Recognition method, device and equipment for disorderly stacked materials and medium

Abstract

The invention relates to the field of image processing and computer vision, in particular to a method, a device, equipment and a medium for recognizing disordered materials, and the method comprises the steps: obtaining a to-be-detected image of a target place and a standard image corresponding to the to-be-detected image, the standard image being an image when no disordered materials exist in the target place; extracting image difference features between the to-be-detected image and the standard image; inputting the image difference features into a pre-trained disordered material identification model to obtain a disordered material identification result; wherein the disordered material identification model is obtained by performing supervised training on a plurality of training sample image pairs, and each training sample image pair comprises a standard image sample in the same place and a to-be-detected image sample with a disordered material label; and determining whether the disordered materials exist in the target place or not according to the disordered material identification result. According to the method for identifying the disorderly-stacked materials, the disorderly-stacked materials of any type can be accurately identified, and the adaptability to complex backgrounds is high.

Description

Technical Field

[0001] The present invention relates to the fields of image processing and computer vision, and in particular to a method, device, equipment and medium for identifying disorderly piled materials. Background Art

[0002] In industrial production and logistics management, the disorderly stacking of materials will have a serious impact on safety and efficiency, especially the problem of disorderly stacking of materials occupying passages, which will cause passage blockages, affect logistics transportation and personnel passage, and may even cause safety accidents.

[0003] Traditional methods for identifying bulk materials usually rely on manual observation or equipment detection.

[0004] Manual observation refers to sending a large number of management personnel to patrol or check surveillance videos, and record and handle any cases of disorderly stacking of materials. However, this method has the problems of extremely high personnel management costs, low work efficiency, difficult data analysis and low real-time performance, and it is impossible to comprehensively and real-timely monitor the entire production and stacking area.

[0005] Equipment discovery refers to monitoring the channel through sensors to actively discover the behavior of random stacking of materials. Its working principle is usually to determine whether there is random stacking of materials by detecting whether there is an object at a fixed position. However, this method has problems such as limited detection range, low recognition efficiency, and high false alarm rate.

[0006] With the development of artificial intelligence technology, especially the gradual maturity of image processing and computer vision technology, it has become possible to use intelligent monitoring systems to conduct real-time analysis and automatic identification of material stacking at production sites. However, existing monitoring systems usually use target detection, image segmentation and other technologies to identify randomly piled materials. These methods usually rely on complex feature extraction methods to distinguish materials from the background. They are not adaptable to complex backgrounds, are greatly affected by light changes, can only identify specific types of materials (such as cartons, steel, foam, etc.), and are difficult to identify and locate materials of unknown categories. Summary of the invention

[0007] In view of the above problems in the prior art, the purpose of the present invention is to provide a method, device, equipment and medium for identifying bulk materials, which can not only accurately identify any type of bulk materials, but also have strong adaptability to complex backgrounds.

[0008] In order to solve the above problems, the present invention provides a method for identifying disorderly piled materials, comprising:

[0009] Acquire an image to be tested of the target location and a standard image corresponding to the image to be tested, wherein the standard image is an image when there is no random pile of materials in the target location;

[0010] Extracting image difference features between the image to be tested and the standard image;

[0011] Inputting the image difference features into a pre-trained shuffled material recognition model to obtain a shuffled material recognition result; wherein the shuffled material recognition model is obtained by supervised training using a plurality of training sample image pairs, each training sample image pair including a standard image sample in the same place and a test image sample with a shuffled material label;

[0012] Determine whether there are piles of materials in the target location based on the piled materials identification results.

[0013] Furthermore, the image difference features between the image to be tested and the standard image are extracted, including:

[0014] The histogram difference feature, mean difference feature and standard deviation difference feature between the image to be tested and the standard image are extracted respectively;

[0015] The histogram difference features, mean difference features and standard deviation difference features are concatenated to obtain the image difference features.

[0016] Furthermore, the image difference features between the image to be tested and the standard image are extracted, including:

[0017] The image to be tested and the standard image are processed into blocks to obtain a plurality of standard-to-test image block pairs;

[0018] Image block difference features between the standard image block and the image block to be tested in each standard-to-test image block pair are extracted respectively.

[0019] Furthermore, the image difference features are input into a pre-trained piled material recognition model to obtain a piled material recognition result, including:

[0020] Inputting the image block difference features corresponding to each standard-to-test image block pair into the disorderly pile material recognition model to obtain multiple sub-recognition results;

[0021] Among them, determining whether there are disorderly piles of materials in the target place according to the disorderly piles of materials identification results includes:

[0022] If at least one of the multiple sub-identification results indicates that there is a disorderly pile of materials, it is determined that there is a disorderly pile of materials in the target place;

[0023] The method further comprises:

[0024] The position corresponding to the standard-to-be-tested image block pair corresponding to the sub-recognition result indicating the existence of the disorderly piled materials is used as the position of the disorderly piled materials in the target place.

[0025] Furthermore, the method further comprises:

[0026] Preprocessing the image to be tested and the standard image respectively to obtain a preprocessed image to be tested and a preprocessed standard image; wherein the preprocessing may include one or more of adaptive gamma correction, color space conversion and demarcation of the area to be tested;

[0027] The image difference features between the image to be tested and the standard image are extracted, including:

[0028] The image difference features between the preprocessed test image and the preprocessed standard image are extracted.

[0029] Furthermore, the disorderly pile material recognition model is a support vector machine model, or the disorderly pile material recognition model includes a plurality of support vector machine models with different kernel functions.

[0030] Furthermore, the method also includes pre-training a random material identification model, and the training method of the random material identification model includes:

[0031] Acquire a plurality of training sample image pairs, each training sample image pair comprising a standard image sample and an image sample to be tested in the same place;

[0032] respectively extracting image sample difference features between the standard image sample and the image sample to be tested in each training sample image pair;

[0033] The image samples to be tested in each training sample image pair are respectively labeled to obtain the image samples to be tested with a jumbled material label, where the jumbled material label is used to identify whether there is a jumbled material in the corresponding image samples to be tested;

[0034] The image sample difference features corresponding to each training sample image pair are used as input, and the corresponding disorderly pile material labels are used as supervision, and the initial recognition model is trained in a supervised manner to obtain the disorderly pile material recognition model.

