A target detection optimization method and system based on Center Point model

Abstract

The present invention relates to the field of artificial intelligence technology, and discloses a target detection optimization method and system based on a CenterPoint model. The method comprises obtaining an original image of a current frame; inputting the original image into a CenterPoint model for inference to obtain target category, center offset position, and target size information; inputting the target category, center offset position, target size information, and a historical target center position set stored in the previous frame into a pre-trained target center point motion model to obtain a target center position set; rendering the target category, target center position set, and target size information into the original image to obtain a rendered image; performing target detection based on the rendered image, and tracking the target based on the target detection result. The method has the following effects: achieving accurate detection of fast-moving or obscured targets to improve the accuracy of target tracking.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a target detection optimization method and system based on a Center Point model. Background Art

[0002] With the continuous advancement of technology, autonomous driving technology is gradually moving from science fiction to reality. Autonomous vehicles rely on a complex array of systems and technologies to perceive their surroundings, make decisions, and navigate safely. Among these systems, object detection and tracking are key components that enable vehicles to understand and respond to dynamic traffic conditions. In the context of autonomous driving, vehicles must be able to identify various road users, such as pedestrians, cyclists, and other motor vehicles, and predict their behavior. To achieve this, vehicles are equipped with a variety of sensors, including cameras, LiDAR, radar, and ultrasonic sensors, to collect vast amounts of environmental data. However, processing this data is challenging, especially in complex urban environments, which are filled with fast-moving objects, unexpected pedestrians, and frequent occlusions. Therefore, improving object detection and tracking capabilities, especially in rapidly changing and complex conditions, has become a key research direction driving the advancement of autonomous driving technology. Researchers are committed to developing smarter and more efficient algorithms to ensure the stable operation of autonomous vehicles in a variety of conditions, improving passenger safety and comfort, and ensuring transportation efficiency.

[0003] In one existing technology, a real-time detection method and system for multiple targets in a vehicle-road scene are disclosed. The method comprises the following steps: reading a vehicle-road scene image frame, detecting targets in the current image frame using a target detection model; mapping the targets to the original image frame based on the target detection results; tracking targets marked in the current image frame using a target tracking model, visually predicting the targets based on the target tracking results, adding the predictions to a prediction result set, and rechecking the prediction result set as a data set; obtaining a union of multiple target detection results and multiple target tracking prediction results; processing the union of the multiple target detection results and the multiple target tracking prediction results to obtain a final multiple target detection and tracking result, and outputting the final multiple target detection frame to the vehicle-road scene image captured by the camera through a camera projection matrix for display, thereby achieving the purpose of accurate detection.

[0004] The existing technology cannot accurately detect fast-moving or obscured targets, resulting in tracking interruption or incorrect association. Summary of the Invention

[0005] The present invention provides a target detection optimization method and system based on the Center Point model, which can achieve accurate detection of fast-moving or obscured targets, thereby improving the accuracy of target tracking.

[0006] In a first aspect, in order to solve the above technical problems, the present invention provides a target detection optimization method based on a Center Point model, comprising:

[0007] Get the original picture of the current frame;

[0008] Input the original image into the CenterPoint model for inference to obtain the target category, center offset position and target size information;

[0009] Inputting the target category, the center offset position, the target size information, and a historical target center position set stored in a previous frame into a pre-trained target center point motion model to obtain a target center position set;

[0010] Rendering the target category, the target center position set, and the target size information into the original image to obtain a rendered image;

[0011] Target detection is performed based on the rendered image, and target tracking is performed based on the target detection result.

[0012] In an optional embodiment, inputting the original image into the CenterPoint model for inference to obtain target category, center offset position, and target size information includes:

[0013] Input the original image into the ResNet backbone network of the optimized CenterPoint model to extract image features and obtain a feature map;

[0014] The feature pyramid network is used to fuse the scene features of the feature map to obtain a fused feature map.

[0015] The fused feature map is input into the CenterBox Head for position prediction to obtain the center offset position and target size information;

[0016] Input the fused feature map into the RetinaNet network for regression to obtain the target prediction frame;

[0017] The center offset position and the target prediction box are input into the KeyPoint Head for type recognition to obtain the target category.

[0018] In an optional embodiment, the optimization process of the CenterPoint model includes:

[0019] Input the original image into the CNN backbone network of the initial CenterPoint model to extract image features and obtain the initial feature map;

[0020] Input the feature map into the CenterBox Head for position prediction to obtain the initial center offset position;

[0021] Inputting the initial center offset position into the KeyPoint Head to perform target type recognition to obtain an initial target category;

[0022] Inputting the initial feature map into the MaskNetHead model to obtain a pixel-level segmentation map, wherein the pixel-level segmentation map includes pixel information of the target category and the category region area;

[0023] performing binarization processing on the pixel-level segmentation map to obtain a binarized image;

[0024] Performing target judgment based on the binarized image, the initial target category, the category area, and the initial center offset, and counting the initial target category, the category area, and the initial center offset into a target list according to the judgment result;

[0025] The target truth table in the original image is compared with the target list and the positioning loss, classification loss and segmentation loss are calculated. The initial CenterPoint model is iteratively optimized through backpropagation according to the calculation results to obtain the CenterPoint model.

