Foreground object detection method and device based on mixed Gaussian model, electronic device

Abstract

The application discloses a foreground object detection method and device based on a Gaussian mixture model, and an electronic device. Through application of the technical solution of the application, on one hand, a judgment condition of original model matching can be modified by calculating active points in an image. Thus, by performing model matching only on pixel points with a large change in pixel value in the image, consumption of meaningless model matching in a video stream image can be reduced. On the other hand, the result of twice Gaussian model matching can indirectly complete the effect of image denoising, and this way will not cause small targets to be filtered out. Thus, the problem of large foreground detection overhead in related technologies can be avoided.

Description

Technical Field

[0001] This application relates to image data processing technology, and in particular to a foreground target detection method, apparatus, and electronic device based on a Gaussian mixture model. Background Technology

[0002] Image processing technology has experienced rapid development in recent years, evolving in two distinct directions based on varying needs: semantically-based and semantically-free. Semantically-free approaches include simple background segmentation for object detection and weak object detection, primarily applied to infrared scenes. Currently, object detection algorithms are mainly used in the software domain, where ample computing resources support complex processing methods for high-resolution images.

[0003] On the other hand, image data acquisition systems are developing towards higher resolution and higher sampling rates. Thanks to advancements in sensor technology and improvements in the performance of sampling devices, it has become a reality for some advanced cameras to quickly sample and generate large amounts of image data in a short period of time.

[0004] Therefore, the current bottleneck in image processing lies in the mismatch between the high throughput of hardware systems and the high complexity of software algorithms in transmission. In other words, a complete system encompassing image acquisition, image processing, and result output often fails to achieve a balance between real-time performance, detection accuracy, and resource consumption. This is also a problem encountered when porting image detection algorithms to hardware. Software solutions to such problems often employ parallel processing methods; however, as image resolution increases, the hardware resource consumption of the model expands far beyond the capacity of current hardware development boards. Summary of the Invention

[0005] This application provides a foreground target detection method, apparatus, and electronic device based on a Gaussian mixture model. This addresses the problem of high detection overhead in image detection scenarios in related technologies.

[0006] According to one aspect of the embodiments of this application, a foreground target detection method based on a Gaussian mixture model is provided, comprising:

[0007] Obtain an active point matrix map in the first sub-block of the image to be detected where the pixel value change is greater than a preset threshold. The active point matrix map is composed of multiple pixels in the first sub-block.

[0008] Perform Gaussian mixture model matching on each pixel in the active point matrix to obtain the matching result for each pixel;

[0009] The Gaussian mixture model is updated based on the matching results.

[0010] Optionally, in another embodiment based on the method described above in this application, obtaining the active point matrix map in the first sub-block of the image to be detected, where the pixel value change is greater than a preset threshold, includes:

[0011] Obtain the current pixel value of each pixel in the first sub-block, and obtain the historical pixel value of each pixel at the previous time step.

[0012] Calculate the change in pixel value for each pixel, where the change in pixel value is obtained from the difference between the current pixel value and the historical pixel value;

[0013] The active point matrix is ​​formed by combining the pixels whose pixel value changes are greater than a preset threshold.

[0014] Optionally, in another embodiment based on the method described above in this application, after forming the active point matrix by assembling the pixels whose pixel value changes are greater than a preset threshold, the method further includes:

[0015] Obtain the first active point in the active point matrix graph, where the first active point is any active point in the active point matrix graph;

[0016] Centered on the first active point, pixels within a preset range adjacent to the first active point are designated as the first dilation active points;

[0017] The first active point and the first expanded active point are added together to the active point matrix diagram.

[0018] Optionally, in another embodiment based on the method described above in this application, the step of performing Gaussian mixture model matching on each pixel in the active point matrix to obtain the matching result corresponding to each pixel includes:

[0019] Obtain the Gaussian model parameters matched with the image to be detected, the model parameters including mean, variance and weight;

[0020] Based on whether each pixel in the first sub-block belongs to the active point matrix, determine whether to update the Gaussian model parameters corresponding to each pixel.

[0021] Based on the updated Gaussian model parameters, the matching result corresponding to each pixel is obtained.