[0035] Another aspect of the present invention provides a device for identifying disorderly piled materials, comprising:

[0036] An image acquisition module is used to acquire an image to be tested of the target location and a standard image corresponding to the image to be tested, wherein the standard image is an image when there is no random pile of materials in the target location;

[0037] A feature extraction module is used to extract image difference features between the image to be tested and the standard image;

[0038] A disorderly pile material recognition module is used to input image difference features into a pre-trained disorderly pile material recognition model to obtain disorderly pile material recognition results; wherein the disorderly pile material recognition model is obtained by supervised training using multiple training sample image pairs, each training sample image pair includes a standard image sample in the same place and a test image sample with a disorderly pile material label;

[0039] The disorderly pile material determination module is used to determine whether disorderly pile materials exist in the target place according to the disorderly pile material identification result.

[0040] Another aspect of the present invention provides an electronic device, including a processor and a memory, wherein the memory stores at least one instruction or at least one program, and the at least one instruction or the at least one program is loaded and executed by the processor to implement the method for identifying disorderly piled materials as described above.

[0041] Another aspect of the present invention provides a computer-readable storage medium, in which at least one instruction or at least one program is stored. The at least one instruction or the at least one program is loaded and executed by a processor to implement the method for identifying disorderly piled materials as described above.

[0042] Another aspect of the present invention provides a computer program product, comprising a computer program / instruction, which implements the above-mentioned method for identifying disorderly piled materials when executed by a processor.

[0043] Due to the above technical solution, the present invention has the following beneficial effects:

[0044] According to the method for identifying piled materials in an embodiment of the present invention, the difference between the standard image and the image to be tested is used as a feature, and a pre-trained piled material identification model is used to identify the piled materials. This method based on image feature recognition rather than material type recognition can not only cope with the situation of any type of material being piled, but also accurately identify any type of piled materials, and the calculation is simpler in the feature extraction stage. There is no need to design a complex feature extraction method to distinguish between the piled materials and the background, and the interference of different backgrounds can be ignored. It has strong adaptability to complex backgrounds. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the technical solution of the present invention, the following briefly introduces the drawings required for use in the embodiments or prior art descriptions. Obviously, the drawings described below are only some embodiments of the present invention, and for those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative work.

[0046] Figure 1 It is a schematic diagram of an implementation environment provided by an embodiment of the present invention;

[0047] Figure 2 is a flow chart of a method for identifying disorderly piled materials provided by one embodiment of the present invention;

[0048] Figure 3 is a flow chart of a method for identifying disorderly piled materials provided by another embodiment of the present invention;

[0049] Figure 4 It is a flow chart of a method for training a disorderly pile material recognition model provided by one embodiment of the present invention;

[0050] Figure 5 It is a schematic diagram of the structure of a device for identifying disorderly piled materials provided by one embodiment of the present invention;

[0051] Figure 6 is a block diagram of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0052] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0053] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, device, product or equipment that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or equipment.

[0054] In order to make the purpose, technical solution and advantages disclosed in the embodiments of the present invention more clearly understood, the embodiments of the present invention are further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the embodiments of the present invention and are not used to limit the embodiments of the present invention. First, the embodiments of the present invention explain the following concepts:

[0055] RGB space: RGB space is based on the three basic colors of red (Red, R), green (Green, G), and blue (Blue, B), and superimposes them to varying degrees to produce rich and wide-ranging colors. It is commonly known as the three-primary color model. RGB space is the most commonly used model.

[0056] LAB space: LAB space is a color-opposition space (Lab Color Space). Channel L in the LAB space represents brightness, and channel A and channel B represent color opposition dimensions.

[0057] MAX-MIN standardization: Also known as deviation standardization or normalization, it is a data preprocessing technique used to transform numerical data to a common scale, usually in the range of 0 to 1. This transformation helps to eliminate the effects of different dimensions and magnitudes, making the data comparable when compared, weighted, or otherwise analyzed.

[0058] Support Vector Machine (SVM): SVM is a classic supervised learning algorithm, which is widely used in classification, regression and outlier detection tasks. Its core idea is to find an optimal hyperplane in the feature space for classification with the largest margin.

[0059] Reference Manual Attached Figure 1 , which shows a schematic diagram of the implementation environment of a method for identifying disorderly piled materials provided by an embodiment of the present invention, such as Figure 1 As shown, the implementation environment may include at least one monitoring device 110 and a server 120. The server 120 and each monitoring device 110 may be directly or indirectly connected via wired or wireless communication, which is not limited in the embodiment of the present invention. For example, each monitoring device 110 may be connected to the server 120 via a converter / router.

[0060] Among them, the monitoring device 110 can be a common monitoring camera, a hard disk recorder, etc. The server 120 can be an independently operated server, or a server cluster or distributed system composed of multiple servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content delivery networks (CDN), and big data and artificial intelligence platforms.

[0061] In an embodiment of the present invention, the monitoring device 110 can be set at the target place to be monitored, and the monitoring video or monitoring image of the target place can be collected in real time, so as to obtain the image to be tested of the target place. The monitoring device 110 can send the collected monitoring video or monitoring image to the server 120. After receiving the monitoring video or monitoring image sent by the monitoring device 110, the server 120 can obtain the image to be tested of the target place from the monitoring video or can directly use the received monitoring image as the image to be tested of the target place, and use the method for identifying the disorderly piled materials provided in the embodiment of the present invention to identify whether there are disorderly piled materials in the target place based on the image to be tested.

[0062] It should be noted that Figure 1 The implementation environment shown is only exemplary. In actual applications, the implementation environment may also include Figure 1 There may be more or fewer components shown in the figure, and the embodiments of the present invention are not limited to this.

[0063] Reference Manual Attached Figure 2 , which shows the process of a method for identifying disorderly piled materials provided by an embodiment of the present invention, which can be applied to Figure 1 In the server 120, specifically Figure 2 As shown, the method may include the following steps:

[0064] S210: Acquire an image to be tested of the target location and a standard image corresponding to the image to be tested, wherein the standard image is an image when there are no randomly piled materials in the target location.

