Small target motion detection method and device and related equipment

Abstract

The embodiment of the invention provides a small target motion detection method and device, and the method comprises the steps: obtaining continuous image data in a preset time frame; in the image data, motion information of a target object and contrast information of the image data where the target object is located are extracted, and the motion information comprises a motion track; and based on the motion information and the contrast information, determining a target motion track of the corresponding target object in the image data. According to the method, motion information and contrast information of a target object are extracted, background false positive suppression is carried out by fusing two types of visual information, tracking detection of a small target in a complex background image is realized, continuous image data can be quickly acquired and processed in a preset time frame, and the accuracy of the target object tracking detection is improved. And the processing efficiency and accuracy of small target trajectory tracking are improved through an information fusion technology.

Description

Technical Field

[0001] The present invention relates to the field of image recognition, and in particular to a small target motion detection method, device, system, electronic equipment and storage medium thereof. Background Art

[0002] With the continuous development of computer vision technology, motion target detection and tracking have been widely used in many fields, such as video surveillance, autonomous driving, security monitoring, etc. Existing target detection technologies usually rely on static information or motion information of images, and achieve target detection and tracking through traditional methods such as background modeling and optical flow method. However, these traditional methods have some limitations, especially in the detection of small targets and dynamic scenes, the target's motion trajectory is relatively complex and easily interfered by the background, resulting in reduced detection accuracy.

[0003] Therefore, existing motion detection methods have difficulty in accurately distinguishing targets from backgrounds in complex backgrounds, cannot effectively handle small targets in dynamic scenes, and lack effective fusion of motion information and contrast information, resulting in insufficient accuracy and robustness in target object detection. Summary of the invention

[0004] The embodiment of the present invention provides a small target motion detection method to solve the problems that the existing motion detection methods have difficulty in accurately distinguishing the target from the background in a complex background, cannot effectively process small targets in dynamic scenes, and lack effective fusion of motion information and contrast information, resulting in insufficient accuracy and robustness in target object detection.

[0005] In a first aspect, an embodiment of the present invention provides a small target motion detection method, the method comprising the following steps: Acquire continuous image data within a preset time frame; Extracting, from the image data, motion information of the target object and contrast information of the image data where the target object is located, wherein the motion information includes a motion trajectory; Based on the motion information and contrast information, a target motion trajectory of the corresponding target object is determined in the image data.

[0006] Optionally, acquiring continuous image data within a preset time frame includes: Determining a corresponding preset time frame based on the complexity of the pre-collected image data, wherein the complexity is determined by the resolution and color distribution range of the pre-collected image data; Within the preset time frame, the continuous image data is collected and identified by an image collection device to obtain corresponding continuous image data.

[0007] Optionally, extracting the motion information of the target object from the image data includes: Extracting motion information of at least one moving object in the image data by using a motion information extraction model to obtain motion information of at least one moving object; Determine individual data of the moving object, wherein the individual data includes the size of the moving object in the background image, the moving state, and the color difference value between the moving object and the background image; Based on the size of the moving object in the background image, the moving state, and the color difference value between the moving object and the background image, the moving information of the target object is determined in the at least one moving object.

[0008] Optionally, determining the motion information of the target object in the at least one moving object based on the size of the moving object in the background image, the motion state, and the color difference value between the moving object and the background image includes: The size of the at least one moving object in the background image, the moving state, and the color difference value between the at least one moving object and the background image are captured and collected by using the Ommatidia neuron matrix, and a first brightness signal of the moving object to be processed is obtained by performing Gaussian blur processing, wherein the moving object to be processed includes a large moving object and a small moving object; Processing the first brightness signal through an LMC neuron matrix to determine motion information of a moving object to be processed; Extracting the positive and negative signals of the motion information of the motion object to be processed through the Tm3 and Tm1 neuron matrices, and outputting them to the STMD neuron matrix, so that the STMD neuron matrix multiplies the positive and negative signals to obtain STMD neuron output data; The output data of the STMD neurons are processed through a lateral inhibition mechanism to obtain the motion information of the small motion object, and the motion information of the small motion object is used as the motion information of the target object.

[0009] Optionally, extracting contrast information of the image data where the target object is located in the image data includes: Extracting local image data from the image data using a local contrast information extraction model to obtain contrast information corresponding to at least one local image data; Based on the contrast information corresponding to the at least one local image data, the contrast information corresponding to the target area image data is determined.