[0022] Optionally, in another embodiment based on the method described above in this application, updating the Gaussian mixture model according to the matching result includes:

[0023] If the matching result indicates that the second pixel is a foreground target point, the second pixel is matched with the target model queue in the global Gaussian mixture model, and the second pixel is any pixel in the first sub-block.

[0024] If the first match is successful, the second pixel is determined to be the foreground target point in the first sub-block, and the Gaussian mixture model is updated.

[0025] If the first match fails, the second pixel is matched with the queue of models to be processed in the global Gaussian mixture model for a second match.

[0026] If the second match is successful, the second pixel is determined to be either a foreground target point or a noise point in the first sub-block, and the Gaussian mixture model is updated.

[0027] Optionally, in another embodiment based on the method described above, after determining the second pixel as the foreground target point in the first sub-block if the first match is successful, the method further includes:

[0028] The Gaussian mixture model corresponding to the second sub-block is updated. The second sub-block is the sub-block in the image to be detected that is adjacent to the first sub-block.

[0029] According to another aspect of the embodiments of this application, a foreground target detection device based on a Gaussian mixture model is provided, characterized in that it includes:

[0030] The acquisition module is configured to acquire an active point matrix map of a first sub-block of the image to be detected, wherein the pixel value change is greater than a preset threshold, and the active point matrix map is composed of multiple pixels in the first sub-block;

[0031] The matching module is configured to perform Gaussian mixture model matching on each pixel in the active point matrix to obtain the matching result corresponding to each pixel.

[0032] The update module is configured to update the Gaussian mixture model based on the matching results.

[0033] According to another aspect of the embodiments of this application, an electronic device is provided, comprising:

[0034] Memory, used to store executable instructions; and

[0035] A display is configured to work with the memory to execute the executable instructions to perform the operation of any of the foreground target detection methods based on Gaussian mixture models described above.

[0036] According to another aspect of the embodiments of this application, a computer-readable storage medium is provided for storing computer-readable instructions, which, when executed, perform the operations of any of the foreground target detection methods based on Gaussian mixture models described above.

[0037] In this application, an active point matrix is ​​obtained from the first sub-block of the image to be detected, where the pixel value change is greater than a preset threshold. The active point matrix consists of multiple pixels in the first sub-block. Based on the active point matrix, Gaussian mixture model matching is performed on each pixel in the first sub-block to obtain the matching result for each pixel. The Gaussian mixture model is then updated based on the matching result. By applying the technical solution of this application, on the one hand, it is possible to modify the judgment conditions of the original model matching by calculating active points in the image. This reduces the meaningless model matching cost in the video stream image by performing model matching only on pixels with large pixel value changes. On the other hand, the results of two Gaussian model matchings can be used to indirectly achieve the effect of image denoising, and this method does not filter out small targets. This avoids the problem of high foreground detection overhead in related technologies.

[0038] The technical solution of this application will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0039] The accompanying drawings, which form part of this specification, illustrate embodiments of this application and, together with the description, serve to explain the principles of this application.

[0040] This application can be more clearly understood with reference to the accompanying drawings and the following detailed description, wherein:

[0041] Figure 1 This illustration shows a schematic diagram of a foreground target detection method based on a Gaussian mixture model according to an embodiment of this application;

[0042] Figure 2 This illustration shows a schematic diagram of the process of determining the active point matrix in a foreground target detection method based on a Gaussian mixture model provided in an embodiment of this application;

[0043] Figure 3 The diagram illustrates a flowchart of a foreground target detection method based on a Gaussian mixture model according to an embodiment of this application.

[0044] Figure 4 This paper illustrates a flowchart of another foreground target detection method based on a Gaussian mixture model provided in an embodiment of this application.

[0045] Figure 5 This invention provides a schematic diagram of the structure of an electronic device according to an embodiment of the present application.

[0046] Figure 6 This illustration shows a schematic diagram of the structure of an electronic device according to an embodiment of this application;

[0047] Figure 7 A schematic diagram of a storage medium provided in one embodiment of this application is shown. Detailed Implementation

[0048] Various exemplary embodiments of the present application will now be described in detail with reference to the accompanying drawings. It should be noted that, unless otherwise specifically stated, the relative arrangement, numerical expressions, and values ​​of the components and steps set forth in these embodiments do not limit the scope of the present application.

[0049] At the same time, it should be understood that, for ease of description, the dimensions of the various parts shown in the accompanying drawings are not drawn according to actual scale.