[0065] In the embodiment of the present invention, the target place may be various places where bulky materials need to be processed, such as places including but not limited to passages in industrial production and logistics management places, and the embodiment of the present invention does not limit this.

[0066] In an embodiment of the present invention, a monitoring video or monitoring image of the target place in an actual production scenario can be collected in real time by a monitoring device arranged in the target place, and sent to a server. The server can obtain one or more frames of images from the received monitoring video as images to be tested, or can directly use the received monitoring image as the image to be tested. It is also possible to obtain a monitoring image of the target place without random piles of materials in advance through the monitoring device as a standard image corresponding to the image to be tested of the target place. The image to be tested of the target place and the corresponding standard image can constitute a standard-image pair to be tested.

[0067] S220: Extracting image difference features between the image to be tested and the standard image.

[0068] In an embodiment of the present invention, one or more difference features can be first extracted from the test image and the standard image, and then spliced ​​to obtain the final image difference features. The one or more difference features may include but are not limited to histogram difference features, mean difference features, and standard deviation difference features, etc., and the embodiment of the present invention is not limited to this.

[0069] In some embodiments, extracting image difference features between the image to be tested and the standard image may include: respectively extracting histogram difference features, mean difference features, and standard deviation difference features between the image to be tested and the standard image; and splicing the histogram difference features, mean difference features, and standard deviation difference features to obtain image difference features.

[0070] Specifically, the image to be tested and the standard image can be converted into color space respectively, the image can be converted from RGB space to LAB space, and the L channel (i.e., brightness channel) can be discarded to reduce the influence of brightness on image features. Then the histograms of the A and B channels can be calculated for the image to be tested and the standard image respectively, and MAX-MIN standardization can be performed. Finally, the difference between the histograms of the A and B channels of the image to be tested and the histograms of the A and B channels of the standard image can be calculated respectively to obtain the histogram difference features. Among them, the number of histogram features can be predetermined according to actual conditions, for example, it can be 256, 128, 64, 48, 32, 24, 16, etc., preferably the same as the number of features in the histogram difference features in the input data of the adopted random material recognition model, and the embodiment of the present invention does not limit this.

[0071] Specifically, the absolute value of the difference in pixel values ​​in channel A between the image to be tested and the standard image, as well as the absolute value of the difference in pixel values ​​in channel B between the image to be tested and the standard image can be calculated first, and MAX-MIN normalization can be performed. Then, the mean and standard deviation of the absolute values ​​of the difference in pixel values ​​in channels A and B can be calculated respectively to obtain the mean difference feature and the standard deviation difference feature.

[0072] Specifically, the existing feature splicing method can be used to splice the histogram difference feature, the mean difference feature and the standard deviation difference feature to obtain the image difference feature, which will not be described in detail in the embodiment of the present invention.

[0073] It can be understood that the embodiment of the present invention extracts histogram difference features, mean difference features and standard deviation difference features as input features of the disorderly piled material recognition model, which not only simplifies the calculation in the feature extraction stage but also has a good recognition effect.

[0074] In some embodiments, after acquiring the image to be tested and the standard image, the image to be tested and the standard image may be preprocessed to obtain a preprocessed image to be tested and a preprocessed standard image. The preprocessing may include, but is not limited to, one or more of adaptive gamma correction, color space conversion, and demarcation of the area to be tested.

[0075] Specifically, adaptive gamma correction can be performed on the image to be tested and the standard image, respectively, to adjust the brightness of the image. Color space conversion can also be performed on the image to be tested and the standard image, respectively, converting the image from RGB space to LAB space, and discarding the L channel (i.e., brightness channel) to reduce the impact of brightness on image features. And / or, the image to be tested and the standard image can be delineated as a detection area, and all pixel values ​​of the image outside the area can be set to 0 according to the pre-set polygonal detection area. The specific implementation process of the above-mentioned various preprocessing can refer to the prior art, and the embodiments of the present invention will not be repeated here.

[0076] Accordingly, extracting the image difference features between the image to be tested and the standard image (step S220) may be: extracting the image difference features between the preprocessed image to be tested and the preprocessed standard image. The specific extraction method may refer to the relevant content in step S220, and the embodiment of the present invention will not be repeated here.

[0077] It should be noted that the above-mentioned preprocessing methods are only examples. In practical applications, other types of preprocessing may be performed on the image to be tested and the standard image, and the embodiment of the present invention does not limit this.

[0078] It can be understood that the embodiments of the present invention can reduce the impact of different lighting conditions on the recognition results through preprocessing such as adaptive gamma correction and discarding the L channel after converting the LAB space, so as to adapt to complex light changes and further improve the accuracy of identifying disorderly piled materials.

[0079] S230: Inputting the image difference features into a pre-trained shuffled material recognition model to obtain a shuffled material recognition result.

[0080] In the embodiment of the present invention, the disorderly piled material recognition model is obtained by using a plurality of training sample image pairs for supervised training, and each training sample image pair includes a standard image sample in the same place and a to-be-tested image sample with a disorderly piled material label.

[0081] Specifically, multiple training sample image pairs in different places can be acquired in advance, and each training sample image pair includes a standard image sample and an image sample to be tested in the same place. Then, the image sample difference features between the standard image sample and the image sample to be tested in each training sample image pair can be extracted respectively. Then, according to the actual situation of the disorderly piled materials, the image samples to be tested in each training sample image pair can be labeled respectively to obtain the corresponding disorderly piled material labels. Finally, the image sample difference features corresponding to each training sample image pair can be used as input, and the corresponding disorderly piled material labels can be used as supervision to perform supervised training on the initial recognition model to obtain a disorderly piled material recognition model.

[0082] In some embodiments, after obtaining a plurality of training sample image pairs, the standard image samples and the image samples to be tested in each training sample image pair may be subjected to image block processing to obtain a plurality of standard-to-test image sample block pairs. Then, the image sample block difference features between the standard image sample blocks and the image sample blocks to be tested in each standard-to-test image sample block pair are extracted respectively. Then, the image sample block difference features corresponding to each standard-to-test image sample block pair are used as input, and the corresponding piled material labels of the image samples to be tested are used as supervision to perform supervised training on the initial recognition model to obtain a piled material recognition model.