[0010] Optionally, determining the contrast information corresponding to the target area image data based on the contrast information corresponding to the at least one local image data includes: Processing the at least one local image data by using an Ommatidia neuron matrix to determine a corresponding second brightness signal; The second brightness signal is sent to the TI neuron matrix through the AMC neuron matrix, so that the TI neuron matrix processes the second brightness signal to obtain contrast information corresponding to the target area image data.

[0011] Optionally, determining a target motion trajectory of a corresponding target object in the image data based on the motion information and the contrast information includes: Determining a motion trajectory of the target object within a preset time frame; Time-aligning the motion trajectory with contrast information of the target object to determine the motion trajectory and contrast information on the corresponding time series; Based on the motion trajectory and the contrast information on the corresponding time series, a target motion trajectory of the target object is determined.

[0012] In a second aspect, an embodiment of the present invention further provides a small target motion detection device, the small target motion detection device comprising: A first acquisition module, used to acquire continuous image data within a preset time frame; A first extraction module, used to extract motion information of a target object and contrast information of the image data where the target object is located in the image data, wherein the motion information includes a motion trajectory; The first determination module is used to determine a target motion trajectory of a corresponding target object in the image data based on the motion information and the contrast information.

[0013] In a third aspect, an embodiment of the present invention provides an electronic device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the small target motion detection method provided in the embodiment of the present invention when executing the computer program.

[0014] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the small target motion detection method provided in the embodiment of the invention are implemented.

[0015] In an embodiment of the present invention, continuous image data is acquired within a preset time frame; in the image data, motion information of the target object and contrast information of the image data where the target object is located are extracted, and the motion information includes a motion trajectory; based on the motion information and contrast information, the target motion trajectory of the corresponding target object is determined in the image data. The motion information and contrast information of the target object are extracted, and background false positive suppression is performed by fusing the two types of visual information to achieve tracking detection of small targets in complex background images. Through the above method, continuous image data can be quickly acquired and processed within a preset time frame, and the processing efficiency and accuracy of small target trajectory tracking can be improved through information fusion technology. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0017] Figure 1 is a flow chart of a small target motion detection method provided by an embodiment of the present invention; Figure 2 It is a schematic diagram of the overall structure of a small target motion detection model provided by an embodiment of the present invention; Figure 3 This is a schematic diagram of a small target motion trajectory acquisition process provided by an embodiment of the present invention; Figure 4 is a structural schematic diagram of a small target motion detection device provided in an embodiment of the present invention; Figure 5 It is a structural schematic diagram of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0018] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions 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.

[0019] like Figure 1 As shown, Figure 1 : is a flow chart of a small target motion detection method provided by an embodiment of the present invention, the small target motion detection method comprising the steps of: 101. Acquire continuous image data within a preset time frame.

[0020] In an embodiment of the present invention, the above-mentioned small target motion detection method can be applied to the above-mentioned small target motion detection device, which has the functions of small target motion data processing, small target motion data transmission and reception, and small target motion data memory storage, and can be constructed based on a server or a server cluster. The above-mentioned server or server cluster can be an electronic device with small target motion data processing capabilities.

[0021] The above-mentioned preset time frame can be set in motion detection according to the complexity of the image data where the target object is located or the movement speed of the target object. Generally speaking, the preset time frame can be used as a time window to monitor the movement of the target object (such as a small target) within this time window to ensure that the target movement can be captured and analyzed within the time period of this preset time frame.

[0022] It can be understood that the above-mentioned preset time frame can be updated in real time according to the size or state of the motion amplitude during the process of identifying the continuous motion of the target object. For example, when the motion amplitude of the target object increases, the preset time frame will be reduced in real time to ensure that the motion state of the target object can be captured in large-amplitude motion.

[0023] The above image data may be digital image information obtained through an image acquisition device (such as a camera, scanner, sensor, etc.), and the image information may include a series of pixel values, color information, brightness information, etc. It should be noted that the continuity of the image data may be determined according to the above preset time frame, that is, the continuous length of the acquired image data is changed in real time according to the above preset time frame.

[0024] 102. Extract, from the image data, motion information of the target object and contrast information of the image data where the target object is located.

[0025] In an embodiment of the present invention, the target object may be an object in the image data that needs to be motion tracked and detected. Generally speaking, it may be an entity or object that has a tendency to move between adjacent frames in the background of the image data. It is understandable that the target object is generally small in size and occupies fewer pixels in the image or surveillance video. For example, it may be a small aircraft in a sky picture, or a small stone for road monitoring in an unmanned vehicle.