[0050] The following description of at least one exemplary embodiment is merely illustrative and is not intended to limit the scope of this application or its application or use.

[0051] Techniques, methods, and equipment known to those skilled in the art may not be discussed in detail, but where appropriate, such techniques, methods, and equipment should be considered part of the specification.

[0052] It should be noted that similar labels and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be discussed further in subsequent figures.

[0053] Furthermore, the technical solutions of the various embodiments of this application can be combined with each other, but only if they are based on the ability of those skilled in the art to implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be considered that such combination of technical solutions does not exist and is not within the scope of protection claimed by this application.

[0054] It should be noted that all directional indications (such as up, down, left, right, front, back, etc.) in the embodiments of this application are only used to explain the relative positional relationship and movement of each component in a certain specific posture (as shown in the figure). If the specific posture changes, the directional indication will also change accordingly.

[0055] The following is combined with Figures 1-4 This application describes a method for foreground object detection based on a Gaussian mixture model, according to exemplary embodiments thereof. It should be noted that the following application scenarios are shown only to facilitate understanding of the spirit and principles of this application, and the embodiments of this application are not limited in any way. Rather, the embodiments of this application can be applied to any applicable scenario.

[0056] This application also proposes a foreground target detection method, device, and electronic device based on a Gaussian mixture model.

[0057] Figure 1 A schematic flowchart of a foreground target detection method based on a Gaussian mixture model according to an embodiment of this application is shown. Figure 1 As shown, the method includes:

[0058] S101, Obtain the active point matrix map of the first sub-block of the image to be detected, where the pixel value change is greater than a preset threshold. The active point matrix map is composed of multiple pixels in the first sub-block.

[0059] S102, based on the active point matrix diagram, perform Gaussian mixture model matching on each pixel in the first sub-block to obtain the matching result corresponding to each pixel.

[0060] S103, Update the Gaussian mixture model based on the matching results.

[0061] In related technologies, the resource consumption in processing video image data using Gaussian mixture models is mainly due to the repeated matching model. This is because, for scenarios suitable for processing video stream data using Gaussian mixture models, the background portion of pixels in adjacent frames and the internal pixels of the foreground image are mostly unchanged; pixel changes only occur at pixel boundaries.

[0062] According to the principle of Gaussian mixture model, when the pixel value remains unchanged (or changes very little), all models must be repeatedly matched until a match is successful and the model is removed from the queue. In other words, a large portion of the overhead in background points can be eliminated.

[0063] Based on the above ideas, such as Figure 2 As shown, this application can generate an active point matrix for an image by measuring the degree of pixel value change for each pixel in each sub-block (the first sub-block being any one of the sub-blocks) of the image to be detected. It is understood that this application considers only pixels located in this active point matrix to require a re-matching model step.

[0064] It should be noted that, in order to ensure the integrity of the image to be detected, this application also needs to consider that when the pixel value changes at the boundary of the target, the pixel difference will not be very obvious. Therefore, the calculated active point matrix needs to be expanded once, so that several points around each active point (the first active point is any active point) (i.e., pixels within a preset range adjacent to the first active point) participate in the model's rematching process.

[0065] As an example, the preset range can be the range of all expanded active points within an r*r square area centered on a certain active point. It can be understood that the active point and its corresponding expanded active points are added to the active point matrix diagram.

[0066] Furthermore, this application also considers the noise problem caused by image quality during the processing of video stream data. Specifically, according to the principle of Gaussian mixture model, when a pixel that does not match the existing model is detected at a certain point, a new model should be created and regarded as the target point. This results in relatively large noise in the Gaussian mixture model. This is because the detection method of Gaussian mixture model is based on a single pixel and does not use spatial domain information, making it difficult to determine whether the detected point is a target or noise.

[0067] To address the aforementioned issues, this application also proposes a process for processing spatial information during the first image segmentation, specifically including...

[0068] Firstly, a global Gaussian model belonging to the block is set within the sub-block. Since the pixel value is likely to remain unchanged when the target moves within the block, if a new pixel value is input into a pixel within the sub-block and it is found to match the target model in the existing global Gaussian model within the block, it should be assumed that the target has moved to its own position, thus skipping the tracking process and directly responding to the target.