[0026] In an optional embodiment, the target category, the center offset position, the target size information, and a historical target center position set stored in a previous frame are input into a pre-trained target center point motion model to obtain the target center position set, including:

[0027] Use Kalman filter as the basic model to establish the initial motion prediction model;

[0028] Dynamically adjusting the process noise covariance matrix and the measurement noise covariance matrix according to the pre-stored target velocity, pre-stored target acceleration and pre-stored motion position of the previous frame to obtain a motion prediction model;

[0029] Input the target category, the center offset position, the target size information, and the historical target center position set stored in the previous frame into the motion prediction model to perform Kalman filter prediction to generate a preliminary predicted position sequence;

[0030] Optimizing the preliminary predicted position sequence and the pre-stored actual position sequence by a dynamic time warping algorithm to obtain a predicted position sequence;

[0031] Mapping adjustment is performed according to the predicted position sequence to obtain a target center position set.

[0032] In an optional embodiment, optimizing the preliminary predicted position sequence by a dynamic time warping algorithm based on the preliminary predicted position sequence and a pre-stored actual position sequence to obtain a predicted position sequence includes:

[0033] Calculating corresponding Euclidean distances based on the preliminary predicted position sequence and the pre-stored actual position sequence to obtain a distance matrix;

[0034] Creating a cumulative distance matrix based on the distance matrix, and filling the cumulative distance matrix through dynamic programming to obtain an optimal path;

[0035] Backtracking the optimal path by comparing the cumulative distances of adjacent cells of the cumulative distance matrix to form a matching pair list;

[0036] The preliminary predicted position sequence is adjusted according to the matching pair list by an interpolation method to obtain a predicted position sequence.

[0037] In an optional implementation, rendering the object category, the object center position set, and the object size information into the original image to obtain a rendered image includes:

[0038] Calculating the frame corner coordinates based on the target center position set and the target size information;

[0039] matching target color and target shape according to the target category;

[0040] Draw a target logo on the original image using a drawing function in a pre-stored image processing library according to the frame corner coordinates, the target color, and the target shape to obtain a rendered image;

[0041] The frame corner coordinates are calculated using the following formula: ;

[0042] in, Indicates the coordinates of the upper left corner of the border. Indicates the coordinates of the upper right corner of the border. Indicates the coordinates of the lower left corner of the border. Indicates the coordinates of the lower right corner of the border. represents the center point coordinates, Indicates the target width in the target size information. Indicates the target height in the target size information.

[0043] In an optional embodiment, the rendered image is compared with the rendered image of the previous frame to determine whether the target of the current frame is the same as the target of the previous frame. If the target of the current frame is the same as the target of the previous frame, the pre-stored target motion model is not updated and target detection is performed for the next frame.

[0044] If the target of the current frame is different from the target of the previous frame, determine whether the pre-stored tracker list contains the target motion model of the current frame. If the pre-stored tracker list does not contain the target motion model of the current frame, the corresponding tracker is invalid and the invalid tracker is deleted.

[0045] If the pre-stored tracker list contains the target motion model of the current frame, the tracker list is updated and the predicted target position is generated based on the rendered image.

[0046] In a second aspect, the present invention provides a target detection optimization system based on a Center Point model, comprising:

[0047] Data acquisition module, used to obtain the original image of the current frame;

[0048] The target information inference module is used to input the original image into the CenterPoint model for inference to obtain target category, center offset position and target size information;

[0049] A center position generation module is used to input the target category, the center offset position, the target size information, and the historical target center position set stored in the previous frame into a pre-trained target center point motion model to obtain a target center position set;

[0050] An image rendering module renders the target category, the target center position set, and the target size information into the original image to obtain a rendered image;

[0051] The target tracking module is used to perform target detection based on the rendered image and to track the target based on the target detection result.

[0052] In a third aspect, the present invention also provides an electronic device comprising a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, it implements any one of the above-mentioned target detection optimization methods based on the Center Point model.

[0053] In a fourth aspect, the present invention also provides a computer-readable storage medium, which includes a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute any one of the above-mentioned target detection optimization methods based on the Center Point model.

[0054] Compared with the prior art, the present invention has the following beneficial effects:

[0055] The present invention discloses a target detection optimization method based on a CenterPoint model, comprising obtaining an original image of a current frame; inputting the original image into a CenterPoint model for inference to obtain target category, center offset position, and target size information; inputting the target category, center offset position, target size information, and a previously stored historical target center position set into a pre-trained target center point motion model to obtain a target center position set; rendering the target category, target center position set, and target size information into the original image to obtain a rendered image; performing target detection based on the rendered image, and tracking the target based on the target detection result. The present invention analyzes the image of the current frame using the CenterPoint model to obtain the target category, position offset, and size information. Then, combined with the target position data of the previous frame, the target position in the current frame is predicted and updated. This information is annotated on the original image to form a rendered image with the detection result, thereby realizing target recognition and tracking, thereby achieving accurate detection of fast-moving or obscured targets, and improving the accuracy of target tracking. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 1 is a flow chart of a target detection optimization method based on a Center Point model provided by the first embodiment of the present invention;

[0057] Figure 2 2 is a schematic diagram of a target detection optimization system based on a Center Point model provided in a second embodiment of the present invention. DETAILED DESCRIPTION

[0058] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0059] Reference Figure 1The first embodiment of the present invention provides a target detection optimization method based on a Center Point model, comprising the following steps:

[0060] S11, obtain the original image of the current frame;

[0061] S12, inputting the original image into the CenterPoint model for inference to obtain target category, center offset position, and target size information;

[0062] S13, inputting the target category, the center offset position, the target size information, and a historical target center position set stored in a previous frame into a pre-trained target center point motion model to obtain a target center position set;

[0063] S14, rendering the target category, the target center position set, and the target size information into the original image to obtain a rendered image;

[0064] S15, performing target detection according to the rendered image, and performing target tracking according to the target detection result.