[0083] Specifically, the initial recognition model can be pre-built according to actual needs, for example, it can include but not limited to classification models such as support vector machine models, that is, the disorderly pile material recognition model can also include but not limited to classification models such as trained support vector machine models. The specific training method of the model will be described in detail later.

[0084] In some embodiments, the disorderly pile material recognition model may be a SVM model, or the disorderly pile material recognition model may also include multiple SVM models with different kernel functions.

[0085] Optionally, when the bulk material recognition model is an SVM model, the bulk material recognition model may be various SVM models with different kernel functions, for example, an SVM model with a linear kernel, a Gaussian kernel or a polynomial kernel, and the embodiment of the present invention is not limited to this.

[0086] Optionally, when the disorderly pile material recognition model includes multiple SVM models with different kernel functions, the number of SVM models and the kernel function of each SVM model can be pre-set according to actual needs. For example, the number of SVM models can be set to 3, which are SVM models with linear kernel, Gaussian kernel and polynomial kernel respectively. The embodiment of the present invention is not limited to this.

[0087] In an embodiment of the present invention, the extracted image difference features can be directly input into a shuffled material recognition model to obtain a shuffled material recognition result output by the model, and the shuffled material recognition result can indicate whether shuffled materials are recognized in the image to be tested.

[0088] Specifically, when the shuffled material recognition model is an SVM model, a shuffled material recognition result can be obtained, and the result of whether the shuffled material is present can be directly determined based on the shuffled material recognition result. When the shuffled material recognition model includes multiple SVM models with different kernel functions, multiple shuffled material recognition results can be obtained, and then the multiple shuffled material recognition results can be combined to finally determine the result of whether the shuffled material is present.

[0089] It can be understood that the embodiment of the present invention can further ensure the accuracy of identifying the disorderly piled materials by combining the recognition results of multiple SVM models with different kernel functions to comprehensively determine whether there are disorderly piled materials.

[0090] S240: Determine whether there are piled materials in the target place according to the piled materials identification result.

[0091] In an embodiment of the present invention, when the piled material recognition result indicates that the piled materials are recognized in the image to be tested, it can be determined that there are piled materials in the target place; when the piled material recognition result indicates that the piled materials are not recognized in the image to be tested, it can be determined that there are no piled materials in the target place.

[0092] Specifically, when the piled material recognition model is an SVM model, it can be directly determined whether there are piled materials in the target place according to the piled material recognition result output by the SVM model. When the piled material recognition model includes multiple SVM models with different kernel functions, it can be determined that there are piled materials in the target place only when the piled material recognition results output by each SVM model indicate that the piled materials are recognized. If there is at least one SVM model that outputs a piled material recognition result indicating that the piled materials are not recognized, it can be determined that there are no piled materials in the target place.

[0093] In summary, according to the method for identifying piled materials in an embodiment of the present invention, a pre-trained piled material identification model is used to identify piled materials by using the difference between a standard image and an image to be tested as a feature. This method based on image feature recognition rather than material type recognition can not only cope with the situation of any type of material being piled, but also accurately identify any type of piled materials. In addition, the calculation is simpler in the feature extraction stage. There is no need to design a complex feature extraction method to distinguish between piled materials and backgrounds. The interference of different backgrounds can be ignored, and the adaptability to complex backgrounds is strong.

[0094] Reference Manual Attached Figure 3, which shows the process of a method for identifying disorderly piled materials provided by another embodiment of the present invention, which can be applied to Figure 1 In the server 120, specifically Figure 3 As shown, the method may include the following steps:

[0095] S310: Acquire an image to be tested of the target location and a standard image corresponding to the image to be tested, wherein the standard image is an image when there are no randomly piled materials in the target location.

[0096] S320: performing image block processing on the image to be tested and the standard image respectively to obtain a plurality of standard-image block pairs to be tested.

[0097] In the embodiment of the present invention, before extracting the image features, the image to be tested and the standard image may be divided into blocks using an existing block division method to obtain a plurality of standard-image to be tested block pairs.

[0098] Specifically, the number of blocks of the image can be preset according to actual needs. For example, the image can be set to be divided into 10×10 image blocks, which is not limited in the embodiment of the present invention.

[0099] Specifically, image blocks in the image to be tested and the standard image that have corresponding positions may be paired to form a plurality of standard-image block pairs to be tested.

[0100] S330: extracting image block difference features between the standard image block and the image block to be tested in each standard-to-test image block pair.

[0101] In the embodiment of the present invention, the method for extracting the image block difference feature between the standard image block and the image block to be tested in the standard-to-test image block pair can refer to Figure 2 The relevant contents of extracting the image difference features between the image to be tested and the standard image in the illustrated embodiment will not be described in detail in the embodiment of the present invention.

[0102] In some embodiments, after obtaining the image to be tested and the standard image, the image to be tested and the standard image may be preprocessed to obtain a preprocessed image to be tested and a preprocessed standard image. The preprocessing may include but is not limited to one or more processes of adaptive gamma correction, color space conversion, and demarcation of the area to be tested. For specific preprocessing methods, please refer to Figure 2 The relevant contents in the illustrated embodiments will not be described in detail in the embodiments of the present invention.

[0103] Accordingly, the image to be tested and the standard image are respectively subjected to image block processing to obtain a plurality of standard-image to be tested block pairs (step S320), which can be: the preprocessed image to be tested and the preprocessed standard image are respectively subjected to image block processing to obtain a plurality of standard-image to be tested block pairs.

[0104] Specifically, after the preprocessed image to be tested and the preprocessed standard image are divided into blocks, image blocks with effective portions exceeding 50% can be selected as image blocks to be detected according to the area to be detected, and then paired to form multiple standard-image block pairs to be tested.

[0105] S340: Inputting the image block difference features corresponding to each standard-to-test image block pair into the disorderly pile material recognition model to obtain multiple sub-recognition results.

[0106] In the embodiment of the present invention, each standard-to-test image block pair can be individually detected to see if it is in a disorderly pile, thereby realizing the location of the disorderly pile area. Specifically, the image difference features corresponding to each standard-to-test image block pair can be input into a disorderly pile material recognition model to obtain a disorderly pile material recognition result output by the model as a sub-recognition result corresponding to the to-test image block, and the disorderly pile material recognition result can indicate whether a disorderly pile material is recognized in the to-test image block.

