular point R in all singular point sets, Num R is the total number of singular points in the set of all singular points;

[0073] Step b7: taking the inflection point with a repetition frequency higher than the first preset frequency value as the first posture feature, and forming a first posture feature set from all the first posture features; and taking the singular point with a repetition frequency higher than the second preset frequency value as the second posture feature, and forming a second posture feature set from all the second posture features; wherein, in this embodiment, the first preset frequency value is greater than the second preset frequency value; for example, the first preset frequency value is 90%, and the second preset frequency value is 60%; of course, the first preset frequency value and the second preset frequency value here can be adjusted as needed;

[0074] Step b8: Based on the obtained first posture feature set and second posture feature set of the same type of garbage, a multi-posture garbage attribute recognition model for the garbage is formed. In this embodiment, the multi-posture garbage attribute recognition model for the garbage is constructed as follows:

[0075]

[0076] Among them, u represents a type of garbage to be identified, X u Represents the first posture feature set on the contour curve of the garbage u, Y u Represents the second posture feature set on the contour curve of the type of garbage u, P X,u Represents the first posture feature set X on the contour curve of this type of garbage u u The total number of inflection points, P X,th Represents the first posture feature set X on the preset contour curve of the type of garbage u u The total number of inflection points threshold, P Y,u Represents the second posture feature set Y on the contour curve of this type of garbage u u The total number of internal singular points, P Y,th Represents the second posture feature set Y on the preset contour curve of the type of garbage u u The total number of internal singular points threshold, u m For example, in this embodiment, the total number threshold of inflection points is greater than the total number threshold of singular points.

[0077] Although the preferred embodiments of the present invention have been described in detail above, it should be clearly understood that various modifications and variations are possible for those skilled in the art. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention are intended to be included within the scope of protection of the present invention.

Claims

1. Artificial intelligence garbage classification logistics sorting method, characterized by: The steps include: Step 1: Obtain a multi-posture feature set for each type of garbage in advance; Step 2: Based on the obtained multi-posture feature sets corresponding to each type of garbage, multi-posture garbage attribute recognition models are constructed for each type of garbage. Each type of garbage corresponds to a multi-posture garbage attribute recognition model, and each multi-posture garbage attribute recognition model has a multi-posture feature set corresponding to the type of garbage it corresponds to. Step 3: Obtain multi-posture features of each actual garbage to be classified; Step 4: Match the obtained multi-posture features of each actual garbage with all the constructed multi-posture garbage attribute recognition models: When the multi-posture features of any actual garbage are consistent with the multi-posture feature set corresponding to any multi-posture garbage attribute recognition model, it is determined that the actual garbage is the type of garbage corresponding to any multi-posture garbage attribute recognition model, and the sorting operation of the type of garbage corresponding to any multi-posture garbage attribute recognition model is performed; otherwise, it is determined that the actual garbage is garbage of a temporarily unknown type, and the sorting operation of garbage of a temporarily unknown type is performed.

2. The artificial intelligence garbage classification and logistics sorting method according to claim 1 is characterized in that: In step 1, the process of pre-acquiring a multi-pose feature set for each type of garbage includes the following steps: Step a1: pre-collecting posture feature groups corresponding to each type of garbage under multiple shooting angles; wherein each shooting angle corresponds to a group of posture feature groups of the garbage, and each type of garbage corresponds to multiple posture feature groups; Step a2: Calculate the contribution weight of each posture feature in identifying the corresponding type of garbage based on all posture feature groups collected for each type of garbage; wherein the posture feature corresponds to the contribution weight in identifying the corresponding type of garbage on a one-to-one basis; Step a3: Arrange the contribution weights corresponding to the posture features in all posture feature groups of each type of garbage in descending order to obtain a descending sequence of contribution weights of the corresponding types of garbage; Step a4: Check for duplicate posture features on the obtained multi-posture features of all types of garbage: When any posture feature is in the posture feature group of at least two types of garbage, go to step a5; otherwise, go to step a6; Step a5: Mark any posture feature as a coincident posture feature, select the posture feature group in which the coincident posture feature has the maximum contribution weight value, and use the selected posture feature group as the multi-posture feature set for the corresponding garbage type; In step a6, each multi-gesture feature is used as a multi-gesture feature set of the corresponding type of garbage.

3. The artificial intelligence garbage classification and logistics sorting method according to claim 2 is characterized in that: In step 2, the process of constructing the multi-posture garbage attribute recognition model includes the following steps: Step b1, pre-collecting a number of garbage images corresponding to the same type of garbage; Step b2: extracting edge contours of the garbage recorded in each garbage image, and forming an edge contour set corresponding to the same type of garbage from all the extracted edge contours; Step b3, mapping each edge contour in the edge contour set of the same type of garbage into the same plane rectangular coordinate system to obtain corresponding edge contour curves; Step b4: extracting all inflection points and all singular points on each edge contour curve to form an inflection point set and a singular point set corresponding to each edge contour curve; Step b5, counting the number of repetitions of each inflection point in all inflection point sets and the number of repetitions of each singular point in all singular point sets; Step b6, calculating the repetition frequency of each inflection point in the set of all inflection points of the edge contour curve of the same type of garbage and the repetition frequency of each singular point in the set of all singular points of the edge contour curve of the same type of garbage; Step b7, taking the inflection points whose repetition frequency is higher than the first preset frequency value as the first posture feature, and forming a first posture feature set from all the first posture features; and, taking singular points with a repetition frequency higher than a second preset frequency value as second posture features, and forming a second posture feature set from all the second posture features; Step b8: forming a multi-posture garbage attribute recognition model for the same type of garbage based on the obtained first posture feature set and second posture feature set of the same type of garbage.

4. The artificial intelligence garbage classification and logistics sorting method according to claim 3 is characterized in that: The multi-posture garbage attribute recognition model for this type of garbage is constructed as follows: Among them, u represents a type of garbage to be identified, X u Represents the first posture feature set on the contour curve of the garbage u, Y u Represents the second posture feature set on the contour curve of the type of garbage u, P X,u Represents the first posture feature set X on the contour curve of this type of garbage u u The total number of inflection points, P X,th Represents the first posture feature set X on the preset contour curve of the type of garbage u u The total number of inflection points threshold, P Y,u Represents the second posture feature set Y on the contour curve of this type of garbage u u The total number of internal singular points, P Y,th Represents the second posture feature set Y on the preset contour curve of the type of garbage u u The total number of internal singular points threshold, u m Indicates the mth type of garbage.

5. The artificial intelligence garbage classification and logistics sorting method according to claim 3 is characterized in that: The first preset frequency value is greater than the second preset frequency value, and the total number threshold of inflection points is greater than the total number threshold of singular points.

6. The artificial intelligence garbage classification and logistics sorting method according to claim 1 is characterized in that: In step 3, the multi-posture features of each piece of actual garbage to be classified are collected by a camera device in a 360-degree rotation manner.

Images