Target detection method and device, computer equipment and storage medium

Abstract

The invention relates to a target detection method and device, computer equipment, a storage medium and a computer program product. The method comprises the following steps: acquiring multi-frame point cloud data obtained by scanning a to-be-identified target in a target scene; according to the point cloud attribute information of each frame of point cloud data, determining the corresponding weight of each frame of point cloud data in the multiple frames of point cloud data; wherein the weight corresponding to the point cloud data is in positive correlation with the data reliability corresponding to the point cloud data; according to the weight corresponding to each frame of point cloud data, performing weighted fusion on the multiple frames of point cloud data to obtain fused point cloud data; and extracting target feature data corresponding to the fused point cloud data, and obtaining a classification detection result of the to-be-identified target according to the target feature data. By adopting the method, the target classification detection accuracy can be improved.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and in particular, to an object detection method, device, computer device, storage medium, and computer program product. Background Art

[0002] With the development of computer technology, object classification and detection is a research hotspot in fields such as graphics and machine vision. Its main purpose is to distinguish different types of objects according to the different features reflected in each data, and correctly identifying objects is a key task for realizing machine intelligence.

[0003] As an active remote sensing tool, radar is widely used in various industries, such as topographic surveying, atmospheric monitoring, unmanned driving and other industries. Due to the high sampling frequency of radar, the resulting point cloud data after acquisition is also large. In the related art, when performing object classification and detection based on this large amount of point cloud data, the accuracy rate is relatively low.

[0004] Therefore, there is a problem of low accuracy in object classification and detection in the related art. Summary of the Invention

[0005] Based on this, it is necessary to provide an object detection method, device, computer device, computer-readable storage medium, and computer program product that can improve the accuracy of object classification and detection for the above technical problems.

[0006] In a first aspect, this application provides an object detection method, including:

[0007] Obtaining multiple frames of point cloud data obtained by scanning a target to be recognized in a target scene;

[0008] Determining the weight corresponding to each frame of the point cloud data in the multiple frames of point cloud data according to the point cloud attribute information of each frame of the point cloud data; wherein, the weight corresponding to the point cloud data is positively correlated with the data reliability corresponding to the point cloud data;

[0009] Performing weighted fusion on the multiple frames of point cloud data according to the weights corresponding to each frame of the point cloud data to obtain fused point cloud data;

[0010] Extracting target feature data corresponding to the fused point cloud data, and obtaining a classification and detection result of the target to be recognized according to the target feature data.

[0011] In one of the embodiments, the point cloud attribute information includes point cloud attribute values; the determining the weight corresponding to each frame of the point cloud data in the multiple frames of point cloud data according to the point cloud attribute information of each frame of the point cloud data includes:

[0012] Obtain the sum of the point cloud attribute values corresponding to each frame of the point cloud data to obtain the sum of the point cloud attribute values;

[0013] For any frame of the multi-frame point cloud data, obtain the weight corresponding to the any frame of the point cloud data in the multi-frame point cloud data according to the ratio of the point cloud attribute value corresponding to the any frame of the point cloud data to the sum of the point cloud attribute values.

[0014] In one embodiment, when the point cloud attribute information includes point cloud attribute values corresponding to at least two types of point cloud attribute types, the obtaining the sum of the point cloud attribute values corresponding to each frame of the point cloud data to obtain the sum of the point cloud attribute values includes:

[0015] For each type of point cloud attribute, obtain the sum of the point cloud attribute values corresponding to each frame of the point cloud data to obtain the sum of the point cloud attribute values corresponding to each type of point cloud attribute;

[0016] The obtaining the weight corresponding to the any frame of the point cloud data in the multi-frame point cloud data according to the ratio of the point cloud attribute value corresponding to the any frame of the point cloud data to the sum of the point cloud attribute values includes:

[0017] According to the ratio of the point cloud attribute value corresponding to the any frame of the point cloud data under each type of point cloud attribute to the sum of the point cloud attribute values corresponding to the corresponding type of point cloud attribute, obtain the weight corresponding to the any frame of the point cloud data in the multi-frame point cloud data.