[0026] The above-mentioned motion information may include but is not limited to motion trajectory. Generally speaking, since it is necessary to identify the motion trend of the above-mentioned target object in the background image, the position of the target object in the background image will be determined in different frames, and different positions in consecutive frames can be connected to form the motion trajectory of the target object. It should be noted that when extracting the motion information of the above-mentioned target object, the position change, speed, direction, acceleration and motion trajectory of the target object can also be extracted and integrated, so as to output a representative data motion trajectory.

[0027] In a possible embodiment, the motion information of the target object can be obtained by extracting the image data using a motion information extraction model. Figure 2 To illustrate, the above motion information extraction model can be composed of an Ommatidia neuron matrix, an LMC neuron matrix, an STMD neuron matrix, a Tm3 and a Tm1 neuron matrix, wherein the above Ommatidia neuron matrix can be used as an external brightness receiver to capture external visual brightness signals and perform Gaussian blur processing on the brightness signals, and then transmit the processed brightness signals to the LMC neuron matrix; the above LMC neuron matrix calculates the brightness change information over time for the above brightness signals, thereby realizing the extraction of the target's motion information and transmitting it to the Tm3 and Tm1 neuron matrices, so that the Tm3 and Tm1 neuron matrices respectively extract the positive signal and the negative signal after the brightness signal output by the LMC neuron matrix, and finally transmit them in parallel to the above STMD neuron matrix, and after being processed by the STMD neuron matrix, the motion information of the above target object is output.

[0028] The contrast information may be information about the difference in brightness, color or other visual features between different regions of the image data. Specifically, the contrast information may be obtained by extracting the contrast information of the image data using a local contrast information extraction model. Figure 2To illustrate, the above-mentioned local contrast information extraction model can be composed of an Ommatidia neuron matrix, an AMC neuron matrix and a T1 neuron matrix, wherein the above-mentioned Ommatidia neuron matrix can be used as an external brightness receiver to capture the external visual brightness signal and perform Gaussian blur processing on the brightness signal, and then transmit the processed brightness signal to the AMC neuron matrix; after receiving the output of the Ommatidia neuron matrix through the above-mentioned AMC neuron matrix and directly transmitting it to the T1 neuron matrix, the T1 neuron matrix directly calculates the local contrast, and transmits the calculated local contrast to the STMD neuron matrix, so that the STMD neuron matrix can receive the motion information of the above-mentioned target object and the corresponding contrast information, and after fusing them, output the target motion trajectory of the corresponding target object.

[0029] In a possible embodiment, the target object in the image data is extracted and identified by using the local contrast information extraction model and the motion information extraction model to obtain the motion information and contrast information corresponding to the target object in the image data.

[0030] The above method can effectively extract the motion information and contrast information of the target object in a low-contrast environment, and by combining the motion information and contrast information, it can more accurately identify and track the target. It is especially suitable for the detection and analysis of small targets in dynamic environments, and improves the robustness and real-time performance of the overall system.

[0031] 103. Based on the motion information and the contrast information, determine a target motion trajectory of the corresponding target object in the image data.

[0032] In an embodiment of the present invention, the target motion trajectory can be determined by fusing the motion information and contrast information, and then processing the fused data through a lateral inhibition mechanism. Specifically, a recognition threshold can be set first, and by determining the difference between the fused data at a certain time t and the recognition threshold, whether to use its position as the position in the target motion trajectory can be determined according to the difference.

[0033] Specifically, according to Figure 3 And it is explained by the following steps, where when the position (x 0 ,y 0 ) Output value of the motion detection channel at frame t If it is greater than the set recognition threshold, it is considered that at position (x 0 ,y 0 ), similarly, at the next time t+1 (i.e., the t+1 frame image), the method of the present invention can also track and record a position (x1 ,y 1 ). In particular, when the position (x 1 ,y 1 ) is at the position detected in the tth frame (x 0 ,y 0 ) (the distance between the two points is less than 5 pixels), then the position (x 1 ,y 1 ) and (x 0 ,y 0 ) belong to the same trajectory.

[0034] That is, the above target motion trajectory can be obtained by the following formula: ; in, Indicates the start time and the current time.

[0035] The above contrast information is extracted, for example, the motion trajectory corresponding to the above detected motion information is extracted by the above motion information extraction model, and the position of each motion trajectory and the contrast information on the corresponding time frame obtained by the T1 neuron matrix are searched to obtain the local contrast trajectory CRC, that is, ; Finally, the contrast trajectory variance information is used to suppress background false positives. Specifically, the following is the result: The contrast trajectory variance of the above local contrast trajectory CRC at each time t in the time frame is calculated, namely: ; By comparing the contrast variance at each moment with a preset suppression threshold, background false positives can be suppressed and small objects can be identified. For example, when in the CRC trajectory, When it is greater than or equal to the above recognition threshold, the motion trajectory where the above contrast trajectory variance is located is determined to be a node in the target motion trajectory of the above target object.