[0065] In step S11, the original picture of the current frame is obtained.

[0066] Specifically, acquiring the raw image of the current frame involves receiving real-time data from the vehicle's sensors, including cameras, LiDAR, and radar, which capture image information of the vehicle's surroundings. The raw image is unprocessed image data that directly reflects the real-time state of the autonomous vehicle's environment, providing the necessary visual input for subsequent object detection and tracking. This step ensures that the system can analyze and make decisions based on the latest environmental information, which is crucial for achieving accurate object detection.

[0067] In step S12, the original image is input into the CenterPoint model for inference to obtain target category, center offset position and target size information.

[0068] In a specific embodiment, inputting the original image into the CenterPoint model for inference to obtain target category, center offset position, and target size information includes:

[0069] Input the original image into the ResNet backbone network of the optimized CenterPoint model to extract image features and obtain a feature map;

[0070] The feature pyramid network is used to fuse the scene features of the feature map to obtain a fused feature map.

[0071] The fused feature map is input into the CenterBox Head for position prediction to obtain the center offset position and target size information;

[0072] Input the fused feature map into the RetinaNet network for regression to obtain the target prediction frame;

[0073] The center offset position and the target prediction box are input into the KeyPoint Head for type recognition to obtain the target category.

[0074] In a specific embodiment, the optimization process of the CenterPoint model includes:

[0075] Input the original image into the CNN backbone network of the initial CenterPoint model to extract image features and obtain the initial feature map;

[0076] Input the feature map into the CenterBox Head for position prediction to obtain the initial center offset position;

[0077] Inputting the initial center offset position into the KeyPoint Head to perform target type recognition to obtain an initial target category;

[0078] Inputting the initial feature map into the MaskNetHead model to obtain a pixel-level segmentation map, wherein the pixel-level segmentation map includes pixel information of the target category and the category region area;

[0079] performing binarization processing on the pixel-level segmentation map to obtain a binarized image;

[0080] Performing target judgment based on the binarized image, the initial target category, the category area, and the initial center offset, and counting the initial target category, the category area, and the initial center offset into a target list according to the judgment result;

[0081] The target truth table in the original image is compared with the target list and the positioning loss, classification loss and segmentation loss are calculated. The initial CenterPoint model is iteratively optimized through backpropagation according to the calculation results to obtain the CenterPoint model.

[0082] Specifically, the original image is first fed into the ResNet backbone network to extract image features and generate feature maps. As a deep residual learning framework, ResNet addresses the vanishing gradient problem in deep networks by introducing skip connections, making training more stable and efficient. During this process, the convolutional layer applies a series of filters to the input original image. Each filter is responsible for capturing specific features, such as edges, textures, or shapes, thereby constructing a multi-level abstract representation.

[0083] The convolutional layer applies a series of filters to the input raw image, a core step in feature extraction in convolutional neural networks (CNNs). Each filter is a small, learnable two-dimensional matrix, typically small in size (such as 3x3 or 5x5) and the same depth as the input data. When applied to color images, this means the filter slides not only across the width and height but also across the three RGB color channels.

[0084] Specifically, suppose there is a The original image, where Represents the width, represents the height, and represents the depth, for example, for a color image, , corresponding to the red, green, and blue channels). There is also a The filter here represents the spatial dimensions of the filter (i.e. width and height), and the depth of the filter Must match the depth of the input image. The convolution operation performs a dot product between this filter and the local region of the input image and sums the results to produce a single output value. This process can be formally expressed as:

[0085] ;

[0086] in, represents the original image, represents the filter (also called convolution kernel), is the coordinate position on the output feature map, are relative coordinates within the filter window, represents the depth index of the input image and filter, is the preset bias term.

[0087] The above formula describes how a single filter is applied to a local region of the input image to compute an element in the output feature map. To obtain the complete feature map, the filter slides across the entire input image with a fixed step size, repeating the above calculations. Furthermore, using zero padding allows convolution to be performed without changing the output size.

[0088] Considering that there are multiple such filters in the convolution layer, each filter can capture different features (such as edges, textures, etc.), then the result of all these filters working together is a multi-channel feature map. filters, the final output feature map will be of size A three-dimensional tensor of is the padding size, is the preset step size.

[0089] For example, if there is a The original image is sized The filter has a step size of 1 and no additional padding, then according to the above formula, the size of the output feature map will be This is because each filter produces a There are six layers of activation maps of size , and since there are six different filters, there are six layers of such activation maps in total.

[0090] Next, the resulting feature maps are fused with scene features using a Feature Pyramid Network (FPN) to generate a fused feature map. The FPN enhances feature representation at different scales through a top-down pathway and lateral connections. Specifically, the FPN starts with the top-most features and gradually transfers information downward, combining it with high-resolution features from lower layers. This effectively combines global context and local details, thereby improving detection performance.