[0018] In one embodiment, the obtaining the weight corresponding to the any frame of the point cloud data in the multi-frame point cloud data according to the ratio of the point cloud attribute value corresponding to the any frame of the point cloud data under each type of point cloud attribute to the sum of the point cloud attribute values corresponding to the corresponding type of point cloud attribute includes:

[0019] According to the ratio of the point cloud attribute value corresponding to the any frame of the point cloud data under each type of point cloud attribute to the sum of the point cloud attribute values corresponding to the corresponding type of point cloud attribute, obtain the weight corresponding to the any frame of the point cloud data under each type of point cloud attribute;

[0020] According to the average value of the weights corresponding to the any frame of the point cloud data under each type of point cloud attribute, obtain the weight corresponding to the any frame of the point cloud data in the multi-frame point cloud data.

[0021] In one embodiment, the point cloud attribute value includes at least one of a point cloud signal-to-noise ratio value and a point cloud quantity value.

[0022] In one embodiment, the extracting the target feature data corresponding to the fused point cloud data includes:

[0023] Obtain the features corresponding to the fused point cloud data under the discriminative feature type, and obtain the discriminative features corresponding to the fused point cloud data; wherein, the features corresponding to different categories of targets under the discriminative feature type satisfy a preset difference condition;

[0024] According to the discriminative features corresponding to the fused point cloud data, obtain the target feature data corresponding to the fused point cloud data.

[0025] In one embodiment, the method further includes:

[0026] Under each feature type, perform histogram analysis on the fused sample point cloud data corresponding to sample targets of different categories, and obtain the histograms corresponding to each category under each feature type;

[0027] According to the differences between the histograms corresponding to each category under each feature type, determine the feature type whose differences satisfy the preset difference condition as the discriminative feature type.

[0028] In one embodiment, the discriminative features include histogram features; the step of obtaining the features corresponding to the fused point cloud data under the discriminative feature type and obtaining the discriminative features corresponding to the fused point cloud data includes:

[0029] Obtain the histogram corresponding to the fused point cloud data under the discriminative feature type;

[0030] According to the frequency information of the histogram, obtain the histogram features corresponding to the fused point cloud data under the discriminative feature type.

[0031] In one embodiment, when there are at least two discriminative feature types, the step of obtaining the target feature data corresponding to the fused point cloud data according to the discriminative features corresponding to the fused point cloud data includes:

[0032] Fuse the histogram features corresponding to the fused point cloud data under each discriminative feature type in a cascaded manner to obtain the target feature data.

[0033] In a second aspect, the present application further provides a target detection device, including:

[0034] An acquisition module, configured to acquire multiple frames of point cloud data obtained by scanning a target to be recognized in a target scene;

[0035] A determination module, configured to determine the weights corresponding to each frame of the point cloud data in the multi-frame point cloud data according to the point cloud attribute information of each frame of the point cloud data; wherein, the weight corresponding to the point cloud data is positively correlated with the data reliability corresponding to the point cloud data;

[0036] A fusion module, configured to perform weighted fusion on the multi-frame point cloud data according to the weights corresponding to each frame of the point cloud data to obtain fused point cloud data;

[0037] An identification module, configured to extract target feature data corresponding to the fused point cloud data, and obtain a classification detection result of the target to be identified according to the target feature data.

[0038] In a third aspect, the present application further provides a computer device. The computer device includes a memory and a processor, the memory stores a computer program, and when the computer program is executed by the processor, the steps of the above method are implemented.

[0039] In a fourth aspect, the present application further provides a computer-readable storage medium. The computer-readable storage medium stores a computer program, and when the computer program is executed by the processor, the steps of the above method are implemented.

[0040] In a fifth aspect, the present application further provides a computer program product. The computer program product includes a computer program, and when the computer program is executed by the processor, the steps of the above method are implemented.