[0069] Secondly, without considering the possibility of the target moving over a large distance, adjacent sub-blocks should notify each other when the target crosses the boundary to avoid the problem of the target being undetectable when crossing the area boundary.

[0070] In this application, an active point matrix is ​​obtained in the first sub-block of the image to be detected, where the pixel value change is greater than a preset threshold. The active point matrix is ​​composed of multiple pixels in the first sub-block. Based on the active point matrix, Gaussian mixture model matching is performed on each pixel in the first sub-block to obtain the matching result corresponding to each pixel. The Gaussian mixture model is updated according to the matching result.

[0071] By applying the technical solution of this application, on the one hand, it is possible to modify the judgment conditions of the original model matching by calculating active points in the image. This reduces meaningless model matching overhead in video stream images by performing model matching only on pixels with large pixel value changes. On the other hand, it can indirectly achieve image denoising by using the results of two Gaussian model matchings, and this method does not filter out small targets. This avoids the problem of high foreground detection overhead in related technologies.

[0072] Optionally, in another embodiment based on the method described above in this application, obtaining the active point matrix map in the first sub-block of the image to be detected, where the pixel value change is greater than a preset threshold, includes:

[0073] Obtain the current pixel value of each pixel in the first sub-block, and obtain the historical pixel value of each pixel at the previous time step.

[0074] Calculate the change in pixel value for each pixel, where the change in pixel value is obtained from the difference between the current pixel value and the historical pixel value;

[0075] The active point matrix is ​​formed by combining the pixels whose pixel value changes are greater than a preset threshold.

[0076] Optionally, in another embodiment based on the method described above in this application, after forming the active point matrix by assembling the pixels whose pixel value changes are greater than a preset threshold, the method further includes:

[0077] Obtain the first active point in the active point matrix graph, where the first active point is any active point in the active point matrix graph;

[0078] Centered on the first active point, pixels within a preset range adjacent to the first active point are designated as the first dilation active points;

[0079] The first active point and the first expanded active point are added together to the active point matrix diagram.

[0080] Optionally, in another embodiment based on the method described above in this application, the step of performing Gaussian mixture model matching on each pixel in the active point matrix to obtain the matching result corresponding to each pixel includes:

[0081] Obtain the Gaussian model parameters matched with the image to be detected, the model parameters including mean, variance and weight;

[0082] Based on whether each pixel in the first sub-block belongs to the active point matrix, determine whether to update the Gaussian model parameters corresponding to each pixel.

[0083] Based on the updated Gaussian model parameters, the matching result corresponding to each pixel is obtained.

[0084] Optionally, in another embodiment based on the method described above in this application, updating the Gaussian mixture model according to the matching result includes:

[0085] If the matching result indicates that the second pixel is a foreground target point, the second pixel is matched with the target model queue in the global Gaussian mixture model, and the second pixel is any pixel in the first sub-block.

[0086] If the first match is successful, the second pixel is determined to be the foreground target point in the first sub-block, and the Gaussian mixture model is updated.

[0087] If the first match fails, the second pixel is matched with the queue of models to be processed in the global Gaussian mixture model for a second match.

[0088] If the second match is successful, the second pixel is determined to be either a foreground target point or a noise point in the first sub-block, and the Gaussian mixture model is updated.

[0089] Optionally, in another embodiment based on the method described above, after determining the second pixel as the foreground target point in the first sub-block if the first match is successful, the method further includes:

[0090] The Gaussian mixture model corresponding to the second sub-block is updated. The second sub-block is the sub-block in the image to be detected that is adjacent to the first sub-block.

[0091] Furthermore, this application combines Figure 3 The specific steps of the plan are explained below:

[0092] Step 1: Divide the image to be detected into blocks according to the size of the sub-blocks to obtain multiple sub-block objects, including the first sub-block.

[0093] Step 2: Based on the image to be detected and the recorded historical images to be detected, obtain the active point matrix corresponding to the image to be detected.

[0094] In one approach, the method for generating the active point matrix includes: calculating the difference between the current pixel and the historical pixel and determining whether it is greater than a predetermined threshold, and generating the active point matrix before expansion.

[0095] In one approach, the historical frame update process includes: calculating the updated historical frame based on the difference between the current frame and the historical frame as the independent variable, using the following formula:

[0096] S t1 =S t0 +k*abs(S in -S t0 )

[0097] Wherein: S t1 For the next historical frame, S t0 For the current historical frame, S in The input pixel value is k, which is a preset coefficient.