[0091] Subsequently, the fused feature maps are input into two different head modules respectively: one is CenterBox Head for position prediction, and the other is RetinaNet for regressing the target prediction box. The task of CenterBox Head is to estimate the offset of the center of the object relative to the center of the grid cell and the size of the object, that is, the width and height. This process involves key point detection, where each category has a heat map representing the potential target center position. For each positive sample (that is, the true object center), its position deviation relative to the center of the grid cell is calculated, and the actual size of the object is regressed. RetinaNet focuses on the precise positioning of the bounding box, predicts the candidate area through the anchor mechanism, and classifies and regresses it.

[0092] To determine the object category, the fused feature map is input into the KeyPoint Head. The KeyPoint Head performs type recognition based on the previously obtained center offset position and the predicted object bounding box. The key here is to locate the key points of the object, such as the four corners of a vehicle or the joints of a pedestrian. Using this keypoint information, the system can accurately determine the specific object category.

[0093] The optimization process for the CenterPoint model begins by feeding the original image into the CNN backbone of the initial CenterPoint model to extract the initial feature maps. These feature maps then enter the CenterBox Head for preliminary position prediction, generating the initial center offset position. Simultaneously, the initial feature maps are fed into the MaskNetHead model to obtain pixel-level segmentation maps. These pixel-level segmentation maps contain pixel information about the target class and its coverage area, which is crucial for understanding the spatial layout of objects.

[0094] Next, the pixel-level segmentation map is binarized, converting it into a binary image containing only foreground (target) and background categories. This step simplifies subsequent analysis, making it easier to distinguish which areas belong to objects of interest. A comprehensive evaluation is then performed based on the binarized image, the initial target category, the area of ​​the category region, and the initial center offset to form an object list. This list records the relevant attributes of all detected objects.

[0095] Finally, the true object information (truth table) annotated in the original image is compared with the object list, and localization loss, classification loss, and segmentation loss are calculated based on this. Localization loss measures the difference between the predicted and true bounding boxes; classification loss reflects the degree of match between the predicted and actual categories; and segmentation loss focuses on the consistency between the pixel-level segmentation results and the true labels. Through the backpropagation algorithm, model parameters are adjusted based on these loss functions, and continuous iteration is performed until the optimal solution is reached, ultimately completing the optimization of the CenterPoint model.

[0096] In summary, this series of operations not only improves the model's robustness and accuracy in complex environments, but also ensures that the model can quickly respond to changing driving scenarios while maintaining high accuracy. This design approach reflects the innovative application and development trend of modern computer vision technology in the field of autonomous driving.

[0097] In step S13, the target category, the center offset position, the target size information and the historical target center position set stored in the previous frame are input into the pre-trained target center point motion model to obtain the target center position set.

[0098] In a specific embodiment, the target category, the center offset position, the target size information, and the historical target center position set stored in the previous frame are input into a pre-trained target center point motion model to obtain the target center position set, including:

[0099] Use Kalman filter as the basic model to establish the initial motion prediction model;

[0100] Dynamically adjusting the process noise covariance matrix and the measurement noise covariance matrix according to the pre-stored target velocity, pre-stored target acceleration and pre-stored motion position of the previous frame to obtain a motion prediction model;

[0101] Input the target category, the center offset position, the target size information, and the historical target center position set stored in the previous frame into the motion prediction model to perform Kalman filter prediction to generate a preliminary predicted position sequence;

[0102] Optimizing the preliminary predicted position sequence and the pre-stored actual position sequence by a dynamic time warping algorithm to obtain a predicted position sequence;

[0103] Mapping adjustment is performed according to the predicted position sequence to obtain a target center position set.

[0104] In a specific embodiment, the optimizing the preliminary predicted position sequence by a dynamic time warping algorithm based on the preliminary predicted position sequence and the pre-stored actual position sequence to obtain the predicted position sequence includes:

[0105] Calculating corresponding Euclidean distances based on the preliminary predicted position sequence and the pre-stored actual position sequence to obtain a distance matrix;

[0106] Creating a cumulative distance matrix based on the distance matrix, and filling the cumulative distance matrix through dynamic programming to obtain an optimal path;

[0107] Backtracking the optimal path by comparing the cumulative distances of adjacent cells of the cumulative distance matrix to form a matching pair list;

[0108] The preliminary predicted position sequence is adjusted according to the matching pair list by an interpolation method to obtain a predicted position sequence.

[0109] Specifically, an initial motion prediction model was first established based on a Kalman filter. This model dynamically adjusts the process noise covariance matrix and the measurement noise covariance matrix by incorporating historical target velocity, acceleration, and position information, resulting in a more accurate motion prediction model. Specifically, the process noise covariance matrix reflects an estimate of the uncertainty of the system's internal variations, while the measurement noise covariance matrix measures the uncertainty of the observations. For different motion modes (such as acceleration, constant speed, or deceleration), selecting an appropriate process noise covariance matrix can improve the filter's tracking stability and convergence speed.

[0110] In the process of dynamically adjusting the process noise covariance matrix and the measurement noise covariance matrix, the core is to reflect the uncertainty of the internal changes in the system and the uncertainty of the observation value based on the historical target speed, acceleration and motion position information. This adjustment is crucial to improving the performance of the Kalman filter, ensuring that the filter can adapt to the actual behavior of the system and maintain good robustness and accuracy in the face of environmental changes. First, consider the process noise covariance matrix, which reflects the uncertainty in the system model and external disturbances. When new measurement data arrives, the difference between the predicted state and the actual measurement can be evaluated to determine whether the currently used process noise covariance matrix is ​​appropriate. If the predicted state error is found to be large, it indicates that the changes within the system are beyond expectations. At this time, the process noise covariance matrix needs to be increased to allow for greater state uncertainty. Conversely, if the predicted state is consistent with the actual measurement, the process noise covariance matrix can be appropriately reduced, thereby reducing the trust in the new measurement data. The specific adjustment formula is as follows:

[0111] ;

[0112] here, Indicates the The process noise covariance matrix after iterations; is the current estimated state covariance matrix; is the state covariance matrix saved after the last iteration.