[0041] The above object detection method, device, computer device, storage medium, and computer program product obtain multi-frame point cloud data obtained by scanning a target to be identified in a target scene; determine the weights corresponding to each frame of the point cloud data in the multi-frame point cloud data according to the point cloud attribute information of each frame of the point cloud data; wherein, the weight corresponding to the point cloud data is positively correlated with the data reliability corresponding to the point cloud data; perform weighted fusion on the multi-frame point cloud data according to the weights corresponding to each frame of the point cloud data to obtain fused point cloud data; extract target feature data corresponding to the fused point cloud data, and obtain a classification detection result of the target to be identified according to the target feature data.

[0042] Thus, since the attribute information corresponding to different frame point cloud data may be different, resulting in different data reliabilities for different frame point cloud data, it is possible to assign a weight to each frame of point cloud data that is positively correlated with the data reliability corresponding to the point cloud data according to the point cloud attribute information related to the point cloud data reliability of each frame of point cloud data. Thus, it is possible to accurately set the corresponding weights for each frame of point cloud data in multiple frames of point cloud data based on the point cloud data reliability, and then perform weighted fusion on the multiple frames of point cloud data through the weights corresponding to each frame of point cloud data. This can not only improve the density of point cloud information and solve the problem that single-frame point cloud data cannot fully and accurately represent the information of the target to be recognized, but also more accurately determine the classification and detection results of the target to be recognized based on the target feature data corresponding to the fused point cloud data, effectively improving the accuracy of target classification, recognition, and tracking. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the following will briefly introduce the drawings required for use in the description of the embodiments or related technologies. Obviously, the drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0044] Figure 1 It is a schematic flowchart of a target detection method in an embodiment;

[0045] FIG. 2(a) is a schematic histogram of sample targets belonging to different categories under height features in an embodiment;

[0046] FIG. 2(b) is a schematic histogram of sample targets belonging to different categories under size features in an embodiment;

[0047] FIG. 2(c) is a schematic histogram of sample targets belonging to different categories under SNR features in an embodiment;

[0048] Figure 3 It is an overall architecture diagram of a target detection method in an embodiment;

[0049] Figure 4 It is a schematic flowchart of a target detection method in another embodiment;

[0050] Figure 5 It is a structural block diagram of a target detection device in an embodiment;

[0051] Figure 6 It is an internal structure diagram of a computer device in an embodiment. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0052] To make the objectives, technical solutions and advantages of this application more clear and understandable, the following further elaborates on this application in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely used to explain this application and are not used to limit this application.

[0053] It should be noted that the terms "first", "second", etc. in the specification and claims of this disclosure and the above-mentioned drawings are used to distinguish similar objects and do not necessarily need to describe a specific order or sequence. It should be understood that such used data can be interchanged under appropriate circumstances so that the embodiments of this disclosure described herein can be implemented in an order other than those illustrated or described herein. The implementation manners described in the following exemplary embodiments do not represent all implementation manners consistent with this disclosure. On the contrary, they are merely examples of devices and methods consistent with some aspects of this disclosure as detailed in the appended claims.

[0054] In one embodiment, as Figure 1 shown, a target detection method is provided. In this embodiment, taking the application of this method to a terminal as an example, it can be understood that this method can also be applied to a server and can also be applied to a system including a terminal and a server and be implemented through the interaction between the terminal and the server. Among them, the terminal can be, but is not limited to, various personal computers, laptop computers, smart phones, tablet computers, Internet of Things devices and portable wearable devices. The Internet of Things devices can be smart speakers, smart TVs, smart air conditioners, smart in-vehicle devices, radar devices, etc. The portable wearable devices can be smart watches, smart bracelets, head-mounted devices, etc. In this embodiment, this method includes the following steps:

[0055] Step S110, obtain multiple frames of point cloud data obtained by scanning a target to be recognized in a target scene.

[0056] Among them, the target scene can be a scene where there are target objects to be classified and recognized.

[0057] Among them, the target to be recognized is the target object to be classified and recognized.