[0107] In some embodiments, the disorderly pile material recognition model may be a SVM model, or the disorderly pile material recognition model may also include multiple SVM models with different kernel functions.

[0108] Specifically, when the jumbled material recognition model is an SVM model, a jumbled material recognition result can be obtained for each standard-to-test image block pair, and the sub-recognition result corresponding to the to-be-tested image block can be directly determined based on the jumbled material recognition result. When the jumbled material recognition model includes multiple SVM models with different kernel functions, multiple jumbled material recognition results can be obtained for each standard-to-test image block pair, and then the sub-recognition result corresponding to the to-be-tested image block can be determined by combining the multiple jumbled material recognition results. Among them, the existence of jumbled materials can be determined only when the jumbled material recognition results output by each SVM model indicate that jumbled materials are recognized. If there is at least one SVM model that outputs a jumbled material recognition result indicating that jumbled materials are not recognized, it can be determined that no jumbled materials exist.

[0109] S350: If at least one of the multiple sub-recognition results indicates that there is a pile of materials, it is determined that there is a pile of materials in the target place.

[0110] In an embodiment of the present invention, the sub-recognition results corresponding to all image blocks in a test image can be summarized to determine whether there are piled materials in the target place. Specifically, if at least one sub-recognition result among multiple sub-recognition results indicates that there are piled materials, it can be determined that there are piled materials in the target place. If all sub-recognition results indicate that there are no piled materials, it can be determined that there are no piled materials in the target place.

[0111] In some embodiments, the method may further include: using the position corresponding to the standard-to-be-tested image block pair corresponding to the sub-recognition result indicating the existence of the disorderly piled materials as the position of the disorderly piled materials in the target location. Specifically, the position corresponding to the standard-to-be-tested image block pair that identifies the disorderly piled materials may be directly used as the position of the disorderly piled materials, thereby realizing the positioning function of the disorderly piled materials.

[0112] It should be noted that other contents of each step in the embodiment of the present invention can be referred to Figure 2 The relevant contents of the method provided in the illustrated embodiment will not be repeated in the embodiments of the present invention.

[0113] To sum up, according to the method for identifying disorderly piled materials in an embodiment of the present invention, by dividing the detection image and the standard image into blocks, the disorderly piled materials in each standard-to-be-tested image block pair can be identified respectively. This method can not only identify the disorderly piled materials in the production environment, but also locate the disorderly piled materials, and is more widely applicable.

[0114] The following is a detailed description of the training method for the disorderly pile material recognition model provided by the embodiment of the present invention.

[0115] Reference Manual Attached Figure 4 , which shows the process of a training method for a disorderly pile material recognition model provided by an embodiment of the present invention, which can be applied to Figure 1 In the server 120, specifically Figure 4 As shown, the method may include the following steps:

[0116] S410: Acquire a plurality of training sample image pairs, each training sample image pair including a standard image sample and a to-be-tested image sample in the same location.

[0117] In an embodiment of the present invention, multiple training sample image pairs in different places can be pre-acquired as a training data set for training a piled material recognition model. Each training sample image pair may include a standard image sample and a test image sample in the same place. In other words, the two image samples in the same training sample image pair are images in the same place. In addition, the images in different training sample image pairs may be images in the same place or in different places, which is not limited in the embodiment of the present invention.

[0118] In practical applications, images captured by various surveillance cameras in actual production scenarios can be collected, including those with and without cluttered materials. For each surveillance camera, an image without cluttered materials can be selected as a standard image sample, and the remaining images can be selected as test image samples. Each standard image sample is paired with the test image sample corresponding to the surveillance camera one by one to form multiple training sample image pairs.

[0119] S420: extracting image sample difference features between the standard image sample and the image sample to be tested in each training sample image pair respectively.

[0120] In an embodiment of the present invention, one or more difference features can be first extracted from the standard image sample and the image sample to be tested, and then spliced ​​to obtain the final image sample difference features. The one or more difference features may include but are not limited to histogram difference features, mean difference features, standard deviation difference features, etc., and the embodiment of the present invention does not limit this.

[0121] In the embodiment of the present invention, when extracting histogram difference features, the number of histogram features can be predetermined according to actual conditions, for example, it can be 256, 128, 64, 48, 32, 24, 16, etc., and the embodiment of the present invention does not limit this.

[0122] In some embodiments, since the total number of histogram features is 256, when extracting histogram difference features, different numbers of histogram features can be extracted, and histogram difference features with different feature numbers can be used to train different random material recognition models, and then the random material recognition model with the best recognition effect is selected as the final model.

[0123] For example, as shown in Table 1 below, histogram features with feature numbers of 256, 128, 64, 48, 32, 24, and 16 can be extracted, and then 7 chaotic material recognition models as shown in Table 1 below can be trained to obtain, and the image block precision, image block recall, training set false alarm rate, and test set false alarm rate of each model can be determined. Then, model 5 with the lowest test set false alarm rate can be selected as the final chaotic material recognition model.

[0124] Table 1 Output results of piled material recognition

[0125]

[0126] It should be noted that the method for extracting the image sample difference features between the standard image sample and the image sample to be tested can refer to Figure 2 The relevant contents of extracting the image difference features between the image to be tested and the standard image in the illustrated embodiment will not be described in detail in the embodiment of the present invention.

[0127] In some embodiments, before extracting the difference features of the image samples, the standard image samples and the image samples to be tested may be preprocessed to obtain the preprocessed standard image samples and the image samples to be tested. The preprocessing may include but is not limited to one or more of adaptive gamma correction, color space conversion, and demarcation of the area to be detected. For specific preprocessing methods, please refer to Figure 2 The relevant contents in the illustrated embodiments will not be described in detail in the embodiments of the present invention.

[0128] S430: Label the image samples to be tested in each training sample image pair respectively to obtain the image samples to be tested with a jumbled material label, where the jumbled material label is used to identify whether there is a jumbled material in the corresponding image sample to be tested.