[0113] Next is the adjustment of the measurement noise covariance matrix. The measurement noise covariance matrix describes the level of random error introduced during the measurement process. Similarly, by comparing the residuals (i.e., errors) between the predicted state and the actual measurement results, it can be determined whether the existing measurement noise covariance matrix is ​​reasonable. If the residuals deviate significantly from zero, it means that the measurement noise is larger than expected, and the measurement noise covariance matrix should be increased. Conversely, if the residuals are small, it means that the measurement is more accurate, and the measurement noise covariance matrix can be reduced. The adjustment formula is:

[0114] ;

[0115] in, Representative The measurement noise covariance matrix after iterations is, is the actual measured value, The predicted measurement value calculated based on the predicted state.

[0116] To achieve this adjustment, an adaptive mechanism is incorporated into the algorithm. After each update step, the prediction error is checked and the process noise covariance matrix and the measurement noise covariance matrix are modified accordingly. For example, in a specific implementation, suppose a moving target is being tracked, and its position is measured using a radar. Because the target may be subject to unknown random perturbations, an adaptive unscented Kalman filter (AUKF) is used to estimate the target's position and velocity. The AUKF not only performs the standard prediction and update operations but also dynamically updates the process noise covariance matrix and the measurement noise covariance matrix based on the prediction error, ensuring that the filter is always operating optimally.

[0117] The significance of this adaptive adjustment lies in that it enables the Kalman filter to automatically adjust parameters without fully understanding the system's characteristics, thereby better adapting to changing conditions. This approach maintains high estimation accuracy even in nonlinear or non-Gaussian noise environments. Furthermore, this approach reduces the workload of manual parameter adjustment and improves the system's automation.

[0118] Next, the target category, center offset, and size information, along with the previously stored historical target center position set from the previous frame, are fed into the optimized motion prediction model to perform the Kalman filter prediction phase. In this phase, the model predicts the next state based on the current state estimate, generating a preliminary predicted position sequence. The prediction formula is:

[0119] ;

[0120] in, is in the time step Predictions of future states, is a pre-stored state transition matrix that describes how the system state evolves over time. is the state estimate for the previous time step, is the pre-stored control input matrix, is the pre-stored external control vector; is the forecast error covariance matrix, is the prediction error covariance matrix of the previous time step, is the process noise covariance matrix.

[0121] To further optimize the preliminary predicted position sequence, a dynamic time warping (DTW) algorithm was introduced. DTW compares two time series of potentially different lengths and minimizes the cumulative distance between them by finding the best matching path between the two sequences. Here, the preliminary predicted position sequence is compared with a pre-stored actual position sequence to find the most similar correspondence between the two. Euclidean distances are calculated to form a distance matrix. Next, a cumulative distance matrix is ​​created and populated using a dynamic programming method until the optimal path is found. Once the optimal path is determined, adjacent cells in the cumulative distance matrix are backtracked to form a list of matching pairs. These matching pairs define how each point in the preliminary predicted position sequence should be adjusted to better match the actual position sequence. Finally, interpolation methods are used to adjust the preliminary predicted position sequence, ensuring that the adjusted sequence is not only morphologically close to the actual position sequence but also maintains continuity and smoothness. This step improves the quality of the predicted position sequence, making the final output target center position set more accurate and reliable.

[0122] In summary, combining the Kalman filter with the DTW algorithm not only effectively predicts target trajectories but also enhances the model's ability to understand target behavior in complex environments. This approach provides more accurate target tracking results while ensuring real-time performance, which is of great significance for applications such as autonomous driving.

[0123] In step S14, the target category, the target center position set, and the target size information are rendered into the original image to obtain a rendered image.

[0124] In a specific embodiment, rendering the object category, the object center position set, and the object size information into the original image to obtain a rendered image includes:

[0125] Calculating the frame corner coordinates according to the target center position set and the target size information;

[0126] matching target color and target shape according to the target category;

[0127] Draw a target logo on the original image using a drawing function in a pre-stored image processing library according to the frame corner coordinates, the target color, and the target shape to obtain a rendered image;

[0128] The frame corner coordinates are calculated using the following formula: ;

[0129] in, Indicates the coordinates of the upper left corner of the border. Indicates the coordinates of the upper right corner of the border. Indicates the coordinates of the lower left corner of the border. Indicates the coordinates of the lower right corner of the border. represents the center point coordinates, Indicates the target width in the target size information. Indicates the target height in the target size information.

[0130] Specifically, we first need to determine the coordinates of the four corner points of the rectangular frame surrounding the target based on the known target center position and target size information. The positions of these corner points can be accurately calculated using the following formula:

[0131] ;

[0132] in, Indicates the coordinates of the upper left corner of the border. Indicates the coordinates of the upper right corner of the border. Indicates the coordinates of the lower left corner of the border. Indicates the coordinates of the lower right corner of the border. represents the center point coordinates, Indicates the target width in the target size information. Indicates the target height in the target size information.