[0058] Exemplarily, the target scene can be an indoor scene, and the targets to be recognized can be target objects such as adults, children, floor sweepers, pets, and green plants in the indoor scene.

[0059] In specific implementation, the terminal can obtain multiple frames of point cloud data obtained by scanning the target to be recognized in the target scene. Among them, a radar device is installed in the target scene, and the point cloud data can be obtained by the radar device scanning the target to be recognized in the target scene, so that the terminal can obtain multiple frames of point cloud data scanned by the radar device.

[0060] It can be understood that the radar device may include, but is not limited to, a millimeter-wave radar device. In addition, the radar device may be integrated into a terminal. The terminal located in the target scenario can scan the target to be recognized in the target scenario through the radar device to obtain multiple frames of point cloud data. Exemplarily, the millimeter-wave radar device may be integrated into terminals such as smart homes, medical care, and intelligent transportation devices.

[0061] Step S120: Determine the weights corresponding to each frame of point cloud data in the multiple frames of point cloud data according to the point cloud attribute information of each frame of point cloud data.

[0062] Among them, the weight corresponding to the point cloud data is positively correlated with the data reliability (point cloud data reliability) corresponding to the point cloud data.

[0063] Among them, the point cloud attribute information may be the attribute information of the point cloud data. Specifically, the point cloud attribute information may be the attribute information related to the point cloud data reliability.

[0064] In a specific implementation, since the attribute information corresponding to different frames of point cloud data may be different, resulting in different data reliabilities corresponding to different frames of point cloud data, the terminal can assign a weight positively correlated with the data reliability corresponding to the point cloud data to each frame of point cloud data according to the point cloud attribute information related to the point cloud data reliability of each frame of point cloud data, so as to set the corresponding weights for each frame of point cloud data in the multiple frames of point cloud data respectively. In this way, the higher the data reliability corresponding to the point cloud data, the higher the weight corresponding to it in the multiple frames of point cloud data.

[0065] Step S130: Perform weighted fusion on the multiple frames of point cloud data according to the weights corresponding to each frame of point cloud data to obtain the fused point cloud data.

[0066] In a specific implementation, since a single frame of point cloud data cannot fully and accurately represent the information of the target to be recognized, the terminal can perform weighted fusion on the multiple frames of point cloud data according to the weights corresponding to each frame of point cloud data to obtain the fused point cloud data, so as to improve the density of the point cloud data.

[0067] Step S140: Extract the target feature data corresponding to the fused point cloud data, and obtain the classification and detection result of the target to be recognized according to the target feature data.

[0068] Among them, the target feature data may be discriminative feature data.

[0069] In a specific implementation, the terminal can extract features from the fused point cloud data to obtain the target feature data corresponding to the fused point cloud data, and perform classification and detection on the target to be recognized based on the target feature data to obtain the classification and detection result of the target to be recognized.

[0070] Specifically, after obtaining the target feature data corresponding to the fused point cloud data, the terminal can input the target feature data into the trained target classification model to obtain the classification detection result of the target to be recognized. Further, the target classification model can be a multi-classification model, which predicts the probabilities that the target to be recognized is classified into different classes respectively, and thus determines its final classification detection result according to the probabilities that the target to be recognized is classified into different classes respectively.

[0071] In practical applications, in a smart home, a millimeter-wave radar device can automatically adjust device parameters such as the brightness of a light and the temperature of an air conditioner in combination with the category, position, and dynamic requirements of indoor scene targets, etc., to achieve intelligent environmental control.

[0072] In the above target detection method, multiple frames of point cloud data obtained by scanning the target to be recognized in the target scene are acquired; according to the point cloud attribute information of each frame of point cloud data, the weight corresponding to each frame of point cloud data in the multiple frames of point cloud data is determined; wherein, the weight corresponding to the point cloud data is positively correlated with the data reliability corresponding to the point cloud data; according to the weights corresponding to each frame of point cloud data, the multiple frames of point cloud data are weighted and fused to obtain the fused point cloud data; the target feature data corresponding to the fused point cloud data is extracted, and according to the target feature data, the classification detection result of the target to be recognized is obtained.