[0098] Step 3: Perform Gaussian mixture model matching based on the active point matrix and input pixel values, and update the model parameters corresponding to the corresponding pixels.

[0099] Among them, such as Figure 4 As shown, this application first needs to determine the model parameters of the currently matched Gaussian model, namely the model mean, variance, weights, etc., and then replace the newly matched model with each active pixel (i.e., each pixel in the active point matrix).

[0100] One approach to determine whether a pixel is background or target includes: determining whether it belongs to background or target based on the currently matched model weights and using a method provided by a Gaussian mixture model.

[0101] In one approach, the mean and variance of the currently matched models are updated, and the weights of all models are also updated. The process includes:

[0102] The existing model is updated using the update formula of the Gaussian mixture model method. If it does not match any of the existing models, a new model is created. In one approach, if the number of models reaches the upper limit, the model is replaced according to the replacement strategy of the Gaussian mixture model method.

[0103] Step 4: For the point that was considered a foreground target in Step 3 (i.e., the matching result reflects that the second pixel is a foreground target point), it is necessary to continue matching the second pixel with the global Gaussian mixture model to which it belongs and update the global Gaussian mixture model.

[0104] It should be noted that in step 4, each sub-block maintains a global Gaussian mixture model for access by pixels within the block. The global Gaussian mixture model comprises three parts: model structure, model maintenance, and model matching method.

[0105] The overall approach to model maintenance is a variation of the queue maintenance process based on continuous frame detection. Specifically, it changes continuous frame detection from a single point to continuous frame detection within a block area. This is to enable the detection of targets with relatively fast movement speeds. The principle of continuous frame detection is that if a target is detected consecutively, it is considered a target; if it is not detected consecutively, it is considered noise or the target has disappeared.

[0106] In one approach, the global Gaussian mixture model comprises: a target model queue for storing all Gaussian models identified as targets; a model queue for storing all Gaussian models that cannot be determined as targets or background; each Gaussian model has its own counter for switching between queues; and the Gaussian models in the global model have no weights.

[0107] In one approach, the matching method for the global Gaussian mixture model includes: for a point considered a target in step 103, its input pixel value first participates in the matching of the target model queue. If a match is found, the pixel is considered a target within the sub-block. If its input pixel value does not match any model in the target model queue, it participates in the matching of the model queue to be judged. If a match is found, the pixel is considered an unjudged target or noise. If its input pixel value still does not match any model in the model queue to be judged, the model is placed into the model queue to be judged.

[0108] In one approach, the maintenance process of the global Gaussian mixture model includes updating the model weights and variances according to the Gaussian mixture model method when any model is matched.

[0109] In one approach, the maintenance process of the global Gaussian mixture model also includes: for the entire sub-block, if any point in a sub-block matches a model in the target model queue, then the target is considered to have reappeared in this frame, and the counter of the matched model is updated to the preset maximum value; if a new model is placed into the queue of models to be judged, then the counter of this model needs to be initialized.

[0110] In one approach, the maintenance process of the global Gaussian mixture model also includes: for the entire sub-block, if a model in a target model queue is not matched by any point in the sub-block, the counter of this model is decremented by one; if the counter value is reduced to below a preset threshold, it is considered that this model has not appeared for several consecutive frames, and it is placed in the queue of models to be judged, or it can be directly removed from the queue.

[0111] In one approach, the maintenance process of the global Gaussian mixture model also includes: for the entire sub-block, if a model in the queue of models to be judged is not matched by any point in the sub-block, then the counter of this model is decremented by one; if the counter value is reduced to less than 0 due to the decrease in the counter value, then it is considered that this model has not appeared for several consecutive frames and is removed from the queue.

[0112] In one approach, the maintenance process of the global Gaussian mixture model further includes: for the entire sub-block, if any point in a sub-block matches a model in the model queue to be judged, the counter of the matched model is incremented by a preset value; if the counter value exceeds a preset threshold due to the increase, the model is considered to have appeared in several consecutive frames and is placed into the target model queue.