[0133] Next, each object is labeled with a color and shape based on its category. For example, for pedestrian detection, a red rectangle might be used, while a vehicle might be represented with a blue oval. This categorical labeling not only helps distinguish different types of objects but also makes the results more intuitive and understandable. The color is based on a predefined color map, while the shape depends on the specific application requirements.

[0134] The final step is to actually draw the target logo. This step relies on the drawing functions provided by one or more image processing libraries. For example, calling or function, passing in the previously calculated four corner coordinates as parameters, and specifying the fill color and other style properties.

[0135] In summary, by rendering the target category, center location, and size information into the original image, we can effectively display the detection results. This approach not only helps quickly understand the image content, but also ensures that key information can be quickly identified, improving decision-making efficiency.

[0136] In step S15, target detection is performed based on the rendered image, and target tracking is performed based on the target detection result.

[0137] In a specific embodiment, performing target detection based on the rendered image and optimizing the target motion model based on the target detection result includes:

[0138] Compare the rendered image with the rendered image of the previous frame to determine whether the target in the current frame is the same as the target in the previous frame. If the target in the current frame is the same as the target in the previous frame, do not update the pre-stored target motion model and perform target detection in the next frame.

[0139] If the target of the current frame is different from the target of the previous frame, determine whether the pre-stored tracker list contains the target motion model of the current frame. If the pre-stored tracker list does not contain the target motion model of the current frame, the corresponding tracker is invalid and the invalid tracker is deleted.

[0140] If the pre-stored tracker list contains the target motion model of the current frame, the tracker list is updated and the predicted target position is generated based on the rendered image.

[0141] Specifically, the system first determines whether the target has changed by comparing the rendered image of the current frame with the previous frame. This determination is based on image feature matching technology, such as using a deep learning model to extract a similarity score between the two frames or calculating the degree of change in the target's position, size, and appearance features. If the target in the current frame is found to be the same as the target in the previous frame, it means that the target has not moved significantly or changed in shape. Therefore, there is no need to update the pre-stored target motion model and the target detection process for the next frame can be directly entered.

[0142] However, if the target in the current frame is different from the target in the previous frame, it is necessary to further check whether the corresponding target motion model already exists in the pre-stored tracker list. This step ensures that the system can still identify and track the same target even if the target is occluded, suddenly accelerated, or in other complex situations. For those targets that no longer appear in the current frame, that is, the tracker is invalid, these invalid trackers will be removed from the tracker list to maintain the validity and real-time performance of the list.

[0143] If the pre-stored tracker list contains the target motion model for the current frame, an update of the tracker list is triggered. The update process involves several key steps: state prediction, data association, and state update and trajectory management. Specifically, during the state prediction phase, a Kalman filter is used in conjunction with a simplified motion model (such as uniform velocity or uniform acceleration) to predict the target's position at a specific moment in the future based on historical information from previous frames.

[0144] Next, in the data association phase, the Hungarian algorithm is used to match the predicted target location with the actual detected target location. The goal is to find one or more pairs of predicted and detected boxes that are most likely to belong to the same physical entity. Once the matching is complete, the state update phase begins, adjusting the Kalman filter state parameters while maintaining the target trajectory. At this point, if there are unmatched high-confidence detection boxes whose confidence exceeds a set threshold, a new tracking track is created. Conversely, low-confidence detection boxes or tracks that have not been successfully matched for a long time are marked for deletion.

[0145] Finally, the process of generating the predicted target position relies not only on historical trajectory analysis but also takes into account factors such as the target's velocity and acceleration. Dynamic adjustments to the process noise covariance matrix and the measurement noise covariance matrix ensure that the prediction results are as close to the actual situation as possible. For example, when the target accelerates or decelerates, the process noise covariance matrix is ​​appropriately increased to allow for greater state uncertainty. For more stable motion patterns, the process noise covariance matrix is ​​reduced to reduce sensitivity to external disturbances. Similarly, the measurement noise covariance matrix is ​​adjusted based on the residual size to ensure the reliability of the measured values.

[0146] In summary, by performing object detection on rendered images and optimizing the target motion model based on the detection results, we achieve effective tracking of moving targets. This approach not only improves the stability and accuracy of the tracking system but also demonstrates good adaptability in complex scenarios. Furthermore, by continuously updating and maintaining the tracker list, we ensure that each target receives continuous attention, maintaining tracking continuity even in the presence of occlusion or rapid movement. The entire process achieves a seamless transition from detection to tracking.

[0147] The following describes the working process of the present invention using a common scenario as an example. Figure 1 , a target detection optimization method based on Center Point model, comprising the following steps:

[0148] When applying the CenterPoint model-based object detection optimization method, the autonomous vehicle captures a raw image of the current road scene through its cameras. As the vehicle drives, the sensors continuously update the environment information. Each new image frame is immediately input into the pre-trained CenterPoint model for inference analysis.

[0149] Next, the system uses the ResNet backbone to extract image features and fuses scene features from different levels using a feature pyramid network. After processing by the CenterBox Head and RetinaNet networks, the object's location prediction and bounding box are determined. The KeyPoint Head then further identifies the object's specific category. These steps work together to enable the system to accurately understand each element in the image.

[0150] For continuous video streams, the system relies not only on information from the current frame but also combines historical data from previous frames to estimate the target's motion trends. A motion prediction model built using a Kalman filter dynamically adjusts parameters to ensure accurate tracking of the target's position, even during rapid movement or under occlusion. The dynamic time warping algorithm employed in this process helps optimize the predicted path, ensuring consistent and coherent trajectory.