[0036] In an embodiment of the present invention, continuous image data is acquired within a preset time frame; in the image data, motion information of the target object and contrast information of the image data where the target object is located are extracted, and the motion information includes a motion trajectory; based on the motion information and contrast information, the target motion trajectory of the corresponding target object is determined in the image data. The motion information and contrast information of the target object are extracted, and background false positive suppression is performed by fusing the two types of visual information to achieve tracking detection of small targets in complex background images. Through the above method, continuous image data can be quickly acquired and processed within a preset time frame, and the processing efficiency and accuracy of small target trajectory tracking can be improved through information fusion technology.

[0037] Optionally, the step of acquiring continuous image data within a preset time frame also includes determining a corresponding preset time frame based on the complexity of the pre-captured image data; within the preset time frame, the continuous image data is captured and identified by an image capture device to obtain corresponding continuous image data.

[0038] In an embodiment of the present invention, the above-mentioned complexity can be determined by the resolution and color distribution range of the pre-captured image data. Specifically, since the more complex the image data is, the more time it takes to collect it, in order to improve the efficiency of image data collection, the complexity of the pre-captured image data can be confirmed. Generally speaking, the more complex the image data is, the higher the resolution is, and the more memory is required. Therefore, when the pre-captured image data is more complex and the collection difficulty is higher, the time span relative to the preset time frame is larger.

[0039] The pre-collected image data may be image data collected in advance before formal image collection. The purpose of the pre-collected data is to analyze the image content, evaluate the complexity of the image, and provide necessary background information for subsequent image processing, target detection and analysis.

[0040] In a possible embodiment, before performing image data recognition through an image acquisition device, the complexity of the pre-captured image data can be confirmed, and then a corresponding preset time frame can be set. Then, the pre-captured image data can be recognized and captured according to the set preset time frame to obtain corresponding continuous image data.

[0041] Optionally, in the step of extracting the motion information of the target object in the image data, the step also includes extracting the motion information of at least one moving object in the image data through a motion information extraction model to obtain the motion information of at least one moving object; determining the individual data of the moving object; and determining the motion information of the target object in at least one moving object based on the size of the moving object in the background image, the motion state, and the color difference value with the background image.

[0042] In an embodiment of the present invention, since there may be multiple entities that are constantly moving in the image data of continuous frames, motion information can be extracted from all of these multiple entities that are constantly moving, and the target object that needs to perform motion trajectory detection in this embodiment can be determined from among the multiple moving entities.

[0043] The above individual data may include but are not limited to the size of the moving object in the background image, the moving state, and the color difference value with the background image. Specifically, if the moving object occupies a larger area in the above image data and moves faster, it is considered to be the main target object, or the significance of the target object can be verified by calculating the color difference value between the moving object and the background image, that is, when the color difference value is larger, it means that the contrast between the moving object and the background is higher, and it can be identified as a target object.

[0044] The above method can effectively combine the spatial information and color information in the image to improve the detection accuracy of the target object, especially in complex backgrounds, and can more accurately identify moving targets, ensuring that the motion information of the target object can be accurately obtained under different environmental conditions.

[0045] Optionally, in the step of determining the motion information of the target object in at least one moving object based on the size of the moving object in the background image, the motion state and the color difference value between the moving object and the background image, the step also includes capturing and collecting the size of the moving object in the background image, the motion state and the color difference value between the moving object and the background image through the Ommatidia neuron matrix, and obtaining a first brightness signal of the moving object to be processed through Gaussian blur processing; processing the first brightness signal through the LMC neuron matrix to determine the motion information of the moving object to be processed; extracting the positive and negative signals of the motion information of the moving object to be processed through the Tm3 and Tm1 neuron matrices, and outputting them to the STMD neuron matrix, so that the STMD neuron matrix multiplies the positive and negative signals to obtain STMD neuron output data; processing the STMD neuron output data through the lateral inhibition mechanism to obtain the motion information of the small moving object, and using the motion information of the small moving object as the motion information of the target object.