[0113] In one approach, the maintenance process of the global Gaussian mixture model further includes: for spatially adjacent (eight directions) blocks A and B, if a model in the target model queue of block A is placed into the model queue to be judged due to the decrease of the counter value, and the pixel value of this model does not match any of the target model queues of block B, then this model is placed into the target model queue of block B, and its counter is assigned a preset maximum value.

[0114] Step 5: Integrate the results of the above steps to obtain the target detection results for the image to be detected.

[0115] In this application, an active point matrix is ​​obtained in the first sub-block of the image to be detected, where the pixel value change is greater than a preset threshold. The active point matrix is ​​composed of multiple pixels in the first sub-block. Based on the active point matrix, Gaussian mixture model matching is performed on each pixel in the first sub-block to obtain the matching result corresponding to each pixel. The Gaussian mixture model is updated according to the matching result.

[0116] By applying the technical solution of this application, on the one hand, it is possible to modify the judgment conditions of the original model matching by calculating active points in the image. This reduces meaningless model matching overhead in video stream images by performing model matching only on pixels with large pixel value changes. On the other hand, it can indirectly achieve image denoising by using the results of two Gaussian model matchings, and this method does not filter out small targets. This avoids the problem of high foreground detection overhead in related technologies.

[0117] Optionally, in another embodiment of this application, such as Figure 5 As shown, this application also provides a foreground target detection device based on a Gaussian mixture model. This includes:

[0118] The acquisition module 201 is configured to acquire an active point matrix map of a first sub-block of the image to be detected, wherein the pixel value change is greater than a preset threshold, and the active point matrix map is composed of multiple pixels in the first sub-block.

[0119] Matching module 202 is configured to perform Gaussian mixture model matching on each pixel in the active point matrix to obtain the matching result corresponding to each pixel;

[0120] The update module 203 is configured to update the Gaussian mixture model based on the matching results.

[0121] In this application, an active point matrix is ​​obtained from the first sub-block of the image to be detected, where the pixel value change is greater than a preset threshold. The active point matrix consists of multiple pixels in the first sub-block. Based on the active point matrix, Gaussian mixture model matching is performed on each pixel in the first sub-block to obtain the matching result for each pixel. The Gaussian mixture model is then updated based on the matching result. By applying the technical solution of this application, on the one hand, it is possible to modify the judgment conditions of the original model matching by calculating active points in the image. This reduces the meaningless model matching cost in the video stream image by performing model matching only on pixels with large pixel value changes. On the other hand, the results of two Gaussian model matchings can be used to indirectly achieve the effect of image denoising, and this method does not filter out small targets. This avoids the problem of high foreground detection overhead in related technologies.

[0122] In another embodiment of this application, the matching module 202 is configured to perform the following steps:

[0123] Obtain the current pixel value of each pixel in the first sub-block, and obtain the historical pixel value of each pixel at the previous time step.

[0124] Calculate the change in pixel value for each pixel, where the change in pixel value is obtained from the difference between the current pixel value and the historical pixel value;

[0125] The active point matrix is ​​formed by combining the pixels whose pixel value changes are greater than a preset threshold.

[0126] In another embodiment of this application, the matching module 202 is configured to perform the following steps:

[0127] Obtain the first active point in the active point matrix graph, where the first active point is any active point in the active point matrix graph;

[0128] Centered on the first active point, pixels within a preset range adjacent to the first active point are designated as the first dilation active points;

[0129] The first active point and the first expanded active point are added together to the active point matrix diagram.

[0130] In another embodiment of this application, the matching module 202 is configured to perform the following steps:

[0131] Obtain the Gaussian model parameters matched with the image to be detected, the model parameters including mean, variance and weight;

[0132] Based on whether each pixel in the first sub-block belongs to the active point matrix, determine whether to update the Gaussian model parameters corresponding to each pixel.

[0133] Based on the updated Gaussian model parameters, the matching result corresponding to each pixel is obtained.

[0134] In another embodiment of this application, the matching module 202 is configured to perform the following steps:

[0135] If the matching result indicates that the second pixel is a foreground target point, the second pixel is matched with the target model queue in the global Gaussian mixture model corresponding to the first sub-block, and the second pixel is any pixel in the first sub-block.

[0136] If the first match is successful, the second pixel is determined to be the foreground target point in the first sub-block, and the Gaussian mixture model is updated.