[0151] After completing the above processing, the system renders the identified target information back onto the original image, presenting the detection results in an intuitive manner. Each target is assigned a specific color and shape to clearly distinguish different object types. Finally, a new round of target detection and tracking is performed based on the rendered image, maintaining the system's real-time responsiveness.

[0152] Throughout the process, the method significantly enhances the ability to understand target behavior in complex traffic environments, particularly challenging scenarios involving fast-moving objects or partial occlusion. This approach ensures that autonomous vehicles can more reliably perceive their surroundings, improving driving safety while also boosting the overall performance of intelligent transportation systems.

[0153] Reference Figure 2 A second embodiment of the present invention provides a target detection optimization system based on a Center Point model, comprising:

[0154] Data acquisition module, used to obtain the original image of the current frame;

[0155] The target information inference module is used to input the original image into the CenterPoint model for inference to obtain target category, center offset position and target size information;

[0156] A center position generation module is used to input the target category, the center offset position, the target size information, and the historical target center position set stored in the previous frame into a pre-trained target center point motion model to obtain a target center position set;

[0157] An image rendering module renders the target category, the target center position set, and the target size information into the original image to obtain a rendered image;

[0158] The target tracking module is used to perform target detection based on the rendered image and to track the target based on the target detection result.

[0159] It should be noted that the target detection optimization device based on the Center Point model provided in an embodiment of the present invention is used to execute all the process steps of the target detection optimization method based on the Center Point model in the above embodiment. The working principles and beneficial effects of the two correspond one to one, so they will not be repeated here.

[0160] An embodiment of the present invention further provides an electronic device. The electronic device includes: a processor, a memory, and a computer program stored in the memory and executable on the processor, such as a target detection optimization program based on a Center Point model. When the processor executes the computer program, the steps in each of the above-mentioned target detection optimization method embodiments based on a Center Point model are implemented, such as Figure 1 Alternatively, when the processor executes the computer program, the processor implements the functions of the modules / units in the above-mentioned device embodiments, such as the target detection optimization module based on the Center Point model.

[0161] Exemplarily, the computer program may be divided into one or more modules / units, which are stored in the memory and executed by the processor to implement the present invention. The one or more modules / units may be a series of computer program instruction segments capable of implementing specific functions, and the instruction segments are used to describe the execution process of the computer program in the electronic device.

[0162] The electronic device may be a computing device such as a desktop computer, notebook, PDA, or smart tablet. The electronic device may include, but is not limited to, a processor and memory. Those skilled in the art will appreciate that the aforementioned components are merely examples of electronic devices and do not constitute a limitation of the electronic device. The electronic device may include more or fewer components than those described above, or a combination of certain components, or different components. For example, the electronic device may also include input / output devices, network access devices, buses, and the like.

[0163] The processor may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor. The processor is the control center of the electronic device and connects various parts of the entire electronic device using various interfaces and lines.

[0164] The memory can be used to store the computer programs and / or modules. The processor implements the various functions of the electronic device by running or executing the computer programs and / or modules stored in the memory and accessing the data stored in the memory. The memory may primarily include a program storage area and a data storage area. The program storage area may store an operating system and at least one application required for a function (such as a sound playback function or an image playback function); the data storage area may store data generated based on the use of the mobile phone (such as audio data, a phone book, etc.). Furthermore, the memory may include high-speed random access memory and non-volatile memory, such as a hard disk, internal memory, a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, at least one disk storage device, a flash memory device, or other volatile solid-state storage device.

[0165] If the module / unit integrated into the electronic device is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present invention can implement all or part of the process steps in the above-mentioned method embodiments by using a computer program to instruct the relevant hardware. The computer program can be stored in a computer-readable storage medium. When executed by a processor, the computer program can implement the steps of each of the above-mentioned method embodiments. The computer program includes computer program code, which can be in source code form, object code form, executable file, or some intermediate form. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard drive, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signal, telecommunication signal, and software distribution medium. It should be noted that the content of the computer-readable medium can be appropriately increased or decreased based on the requirements of legislation and patent practice in a jurisdiction. For example, in some jurisdictions, based on legislation and patent practice, computer-readable media does not include electric carrier signals and telecommunication signals.

[0166] It should be noted that the device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed across multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the present embodiment. In addition, in the drawings of the device embodiments provided by the present invention, the connection relationship between the modules indicates that there is a communication connection between them, which may be specifically implemented as one or more communication buses or signal lines. A person of ordinary skill in the art can understand and implement the present invention without inventive effort.

[0167] The specific embodiments described above further illustrate the objectives, technical solutions, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included within the scope of protection of the present invention for those skilled in the art.