[0046] In an embodiment of the present invention, the above-mentioned motion objects to be processed may include but are not limited to large motion objects and small motion objects. Specifically, large motion objects and small motion objects can generally be determined based on the minimum output matrix of the above-mentioned multiple neuron matrices, that is, output data smaller than the minimum output matrix is ​​considered to be a small motion object, and output data larger than the minimum output matrix is ​​considered to be a large motion object.

[0047] Specifically, the process of determining the motion information can be implemented by the following steps: Assume that the external input information is , the Ommatidia neuron matrix performs Gaussian smoothing on the above input information through a Gaussian filter, for example, After convolution, the first brightness signal is output: ,Right now ; in, , Represents the standard deviation of the Gaussian function.

[0048] Then, the LMC neuron matrix is ​​used to output the first brightness signal Processing, outputting the first brightness signal through a time domain bandpass filter The information that changes over time is calculated to obtain , it is understandable that By using the time core Convolution is performed, that is: ; in, , ,and and yes The order and time delay parameters.

[0049] Then, the positive and negative information in the output of the LMC neuron matrix is ​​extracted through the half-wave rectifiers in the Tm3 and Tm1 neuron matrices, respectively, and the output of the Tm3 neuron matrix is ​​defined as ,Right now ; The output of the Tm1 neuron matrix is ​​defined as ,Right now ; in, , They are The order and time delay parameters of .

[0050] Finally, the output of the Tm3 and Tm1 neuron matrices is multiplied by the above STMD neuron matrix to obtain the motion information of the moving object. ; Right now: ; It should be noted that since the motion information output by the STMD neuron matrix cannot filter the motion information of large moving objects, that is, it cannot accurately output the motion information of small moving objects, it is also necessary to filter the motion information of the above motion information. To process the lateral inhibition mechanism, you can do the following steps: ; in, is the motion information output of the final target object, and is a lateral inhibition kernel, which is used to filter the motion information corresponding to the large moving object in the motion information of the above moving objects. ,and are all constants. For specific constant settings, please refer to the parameter settings in Table 1 below. It can be understood that the constant setting table can set the movement speed and the size of the image data occupied by the target object in the above image data according to the above STMD neuron matrix.

[0051]

[0052] Table 1 Optionally, in the step of extracting contrast information of the image data where the target object is located in the image data, it also includes extracting local image data in the image data through a local contrast information extraction model to obtain contrast information corresponding to at least one local image data; based on the contrast information corresponding to the at least one local image data, determining the contrast information corresponding to the image data of the target area.

[0053] In an embodiment of the present invention, the above-mentioned local image data may be image data of a small area extracted from the above-mentioned image data. This area may generally contain most of the image factors in the above-mentioned image data, such as color, contrast information, etc. Specifically, the division of the area may be based on the size of the above-mentioned image data, and the image data may be evenly divided to meet the efficiency of collecting data in the image data.

[0054] The target area image data is selected from the image data or the local image data, and has area data that can contain the target object or an interested part associated with the target object, such as motion information of the target object.

[0055] In a possible embodiment, by confirming other information such as contrast or brightness in multiple local image data, since the contrast and brightness will change rapidly when the target object moves on the background, the corresponding target area image data can be determined by determining the corresponding brightness and contrast changes. Finally, the contrast information of the target area image data is extracted through a local contrast information extraction model to obtain the contrast information corresponding to the target area image data.

[0056] Optionally, in the step of determining the contrast information corresponding to the target area image data based on the contrast information corresponding to at least one local image data, it also includes processing at least one local image data through the Ommatidia neuron matrix to determine the corresponding second brightness signal; sending the second brightness signal to the TI neuron matrix through the AMC neuron matrix, so that the TI neuron matrix processes the second brightness signal to obtain the contrast information corresponding to the target area image data.

[0057] In the embodiment of the present invention, the step of obtaining the contrast information corresponding to the target area image data may be implemented by the following steps: The Ommatidia neuron matrix receives external stimuli and performs smoothing to obtain a second brightness signal and output it to the AMC neuron matrix, which is then directly transmitted to the T1 neuron matrix, so that the T1 neuron matrix can directly calculate the local contrast corresponding to its second brightness signal, and transmit the calculated local contrast to the STMD neuron matrix for processing.

[0058] Specifically, after the AMC neuron matrix receives the output of the Ommatidia neuron matrix, it extracts the sum of the local area information centered on each pixel and the sum of the surrounding area information after removing the central local area. Suppose the AMC neuron matrix receives the second brightness signal of the central local area centered on each pixel as ,Right now ; in, Indicates a The local area as the center, similarly, let the sum of the information of the surrounding area after removing the central local area received by the AMC neuron matrix be ,Right now ; in, Indicates a The local area centered on , are all constants.