[0137] If the first match fails, the second pixel is matched with the queue of models to be processed in the global Gaussian mixture model for a second match.

[0138] If the second match is successful, the second pixel is determined to be either a foreground target point or a noise point in the first sub-block, and the Gaussian mixture model is updated.

[0139] In another embodiment of this application, the matching module 202 is configured to perform the following steps:

[0140] The Gaussian mixture model corresponding to the second sub-block is updated. The second sub-block is the sub-block in the image to be detected that is adjacent to the first sub-block.

[0141] This application also provides an electronic device for performing the above-described foreground target detection method based on a Gaussian mixture model. Please refer to... Figure 6 This illustrates a schematic diagram of an electronic device provided by some embodiments of this application. For example... Figure 6 As shown, the electronic device 3 includes: a processor 300, a memory 301, a bus 302, and a communication interface 303. The processor 300, the communication interface 303, and the memory 301 are connected via the bus 302. The memory 301 stores a computer program that can run on the processor 300. When the processor 300 runs the computer program, it executes the foreground target detection method based on the Gaussian mixture model provided in any of the foregoing embodiments of this application.

[0142] The memory 301 may include high-speed random access memory (RAM) or non-volatile memory, such as at least one disk storage device. Communication between this device network element and at least one other network element is achieved through at least one communication interface 303 (which can be wired or wireless), such as the Internet, wide area network, local area network, or metropolitan area network.

[0143] Bus 302 can be an ISA bus, PCI bus, or EISA bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. Memory 301 is used to store programs. After receiving an execution instruction, the processor 300 executes the program. The foreground target detection method based on a Gaussian mixture model disclosed in any of the foregoing embodiments of this application can be applied to the processor 300, or implemented by the processor 300.

[0144] The processor 300 may be an integrated circuit chip with signal processing capabilities. In implementation, each step of the above method can be completed by the integrated logic circuitry in the hardware of the processor 300 or by instructions in software form. The processor 300 may be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it may also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), an off-the-shelf programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this application. The general-purpose processor may be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of this application can be directly embodied in the execution of a hardware decoding processor, or executed by a combination of hardware and software modules in the decoding processor. The software modules may reside in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. The storage medium is located in memory 301. The processor 300 reads the information in memory 301 and, in conjunction with its hardware, completes the steps of the above method.

[0145] The electronic device provided in this application embodiment and the foreground target detection method based on Gaussian mixture model provided in this application embodiment are based on the same inventive concept and have the same beneficial effects as the methods they adopt, operate or implement.

[0146] This application also provides a computer-readable storage medium corresponding to the foreground target detection method based on Gaussian mixture model provided in the foregoing embodiments. Please refer to [link / reference]. Figure 7 The computer-readable storage medium shown is an optical disc 40, on which a computer program (i.e., a program product) is stored. When the computer program is run by a processor, it executes the foreground target detection method based on the Gaussian mixture model provided in any of the foregoing embodiments.

[0147] It should be noted that examples of the computer-readable storage medium may also include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other optical and magnetic storage media, which will not be elaborated here.

[0148] The computer-readable storage medium provided in the above embodiments of this application and the foreground target detection method based on Gaussian mixture model provided in the embodiments of this application are based on the same inventive concept and have the same beneficial effects as the methods adopted, run or implemented by the application stored therein.

[0149] It should be noted that:

[0150] Numerous specific details are set forth in the specification provided herein. However, it will be understood that embodiments of this application may be practiced without these specific details. In some instances, well-known structures and techniques have not been shown in detail so as not to obscure the understanding of this specification.

[0151] Similarly, it should be understood that, in order to simplify this application and aid in understanding one or more aspects of the invention, in the above description of exemplary embodiments of this application, various features of this application are sometimes grouped together in a single embodiment, figure, or description thereof. However, this disclosure should not be construed as reflecting a schematic diagram in which the claimed application requires more features than are expressly recited in each claim.

[0152] More precisely, as reflected in the following claims, the inventive aspect lies in fewer than all the features of the single embodiment disclosed above. Therefore, the claims following the detailed description are hereby expressly incorporated into that detailed description, wherein each claim itself is a separate embodiment of this application.

[0153] Furthermore, those skilled in the art will understand that although some embodiments described herein include certain features included in other embodiments but not others, combinations of features from different embodiments are meant to be within the scope of this application and form different embodiments.