Claims

1. A target detection optimization method based on Center Point model, characterized in that: Executed by the controller, including: Get the original picture of the current frame; The original image is input into the CenterPoint model for inference to obtain the target category, center offset position, and target size information, including: Input the original image into the ResNet backbone network of the optimized CenterPoint model to extract image features and obtain a feature map; Using a feature pyramid network to perform scene feature fusion on the feature map to obtain a fused feature map; The fused feature map is input into the CenterBox Head for position prediction to obtain the center offset position and target size information; Input the fused feature map into the RetinaNet network for regression to obtain the target prediction frame; Input the center offset position and the target prediction box into the KeyPoint Head for type recognition to obtain the target category; The optimization process of the CenterPoint model includes: Input the original image into the CNN backbone network of the initial CenterPoint model to extract image features and obtain the initial feature map; Input the feature map into the CenterBox Head for position prediction to obtain the initial center offset position; Inputting the initial center offset position into the KeyPoint Head to perform target type recognition to obtain an initial target category; Inputting the initial feature map into the MaskNetHead model to obtain a pixel-level segmentation map, wherein the pixel-level segmentation map includes pixel information of the target category and the category region area; performing binarization processing on the pixel-level segmentation map to obtain a binarized image; Performing target judgment based on the binarized image, the initial target category, the category area, and the initial center offset, and counting the initial target category, the category area, and the initial center offset into a target list according to the judgment result; Comparing the target truth table in the original image with the target list and calculating the positioning loss, classification loss, and segmentation loss, and iteratively optimizing the initial CenterPoint model through backpropagation based on the calculation results to obtain the CenterPoint model; Inputting the target category, the center offset position, the target size information, and a historical target center position set stored in a previous frame into a pre-trained target center point motion model to obtain a target center position set; Rendering the target category, the target center position set, and the target size information into the original image to obtain a rendered image; Target detection is performed based on the rendered image, and target tracking is performed based on the target detection result.

2. The target detection optimization method based on the Center Point model according to claim 1, characterized in that: The target category, the center offset position, the target size information, and the historical target center position set stored in the previous frame are input into a pre-trained target center point motion model to obtain the target center position set, including: Use Kalman filter as the basic model to establish the initial motion prediction model; Dynamically adjusting the process noise covariance matrix and the measurement noise covariance matrix according to the pre-stored target velocity, pre-stored target acceleration and pre-stored motion position of the previous frame to obtain a motion prediction model; Input the target category, the center offset position, the target size information, and the historical target center position set stored in the previous frame into the motion prediction model to perform Kalman filter prediction to generate a preliminary predicted position sequence; Optimizing the preliminary predicted position sequence and the pre-stored actual position sequence by a dynamic time warping algorithm to obtain a predicted position sequence; Mapping adjustment is performed according to the predicted position sequence to obtain a target center position set.

3. The target detection optimization method based on the Center Point model according to claim 2, characterized in that: The step of optimizing the preliminary predicted position sequence and the pre-stored actual position sequence by a dynamic time warping algorithm to obtain a predicted position sequence includes: Calculating corresponding Euclidean distances based on the preliminary predicted position sequence and the pre-stored actual position sequence to obtain a distance matrix; Creating a cumulative distance matrix based on the distance matrix, and filling the cumulative distance matrix through dynamic programming to obtain an optimal path; Backtracking the optimal path by comparing the cumulative distances of adjacent cells of the cumulative distance matrix to form a matching pair list; The preliminary predicted position sequence is adjusted according to the matching pair list by an interpolation method to obtain a predicted position sequence.

4. The target detection optimization method based on the Center Point model according to claim 1, characterized in that: The rendering of the target category, the target center position set, and the target size information into the original image to obtain a rendered image includes: Calculating the frame corner coordinates according to the target center position set and the target size information; matching target color and target shape according to the target category; Draw a target logo on the original image using a drawing function in a pre-stored image processing library according to the frame corner coordinates, the target color, and the target shape to obtain a rendered image; The frame corner coordinates are calculated using the following formula: ； in, Indicates the coordinates of the upper left corner of the border. Indicates the coordinates of the upper right corner of the border. Indicates the coordinates of the lower left corner of the border. Indicates the coordinates of the lower right corner of the border. represents the center point coordinates, Indicates the target width in the target size information. Indicates the target height in the target size information.

5. The target detection optimization method based on the Center Point model according to claim 1, characterized in that: The performing target detection according to the rendered image and optimizing the target motion model according to the target detection result includes: Compare the rendered image with the rendered image of the previous frame to determine whether the target in the current frame is the same as the target in the previous frame. If the target in the current frame is the same as the target in the previous frame, do not update the pre-stored target motion model and perform target detection in the next frame. If the target of the current frame is different from the target of the previous frame, determine whether the pre-stored tracker list contains the target motion model of the current frame. If the pre-stored tracker list does not contain the target motion model of the current frame, the corresponding tracker is invalid and the invalid tracker is deleted. If the pre-stored tracker list contains the target motion model of the current frame, the tracker list is updated and the predicted target position is generated based on the rendered image.

6. A target detection optimization system based on Center Point model, characterized in that: The method for implementing the target detection optimization method based on the Center Point model according to any one of claims 1 to 5 comprises: Data acquisition module, used to obtain the original image of the current frame; The target information inference module is used to input the original image into the CenterPoint model for inference to obtain target category, center offset position and target size information; A center position generation module is used to input the target category, the center offset position, the target size information, and the historical target center position set stored in the previous frame into a pre-trained target center point motion model to obtain a target center position set; An image rendering module renders the target category, the target center position set, and the target size information into the original image to obtain a rendered image; The target tracking module is used to perform target detection based on the rendered image and to track the target based on the target detection result.

7. An electronic device, characterized in that: The method comprises a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, the target detection optimization method based on the Center Point model as described in any one of claims 1 to 5 is implemented.

8. A computer-readable storage medium, characterized in that The computer-readable storage medium includes a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute the target detection optimization method based on the Center Point model according to any one of claims 1 to 5.

Images