[0059] The output from the AMC neuron matrix is ​​received through the T1 neuron matrix, and the local contrast of all pixels is calculated: ; in, Indicates area The number of pixels, Indicates area The number of pixels in ,It should be explained that k and z can be determined based on the size preference of the ,STMD neuron matrix for the target object as 5x5.

[0060] Optionally, in the step of determining the target motion trajectory of the corresponding target object in the image data based on the motion information and the contrast information, it also includes determining the motion trajectory of the target object within a preset time frame; time-aligning the motion trajectory with the contrast information of the target object to determine the motion trajectory and the contrast information on the corresponding time series; and determining the target motion trajectory of the target object based on the motion trajectory and the contrast information on the corresponding time series.

[0061] In the implementation of the present invention, the above-mentioned time alignment process can synchronize information at different time points or in different data sources to the same time base to ensure that during the analysis or processing, the data can accurately correspond to the same time series. In multi-source data processing, time alignment ensures that data from different sensors or data channels (such as target motion trajectory and contrast information) can be compared or merged at the same time to ensure the accuracy and consistency of the analysis.

[0062] In a possible embodiment, the extracted motion trajectory is time-aligned with the contrast information of the target object in the corresponding time series. For example, the contrast information of the target area in each time frame is paired with the corresponding motion trajectory data to ensure that the motion trajectory and contrast information of the target object match at the same moment. Based on the time-aligned motion trajectory and contrast information, the target motion trajectory of the target object is determined.

[0063] By temporally aligning and fusing the motion trajectory of the target object with the contrast information, the motion trajectory of the target object can be accurately tracked in complex backgrounds and dynamic environments. By combining motion information and contrast information, background interference is effectively reduced, the detection accuracy of small targets is improved, and the stability and consistency of the motion trajectory are ensured.

[0064] like Figure 4 As shown, the embodiment of the present invention further provides a small target motion detection device 400, which includes: The first acquisition module 401 is used to acquire continuous image data within a preset time frame; A first extraction module 402 is used to extract the motion information of the target object and the contrast information of the image data where the target object is located in the image data, wherein the motion information includes a motion trajectory; The first determination module 404 is configured to determine a target motion trajectory of a corresponding target object in the image data based on the motion information and the contrast information.

[0065] Optionally, the first acquisition module 401 includes: A first acquisition submodule, configured to determine a corresponding preset time frame based on the complexity of the pre-acquired image data, wherein the complexity is determined by the resolution and color distribution range of the pre-acquired image data; The second acquisition submodule is used to acquire and identify the continuous image data through an image acquisition device within the preset time frame to obtain corresponding continuous image data.

[0066] Optionally, the first extraction module 402 includes: A first extraction submodule, configured to extract motion information of at least one moving object in the image data by using a motion information extraction model to obtain motion information of at least one moving object; A first determination submodule is used to determine individual data of the moving object, wherein the individual data includes the size of the moving object in the background image, the moving state, and the color difference value between the moving object and the background image; The second determination submodule is used to determine the motion information of the target object in the at least one moving object based on the size of the moving object in the background image, the motion state, and the color difference value between the moving object and the background image.

[0067] Optionally, the second determining submodule includes: A first acquisition unit is used to capture and collect the size of the at least one moving object in the background image, the moving state, and the color difference value between the at least one moving object and the background image through the Ommatidia neuron matrix, and obtain a first brightness signal of the moving object to be processed through Gaussian blur processing, wherein the moving object to be processed includes a large moving object and a small moving object; A first determining unit, configured to process the first brightness signal through an LMC neuron matrix to determine motion information of a moving object to be processed; A first extraction unit is used to extract positive and negative signals of the motion information of the motion object to be processed through the Tm4 and Tm1 neuron matrices, and output them to the STMD neuron matrix, so that the STMD neuron matrix multiplies the positive and negative signals to obtain STMD neuron output data; The first acquisition unit is used to process the output data of the STMD neuron through a lateral inhibition mechanism to obtain the motion information of the small motion object, and use the motion information of the small motion object as the motion information of the target object.

[0068] Optionally, the first extraction module 402 includes: A second extraction submodule, configured to extract local image data from the image data by using a local contrast information extraction model, to obtain contrast information corresponding to at least one local image data; The third determination submodule is used to determine the contrast information corresponding to the target area image data based on the contrast information corresponding to the at least one local image data.