[0073] In this way, since the attribute information corresponding to different frames of point cloud data may be different, resulting in different data reliabilities corresponding to different frames of point cloud data, therefore, according to the point cloud attribute information related to the point cloud data reliability of each frame of point cloud data, a weight positively correlated with the data reliability corresponding to the point cloud data is assigned to each frame of point cloud data, so that the corresponding weights of each frame of point cloud data in the multiple frames of point cloud data can be accurately set based on the point cloud data reliability. Furthermore, through the weights corresponding to each frame of point cloud data, the multiple frames of point cloud data are weighted and fused, which can not only improve the density of point cloud information and solve the problem that a single frame of point cloud data cannot fully and accurately represent the information of the target to be recognized, but also more accurately determine the classification detection result of the target to be recognized according to the target feature data corresponding to the fused point cloud data, effectively improving the accuracy of target classification recognition and tracking.

[0074] In an exemplary embodiment, the point cloud attribute information includes point cloud attribute values; determining the weight corresponding to each frame of point cloud data in the multiple frames of point cloud data according to the point cloud attribute information of each frame of point cloud data includes: obtaining the sum of the point cloud attribute values corresponding to each frame of point cloud data to obtain the sum of point cloud attribute values; for any frame of point cloud data in the multiple frames of point cloud data, according to the ratio of the point cloud attribute value corresponding to the any frame of point cloud data to the sum of point cloud attribute values, the weight corresponding to the any frame of point cloud data in the multiple frames of point cloud data is obtained.

[0075] Among them, the point cloud attribute information includes point cloud attribute values, and the point cloud attribute values can be specific values of the point cloud attributes related to the reliability of the point cloud data in the point cloud data. Exemplarily, the terminal can determine the weights corresponding to each frame of point cloud data in the multi-frame point cloud data according to the point cloud attribute values corresponding to each frame of point cloud data.

[0076] In a specific implementation, when the terminal determines the weights corresponding to each frame of point cloud data in the multi-frame point cloud data according to the point cloud attribute values corresponding to each frame of point cloud data, the terminal can obtain the sum of the point cloud attribute values corresponding to each frame of point cloud data to get the sum of point cloud attribute values. At the same time, for any one frame of point cloud data in the multi-frame point cloud data, the terminal can obtain the weight corresponding to the any one frame of point cloud data in the multi-frame point cloud data according to the ratio of the point cloud attribute value corresponding to the any one frame of point cloud data to the sum of the point cloud attribute values. In this way, for each frame of point cloud data in the multi-frame point cloud data, based on the same method, the terminal can obtain the weights corresponding to each frame of point cloud data in the multi-frame point cloud data respectively.

[0077] For example, when the point cloud attribute values corresponding to each frame of point cloud data are point cloud quantity values, assuming that the multi-frame point cloud data is continuous T frames (for example, T = 7, or T = 10, the specific value is not specifically limited here) of point cloud data, the terminal can perform weighted fusion on the continuous T frames of point cloud data by synthesizing the point cloud quantity values corresponding to each frame of point cloud data. Specifically, the weight corresponding to each frame of point cloud data can be the ratio of the point cloud quantity value corresponding to each frame of point cloud data to the sum of the point cloud quantity values corresponding to the continuous T frames of point cloud data, and its calculation method is:

[0078]

[0079] Among them, is the weight corresponding to the i-th frame of point cloud data in the continuous T frames of point cloud data when the point cloud attribute value is the point cloud quantity value. Among them, is the point cloud quantity value of the i-th frame.