[0129] In the embodiment of the present invention, the image samples to be tested in each training sample image pair can be labeled according to the actual situation of the jumbled materials, so as to obtain corresponding jumbled material labels.

[0130] Specifically, when the image sample to be tested is an image with jumbled materials, the image sample to be tested can be marked as jumbled; when the image sample to be tested is an image without jumbled materials, the image sample to be tested can be marked as not jumbled.

[0131] S440: Using the image sample difference features corresponding to each training sample image pair as input and the corresponding random material labels as supervision, the initial recognition model is supervised and trained to obtain a random material recognition model.

[0132] In the embodiment of the present invention, an initial recognition model may be pre-constructed. The initial recognition model may include but is not limited to classification models such as a support vector machine model. The method for constructing the model will not be described in detail here.

[0133] In some embodiments, the initial recognition model may be an SVM model, or the initial recognition model may include multiple SVM models with different kernel functions.

[0134] Specifically, when the initial recognition model is an SVM model, the initial recognition model can be various SVM models with different kernel functions, such as an SVM model with a linear kernel, a Gaussian kernel, or a polynomial kernel, which is not limited in the embodiment of the present invention. At this time, the trained random pile material recognition model can also be an SVM model.

[0135] At this time, an SVM model with a preset kernel function can be pre-constructed as an initial recognition model, and the disorderly pile material recognition model can be directly trained based on the initial recognition model. Among them, the preset kernel function can be pre-set according to actual needs, for example, it can be set to a linear kernel, a Gaussian kernel or a polynomial kernel function, etc., and the embodiment of the present invention does not limit this.

[0136] Specifically, when the initial recognition model includes multiple SVM models with different kernel functions, the number of SVM models and the kernel function of each SVM model can be preset according to actual needs. For example, the number of SVM models can be set to 3, which are SVM models with linear kernel, Gaussian kernel and polynomial kernel, respectively. The embodiment of the present invention does not limit this. At this time, the trained random pile material recognition model can also include multiple SVM models with different kernel functions.

[0137] At this time, multiple SVM models with different kernel functions can be pre-constructed as initial recognition models. Based on the multiple initial recognition models, training is performed separately to obtain multiple trained SVM models, thereby forming a disorderly material recognition model.

[0138] In an embodiment of the present invention, during the training process, the labeled training sample image pairs can be divided into a training set and a test set, and the constructed initial recognition model can be supervised by using the data in the training set to adjust the model parameters so that it can recognize the disorderly piled materials, thereby obtaining the disorderly piled material recognition model. Specifically, the image sample difference features corresponding to each training sample image pair can be used as input, and the corresponding disorderly piled material labels can be used as supervision to train the initial recognition model, thereby obtaining the disorderly piled material recognition model.

[0139] In an embodiment of the present invention, the trained random pile material recognition model can also be tested using the data in the test set to evaluate its detection performance for random pile materials. When the detection performance of the model meets the requirements, the training is terminated to obtain the final random pile material recognition model. When the detection performance of the model does not meet the requirements, the model can be optimized, such as adjusting the model structure, changing the data preprocessing method, etc., and the model can be retrained to improve the detection performance of the model. Among them, the evaluation indicators for evaluating the detection performance of the model during the test may include but are not limited to precision, recall rate, training set false alarm rate and test set false alarm rate.

[0140] It should be noted that other contents of the model training process can refer to the prior art and will not be described in detail in the embodiments of the present invention.

[0141] In some embodiments, after obtaining a plurality of training sample image pairs, the standard image samples and the image samples to be tested in each training sample image pair may be subjected to image block processing to obtain a plurality of standard-to-test image sample block pairs. Then, the image sample block difference features between the standard image sample blocks and the image sample blocks to be tested in each standard-to-test image sample block pair are extracted respectively. Then, the image sample block difference features corresponding to each standard-to-test image sample block pair are used as input, and the corresponding piled material labels of the image samples to be tested are used as supervision, and the initial recognition model is supervised and trained to obtain a piled material recognition model. The specific training method is similar to the training method when the image is not divided into blocks, and the embodiments of the present invention will not be repeated here.

[0142] Exemplarily, as shown in Table 2 below, multiple images including those with and without piled materials can be first acquired to form training sample image pairs as a piled material training data set. The acquired piled material training data set can also be divided into a training set and a test set. The standard image samples and the image samples to be tested in each training sample image pair in the training set are subjected to image block processing to obtain multiple standard-to-test image sample block pairs, so as to further perform model training based on the multiple standard-to-test image sample block pairs.

[0143] Table 2. Training data set of random pile materials

[0144] Chaotic pile situation Number of training sets Number of image sample blocks in the training set Number of test sets huddle 402 2673 40 No clutter 317 12946 44

[0145] It can be understood that the embodiment of the present invention can increase the amount of training data of the model by performing block processing on the detection image samples and the standard image samples, and thus improve the performance and generalization ability of the model.

[0146] It should be noted that other contents in the embodiments of the present invention can be referred to Figure 2 and Figure 3 The relevant contents of the method provided in the illustrated embodiment will not be repeated in the embodiment of the present invention.

[0147] In summary, according to the model training method of the embodiment of the present invention, model training is performed based on the difference between the standard image and the image to be tested as a feature. The trained random pile material recognition model is based on image feature recognition rather than material type recognition, so that when the model is used to identify random pile materials, it can cope with the random pile situation of any type of materials. It can not only accurately identify any type of random pile materials, but also make the calculation simpler in the feature extraction stage. There is no need to design a complex feature extraction method to distinguish between random pile materials and backgrounds. The interference of different backgrounds can be ignored, and it has strong adaptability to complex backgrounds.

[0148] Reference Manual Attached Figure 5 , which shows the structure of a device 500 for identifying disorderly piled materials provided by one embodiment of the present invention. Figure 5 As shown, the apparatus 500 may include:

[0149] The image acquisition module 510 is used to acquire the image to be tested of the target place and the standard image corresponding to the image to be tested, where the standard image is an image when there is no random pile of materials in the target place;

[0150] A feature extraction module 520 is used to extract image difference features between the image to be tested and the standard image;

[0151] The shuffled material recognition module 530 is used to input the image difference features into a pre-trained shuffled material recognition model to obtain a shuffled material recognition result; wherein the shuffled material recognition model is obtained by supervised training using a plurality of training sample image pairs, each training sample image pair includes a standard image sample in the same place and a to-be-tested image sample with a shuffled material label;

[0152] The bulk material determination module 540 is used to determine whether there are bulk materials in the target place according to the bulk material identification result.