[0069] Optionally, the third determining submodule includes: A second determination unit is used to process the at least one local image data through an Ommatidia neuron matrix to determine a corresponding second brightness signal; The second acquisition unit is used to send the second brightness signal to the TI neuron matrix through the AMC neuron matrix, so that the TI neuron matrix processes the second brightness signal to obtain contrast information corresponding to the target area image data.

[0070] Optionally, the first determining module 404 includes: A fourth determination submodule is used to determine the motion trajectory of the target object within a preset time frame; a fifth determination submodule, configured to perform time alignment between the motion trajectory and the contrast information of the target object, and determine the motion trajectory and the contrast information on the corresponding time series; A sixth determination submodule is used to determine the target motion trajectory of the target object based on the motion trajectory and the contrast information on the corresponding time series. like Figure 5 As shown, an embodiment of the present invention further provides an electronic device 500, including a processor, and the processor can execute any of the above-mentioned small target motion detection methods.

[0071] Specifically, it includes a processor 501 and a memory 502, and a computer program for executing a small target motion detection method stored in the memory 502 and capable of running on the processor 501, wherein: The processor 501 runs the computer program of the small target motion detection method stored in the memory 502 to perform the following steps: Acquire continuous image data within a preset time frame; Extracting, from the image data, motion information of the target object and contrast information of the image data where the target object is located, wherein the motion information includes a motion trajectory; Based on the motion information and contrast information, a target motion trajectory of the corresponding target object is determined in the image data.

[0072] Optionally, the processor 501 executes the acquisition of continuous image data within a preset time frame, including: Determining a corresponding preset time frame based on the complexity of the pre-collected image data, wherein the complexity is determined by the resolution and color distribution range of the pre-collected image data; Within the preset time frame, the continuous image data is collected and identified by an image collection device to obtain corresponding continuous image data.

[0073] Optionally, the processor 501 executes the step of extracting the motion information of the target object from the image data, including: Extracting motion information of at least one moving object in the image data by using a motion information extraction model to obtain motion information of at least one moving object; Determine individual data of the moving object, wherein the individual data includes the size of the moving object in the background image, the moving state, and the color difference value between the moving object and the background image; Based on the size of the moving object in the background image, the moving state, and the color difference value between the moving object and the background image, the moving information of the target object is determined in the at least one moving object.

[0074] Optionally, the processor 501 performs the step of determining the motion information of the target object in the at least one moving object based on the size of the moving object in the background image, the motion state, and the color difference value between the moving object and the background image, including: The size of the at least one moving object in the background image, the moving state, and the color difference value between the at least one moving object and the background image are captured and collected by using the Ommatidia neuron matrix, and a first brightness signal of the moving object to be processed is obtained by performing Gaussian blur processing, wherein the moving object to be processed includes a large moving object and a small moving object; Processing the first brightness signal through an LMC neuron matrix to determine motion information of a moving object to be processed; Extracting the positive and negative signals of the motion information of the motion object to be processed through the Tm3 and Tm1 neuron matrices, and outputting them to the STMD neuron matrix, so that the STMD neuron matrix multiplies the positive and negative signals to obtain STMD neuron output data; The output data of the STMD neurons are processed through a lateral inhibition mechanism to obtain the motion information of the small motion object, and the motion information of the small motion object is used as the motion information of the target object.

[0075] Optionally, the processor 501 executes the step of extracting contrast information of the image data where the target object is located in the image data, including: Extracting local image data from the image data using a local contrast information extraction model to obtain contrast information corresponding to at least one local image data; Based on the contrast information corresponding to the at least one local image data, the contrast information corresponding to the target area image data is determined.

[0076] Optionally, the processor 501 further performs the step of determining the contrast information corresponding to the target area image data based on the contrast information corresponding to the at least one local image data, including: Processing the at least one local image data by using an Ommatidia neuron matrix to determine a corresponding second brightness signal; The second brightness signal is sent to the TI neuron matrix through the AMC neuron matrix, so that the TI neuron matrix processes the second brightness signal to obtain contrast information corresponding to the target area image data.

[0077] Optionally, the processor 501 further performs the step of determining a target motion trajectory of a corresponding target object in the image data based on the motion information and the contrast information, including: determining a motion trajectory of the target object within a preset time frame; Time-aligning the motion trajectory with contrast information of the target object to determine the motion trajectory and contrast information on the corresponding time series; Based on the motion trajectory and the contrast information on the corresponding time series, a target motion trajectory of the target object is determined.

[0078] An embodiment of the present invention also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the various processes of the small target motion detection method or the application-side small target motion detection method provided by the embodiment of the present invention are implemented, and the same technical effect can be achieved. To avoid repetition, it will not be repeated here.