[0154] For example, any of the claimed embodiments can be used in any combination as described in the following claims.

[0155] The above description is merely a preferred embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A foreground target detection method based on a Gaussian mixture model, characterized in that, include: Obtain an active point matrix map in the first sub-block of the image to be detected where the pixel value change is greater than a preset threshold. The active point matrix map is composed of multiple pixels in the first sub-block. Perform Gaussian mixture model matching on each pixel in the active point matrix to obtain the matching result for each pixel; Based on the matching results, the Gaussian mixture model is updated, including: If the matching result indicates that the second pixel is a foreground target point, the second pixel is matched with the target model queue in the global Gaussian mixture model corresponding to the first sub-block, and the second pixel is any pixel in the first sub-block. If the first match is successful, the second pixel is determined to be the foreground target point in the first sub-block, and the Gaussian mixture model is updated. If the first match fails, the second pixel is matched with the queue of models to be processed in the global Gaussian mixture model for a second match. If the second match is successful, the second pixel is determined to be a foreground target point or a noise point in the first sub-block, and the Gaussian mixture model is updated.

2. The method as described in claim 1, characterized in that, The acquisition of the active point matrix map in the first sub-block of the image to be detected, where the pixel value change is greater than a preset threshold, includes: Obtain the current pixel value of each pixel in the first sub-block, and obtain the historical pixel value of each pixel at the previous time step. Calculate the change in pixel value for each pixel, where the change in pixel value is obtained from the difference between the current pixel value and the historical pixel value; The active point matrix is ​​formed by combining the pixels whose pixel value changes are greater than a preset threshold.

3. The method as described in claim 2, characterized in that, After forming the active point matrix by assembling the pixels whose pixel value changes are greater than a preset threshold, the method further includes: Obtain the first active point in the active point matrix graph, where the first active point is any active point in the active point matrix graph; Centered on the first active point, pixels within a preset range adjacent to the first active point are designated as the first dilation active points; The first active point and the first expanded active point are added together to the active point matrix diagram.

4. The method as described in claim 1, characterized in that, The step of performing Gaussian mixture model matching on each pixel in the active point matrix to obtain the matching result for each pixel includes: Obtain the Gaussian model parameters matched with the image to be detected, the model parameters including mean, variance and weight; Based on whether each pixel in the first sub-block belongs to the active point matrix, determine whether to update the Gaussian model parameters corresponding to each pixel. Based on the updated Gaussian model parameters, the matching result corresponding to each pixel is obtained.

5. The method as described in claim 1, characterized in that, After determining that the second pixel is the foreground target point in the first sub-block if the first match is successful, the method further includes: The Gaussian mixture model corresponding to the second sub-block is updated. The second sub-block is the sub-block in the image to be detected that is adjacent to the first sub-block.

6. A foreground target detection device based on a Gaussian mixture model, characterized in that, include: The acquisition module is configured to acquire an active point matrix map of a first sub-block of the image to be detected, wherein the pixel value change is greater than a preset threshold, and the active point matrix map is composed of multiple pixels in the first sub-block; The matching module is configured to perform Gaussian mixture model matching on each pixel in the active point matrix to obtain the matching result corresponding to each pixel. The update module is configured to update the Gaussian mixture model based on the matching results, including: If the matching result indicates that the second pixel is a foreground target point, the second pixel is matched with the target model queue in the global Gaussian mixture model corresponding to the first sub-block, and the second pixel is any pixel in the first sub-block. If the first match is successful, the second pixel is determined to be the foreground target point in the first sub-block, and the Gaussian mixture model is updated. If the first match fails, the second pixel is matched with the queue of models to be processed in the global Gaussian mixture model for a second match. If the second match is successful, the second pixel is determined to be a foreground target point or a noise point in the first sub-block, and the Gaussian mixture model is updated.

7. An electronic device, characterized in that, include: Memory, used to store executable instructions; as well as, A processor for executing the executable instructions to perform the operation of the foreground target detection method based on a Gaussian mixture model as described in any of claims 1-5.

8. A computer-readable storage medium for storing computer-readable instructions, characterized in that, When the instruction is executed by the processor, it performs the operation of any of the foreground target detection methods based on Gaussian mixture models as described in claims 1-5.

Images