[0153] In some embodiments, the device 500 may further include a model training module for pre-training a shuffled material recognition model; the model training module may include:

[0154] A sample acquisition unit, used to acquire a plurality of training sample image pairs, each training sample image pair including a standard image sample and an image sample to be tested in the same place;

[0155] A feature extraction unit, used to extract image sample difference features between the standard image sample and the image sample to be tested in each training sample image pair;

[0156] A sample labeling unit is used to label the image samples to be tested in each training sample image pair respectively to obtain the image samples to be tested with a jumbled material label, where the jumbled material label is used to identify whether there is a jumbled material in the corresponding image sample to be tested;

[0157] The model training unit is used to use the image sample difference features corresponding to each training sample image pair as input and the corresponding disorderly material label as supervision to perform supervised training on the preset neural network model to obtain a disorderly material recognition model.

[0158] It should be noted that the device provided in the above embodiment, when implementing its functions, only uses the division of the above functional modules as an example. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device is divided into different functional modules to complete all or part of the functions described above. In addition, the device provided in the above embodiment and the corresponding method embodiment belong to the same concept, and the specific implementation process is detailed in the corresponding method embodiment, which will not be repeated here.

[0159] An embodiment of the present invention also provides an electronic device, which includes a processor and a memory, wherein the memory stores at least one instruction or at least one program, and the at least one instruction or the at least one program is loaded and executed by the processor to implement the method for identifying disorderly piled materials provided in the above method embodiment.

[0160] The memory can be used to store software programs and modules. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, application programs required for functions, etc.; the data storage area can store data created according to the use of the device, etc. In addition, the memory can include a high-speed random access memory and can also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other volatile solid-state storage devices. Accordingly, the memory can also include a memory controller to provide the processor with access to the memory.

[0161] In conjunction with the reference manual Figure 6 , which is a block diagram of an electronic device 600 according to an embodiment of the present invention. The electronic device 600 may include one or more processors 602, a system control logic 608 connected to at least one of the processors 602, a system memory 604 connected to the system control logic 608, a non-volatile memory (NVM) 606 connected to the system control logic 608, and a network interface 610 connected to the system control logic 608.

[0162] The processor 602 may include one or more single-core or multi-core processors. The processor 602 may include any combination of general-purpose processors and special-purpose processors (e.g., graphics processors, application processors, baseband processors, etc.). In the embodiments of the present invention, the processor 602 may be configured to execute the following operations: Figures 2 to 4 One or more of the various embodiments shown.

[0163] In some embodiments, system control logic 608 may include any suitable interface controller to provide any suitable interface to at least one of processors 602 and / or any suitable device or component in communication with system control logic 608 .

[0164] In some embodiments, the system control logic 608 may include one or more memory controllers to provide an interface to the system memory 604. The system memory 604 may be used to load and store data and / or instructions. In some embodiments, the memory 604 of the device 600 may include any suitable volatile memory, such as a suitable dynamic random access memory (DRAM).

[0165] NVM / memory 606 may include one or more tangible, non-transitory computer-readable media for storing data and / or instructions. In some embodiments, NVM / memory 606 may include any suitable non-volatile memory such as flash memory and / or any suitable non-volatile storage device, such as at least one of a HDD (Hard Disk Drive), a CD (Compact Disc) drive, and a DVD (Digital Versatile Disc) drive.

[0166] NVM / storage 606 may include a portion of storage resources installed on a device of apparatus 600, or it may be accessible to the apparatus but not necessarily a part of the apparatus. For example, NVM / storage 606 may be accessed over a network via network interface 610.

[0167] In particular, system memory 604 and NVM / storage 606 may include, respectively, a temporary copy and a permanent copy of instructions 620. Instructions 620 may include instructions that, when executed by at least one of processors 602, cause device 600 to perform the following operations: Figures 2 to 4 Instructions of the method for identifying shuffled materials shown. In some embodiments, instructions 620, hardware, firmware and / or software components thereof may be additionally / alternatively placed in system control logic 608, network interface 610 and / or processor 602.

[0168] The network interface 610 may include a transceiver for providing a radio interface for the device 600 to communicate with any other suitable device (such as a front-end module, an antenna, etc.) through one or more networks. In some embodiments, the network interface 610 may be integrated with other components of the device 600. For example, the network interface 610 may be integrated with at least one of a communication module of the processor 602, a system memory 604, an NVM / storage 606, and a firmware device (not shown) having instructions. When at least one of the processors 602 executes the instructions, the device 600 implements Figures 2 to 4 One or more of the various embodiments shown.

[0169] The network interface 610 may further include any suitable hardware and / or firmware to provide a multiple-input multiple-output radio interface. For example, the network interface 610 may be a network adapter, a wireless network adapter, a telephone modem and / or a wireless modem.

[0170] In one embodiment, at least one of the processors 602 may be packaged together with logic for one or more controllers of the system control logic 608 to form a system in package (SiP). In one embodiment, at least one of the processors 602 may be integrated on the same die with logic for one or more controllers of the system control logic 608 to form a system on chip (SoC).

[0171] The device 600 may further include an input / output (I / O) device 612. The I / O device 612 may include a user interface to enable a user to interact with the device 600; the design of the peripheral component interface enables the peripheral components to interact with the device 600. In some embodiments, the device 600 further includes a sensor for determining at least one of an environmental condition and location information related to the device 600.

[0172] In some embodiments, the user interface may include, but is not limited to, a display (e.g., an LCD display, a touch screen display, etc.), a speaker, a microphone, one or more cameras (e.g., a still image camera and / or a video camera), a flashlight (e.g., an LED flash), and a keyboard.

[0173] In some embodiments, the peripheral component interface may include, but is not limited to, a non-volatile memory port, an audio jack, and a power interface.