[0079] Those skilled in the art can understand that the implementation of all or part of the processes in the above-mentioned embodiments is a program that can be completed by instructing related hardware through a computer program, and can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. The storage medium can be a disk, an optical disk, a read-only memory (ROM) or a random access memory (RAM).

[0080] The above disclosure is only the preferred embodiment of the present invention, which certainly cannot be used to limit the scope of the present invention. Therefore, equivalent changes made according to the claims of the present invention are still within the scope of the present invention.

Claims

1. A small target motion detection method, characterized in that: include: Acquire continuous image data within a preset time frame; Extracting, from the image data, motion information of the target object and contrast information of the image data where the target object is located, wherein the motion information includes a motion trajectory; Based on the motion information and contrast information, a target motion trajectory of the corresponding target object is determined in the image data.

2. The small target motion detection method according to claim 1, characterized in that: The step of acquiring continuous image data within a preset time frame includes: Determining a corresponding preset time frame based on the complexity of the pre-collected image data, wherein the complexity is determined by the resolution and color distribution range of the pre-collected image data; Within the preset time frame, the continuous image data is collected and identified by an image collection device to obtain corresponding continuous image data.

3. The small target motion detection method according to claim 1, characterized in that: Extracting the motion information of the target object from the image data includes: Extracting motion information of at least one moving object in the image data by using a motion information extraction model to obtain motion information of at least one moving object; Determine individual data of the moving object, wherein the individual data includes the size of the moving object in the background image, the moving state, and the color difference value between the moving object and the background image; Based on the size of the moving object in the background image, the moving state, and the color difference value between the moving object and the background image, the moving information of the target object is determined in the at least one moving object.

4. The small target motion detection method according to claim 3, characterized in that: The step of determining the motion information of the target object in the at least one moving object based on the size of the moving object in the background image, the motion state, and the color difference value between the moving object and the background image comprises: The size of the at least one moving object in the background image, the moving state, and the color difference value between the at least one moving object and the background image are captured and collected by using the Ommatidia neuron matrix, and a first brightness signal of the moving object to be processed is obtained by performing Gaussian blur processing, wherein the moving object to be processed includes a large moving object and a small moving object; Processing the first brightness signal through an LMC neuron matrix to determine motion information of a moving object to be processed; Extracting the positive and negative signals of the motion information of the motion object to be processed through the Tm3 and Tm1 neuron matrices, and outputting them to the STMD neuron matrix, so that the STMD neuron matrix multiplies the positive and negative signals to obtain STMD neuron output data; The output data of the STMD neurons are processed through a lateral inhibition mechanism to obtain the motion information of the small motion object, and the motion information of the small motion object is used as the motion information of the target object.

5. The small target motion detection method according to claim 1, characterized in that: Extracting contrast information of the image data where the target object is located in the image data includes: Extracting local image data from the image data using a local contrast information extraction model to obtain contrast information corresponding to at least one local image data; Based on the contrast information corresponding to the at least one local image data, the contrast information corresponding to the target area image data is determined.

6. The small target motion detection method as claimed in claim 5, characterized in that: The determining, based on the contrast information corresponding to the at least one local image data, the contrast information corresponding to the target area image data comprises: Processing the at least one local image data by using an Ommatidia neuron matrix to determine a corresponding second brightness signal; The second brightness signal is sent to the TI neuron matrix through the AMC neuron matrix, so that the TI neuron matrix processes the second brightness signal to obtain contrast information corresponding to the target area image data.

7. The small target motion detection method according to claim 1, characterized in that: The step of determining a target motion trajectory of a corresponding target object in the image data based on the motion information and the contrast information includes: Determining a motion trajectory of the target object within a preset time frame; Time-aligning the motion trajectory with contrast information of the target object to determine the motion trajectory and contrast information on the corresponding time series; Based on the motion trajectory and the contrast information on the corresponding time series, a target motion trajectory of the target object is determined.

8. A small target motion detection device, characterized in that: include: A first acquisition module, used to acquire continuous image data within a preset time frame; A first extraction module, used to extract motion information of a target object and contrast information of the image data where the target object is located in the image data, wherein the motion information includes a motion trajectory; The first determination module is used to determine a target motion trajectory of a corresponding target object in the image data based on the motion information and the contrast information.

9. An electronic device, characterized in that: include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps in the small target motion detection method as described in any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps in the small target motion detection method according to any one of claims 1 to 7 are implemented.

Images