[0174] In some embodiments, the sensors may include, but are not limited to, gyroscope sensors, accelerometers, proximity sensors, ambient light sensors, and positioning units. The positioning unit may also be part of or interact with the network interface 610 to communicate with components of a positioning network (e.g., global positioning system (GPS) satellites).

[0175] It is to be understood that the structure illustrated in the embodiment of the present invention does not constitute a specific limitation on the electronic device 600. In other embodiments of the present invention, the electronic device 600 may include more or fewer components than shown in the figure, or combine some components, or separate some components, or arrange the components differently. The components shown in the figure may be implemented in hardware, software, or a combination of software and hardware.

[0176] An embodiment of the present invention also provides a computer-readable storage medium, which can be set in an electronic device to store at least one instruction or at least one program related to a method for identifying a pile of materials. The at least one instruction or the at least one program is loaded and executed by the processor to implement the method for identifying a pile of materials provided in the above method embodiment.

[0177] Optionally, in an embodiment of the present invention, the above-mentioned storage medium may include but is not limited to: a USB flash drive, a read-only memory (ROM), a random access memory (RAM), a mobile hard disk, a magnetic disk or an optical disk, and other media that can store program codes.

[0178] One embodiment of the present invention further provides a computer program product or a computer program, the computer program product or the computer program including computer instructions, the computer instructions being stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes the method for identifying disorderly piled materials provided in the above-mentioned various optional implementation examples.

[0179] It should be noted that the sequence of the embodiments of the present invention described above is for description only and does not represent the advantages and disadvantages of the embodiments. The above describes specific embodiments of this specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in an order different from that in the embodiments and still achieve the desired results. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0180] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0181] A person skilled in the art will understand that all or part of the steps to implement the above embodiments may be accomplished by hardware or by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, and the above-mentioned storage medium may be a read-only memory, a disk or an optical disk, etc.

[0182] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for identifying bulky materials, characterized in that: include: Acquire an image to be tested of a target location and a standard image corresponding to the image to be tested, wherein the standard image is an image when there is no randomly piled material in the target location; Extracting image difference features between the image to be tested and the standard image; Inputting the image difference features into a pre-trained shuffled material recognition model to obtain a shuffled material recognition result; wherein the shuffled material recognition model is obtained by supervised training using a plurality of training sample image pairs, each of which includes a standard image sample in the same place and a to-be-tested image sample with a shuffled material label; Determine whether there are piles of materials in the target place according to the pile of materials identification result.

2. The method according to claim 1, characterized in that The extracting of image difference features between the image to be tested and the standard image comprises: Respectively extracting histogram difference features, mean difference features and standard deviation difference features between the image to be tested and the standard image; The histogram difference feature, the mean difference feature and the standard deviation difference feature are concatenated to obtain the image difference feature.

3. The method according to claim 1, characterized in that The extracting of image difference features between the image to be tested and the standard image comprises: Performing image block processing on the image to be tested and the standard image respectively to obtain a plurality of standard-image block pairs to be tested; Image block difference features between the standard image block and the image block to be tested in each of the standard-to-test image block pairs are extracted respectively.

4. The method according to claim 3, characterized in that The step of inputting the image difference features into a pre-trained shuffled material recognition model to obtain a shuffled material recognition result includes: Inputting the image block difference features corresponding to each of the standard-to-test image block pairs into the disorderly pile material recognition model to obtain a plurality of sub-recognition results; Wherein, determining whether there are disorderly piled materials in the target place according to the disorderly piled materials identification result includes: If at least one of the plurality of sub-recognition results indicates that there is a pile of materials, determining that there is a pile of materials in the target place; Wherein, the method further comprises: The position corresponding to the standard-to-be-tested image block pair corresponding to the sub-recognition result indicating the existence of the disorderly piled materials is used as the position of the disorderly piled materials in the target place.

5. The method according to claim 1, characterized in that The method further comprises: Preprocessing the image to be tested and the standard image respectively to obtain a preprocessed image to be tested and a preprocessed standard image; wherein the preprocessing may include one or more of adaptive gamma correction, color space conversion, and demarcation of a region to be tested; The step of extracting the image difference features between the image to be tested and the standard image includes: The image difference features between the preprocessed image to be tested and the preprocessed standard image are extracted.

6. The method according to claim 1, characterized in that The disorderly pile material recognition model is a support vector machine model, or the disorderly pile material recognition model includes a plurality of support vector machine models with different kernel functions.

7. The method according to claim 1, characterized in that The method further includes pre-training the haphazard material identification model, and the training method of the haphazard material identification model includes: Acquire a plurality of training sample image pairs, each of the training sample image pairs comprising a standard image sample and an image sample to be tested in the same place; Respectively extracting image sample difference features between the standard image sample and the image sample to be tested in each of the training sample image pairs; Labeling the image samples to be tested in each of the training sample image pairs respectively to obtain the image samples to be tested with a jumbled material label, wherein the jumbled material label is used to identify whether there is a jumbled material in the corresponding image samples to be tested; The image sample difference features corresponding to each of the training sample images are used as input, and the corresponding random material labels are used as supervision to perform supervised training on the initial recognition model to obtain the random material recognition model.

8. A device for identifying piled materials, characterized in that: include: An image acquisition module, used to acquire an image to be tested of a target location and a standard image corresponding to the image to be tested, wherein the standard image is an image when there is no randomly piled material in the target location; A feature extraction module, used to extract image difference features between the image to be tested and the standard image; A disorderly pile material recognition module is used to input the image difference features into a pre-trained disorderly pile material recognition model to obtain a disorderly pile material recognition result; wherein the disorderly pile material recognition model is obtained by supervised training using a plurality of training sample image pairs, each of which includes a standard image sample in the same place and a to-be-tested image sample with a disorderly pile material label; The disorderly pile material determination module is used to determine whether disorderly pile materials exist in the target place according to the disorderly pile material identification result.

9. An electronic device, characterized in that: The electronic device includes a processor and a memory, wherein the memory stores at least one instruction or at least one program, and the at least one instruction or the at least one program is loaded and executed by the processor to implement the method for identifying disorderly piled materials as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores at least one instruction or at least one program, and the at least one instruction or at least one program is loaded and executed by a processor to implement the method for identifying disorderly piled materials as described in any one of claims 1-7.

Images