[0080] For another example, when the point cloud attribute values corresponding to each frame of point cloud data are point cloud signal-to-noise ratio values (SNR values), assuming that the multi-frame point cloud data is continuous T frames (for example, T = 7, or T = 10, the specific value is not specifically limited here) of point cloud data, the terminal can perform weighted fusion on the continuous T frames of point cloud data by synthesizing the point cloud signal-to-noise ratio values corresponding to each frame of point cloud data. Specifically, the weight corresponding to each frame of point cloud data can be the ratio of the point cloud signal-to-noise ratio value corresponding to each frame of point cloud data to the sum of the point cloud signal-to-noise ratio values corresponding to the continuous T frames of point cloud data, and its calculation method is:

[0081]

[0082] Among them, is the weight corresponding to the i-th frame of point cloud data in consecutive T frames of point cloud data when the point cloud attribute value is the point cloud signal-to-noise ratio. Among them, is the point cloud signal-to-noise ratio of the i-th frame.

[0083] In the technical solution of this embodiment, the point cloud attribute information includes point cloud attribute values; by obtaining the sum of the point cloud attribute values corresponding to each frame of point cloud data, the sum of point cloud attribute values is obtained; for any frame of point cloud data among multiple frames of point cloud data, according to the ratio of the point cloud attribute value corresponding to any frame of point cloud data to the sum of point cloud attribute values, the weight corresponding to any frame of point cloud data in multiple frames of point cloud data is obtained. In this way, through the point cloud attribute value in the point cloud attribute information related to the reliability of point cloud data, the reliability of point cloud data can be accurately quantified. Therefore, in the process of obtaining the weight corresponding to this point cloud data in multiple frames of point cloud data according to the ratio of the point cloud attribute value corresponding to the point cloud data to the sum of point cloud attribute values corresponding to multiple frames of point cloud data, the greater the point cloud attribute value, that is, the higher the data reliability, the greater the weight corresponding to the point cloud data, realizing the accurate assignment of weights corresponding to each frame of point cloud data based on the reliability of point cloud data.

[0084] In one embodiment, when the point cloud attribute information includes point cloud attribute values corresponding to at least two types of point cloud attributes, obtaining the sum of the point cloud attribute values corresponding to each frame of point cloud data to obtain the sum of point cloud attribute values includes: respectively obtaining the sum of the point cloud attribute values corresponding to each frame of point cloud data under each type of point cloud attribute to obtain the sum of point cloud attribute values corresponding to each type of point cloud attribute.

[0085] According to the ratio of the point cloud attribute value corresponding to any frame of point cloud data to the sum of point cloud attribute values, obtaining the weight corresponding to any frame of point cloud data in multiple frames of point cloud data includes: according to the ratio of the point cloud attribute value corresponding to any frame of point cloud data under each type of point cloud attribute to the sum of point cloud attribute values corresponding to the corresponding type of point cloud attribute, obtaining the weight corresponding to any frame of point cloud data in multiple frames of point cloud data.

[0086] Among them, the point cloud attribute value can be the specific value of the point cloud attribute related to the reliability of point cloud data of the point cloud data. The point cloud data can have corresponding values under different types of point cloud attributes. As an optional embodiment of this application, the point cloud attribute value of the point cloud data can include point cloud attribute values corresponding to at least two types of point cloud attributes. For example, the point cloud number value and the point cloud signal-to-noise ratio of the point cloud data are point cloud attribute values corresponding to two types of point cloud attributes (point cloud number and point cloud signal-to-noise ratio respectively).

[0087] In an alternative embodiment of the present application, the point cloud attribute value may include at least one of the point cloud signal-to-noise ratio value and the point cloud quantity value. It can be understood that the weight calculation can also be performed through other point cloud attributes related to the reliability of point cloud data, and the present application does not make specific limitations.

[0088] In a specific implementation, when the point cloud attribute information includes point cloud attribute values corresponding to at least two types of point cloud attributes, in the process of the terminal obtaining the sum of the point cloud attribute values corresponding to each frame of point cloud data to obtain the sum of point cloud attribute values, the terminal can respectively obtain the sum of the point cloud attribute values corresponding to each frame of point cloud data under each type of point cloud attribute to obtain the sum of point cloud attribute values corresponding to each type of point cloud attribute.

[0089] Furthermore, in the process of the terminal obtaining the weight corresponding to any frame of point cloud data in multiple frames of point cloud data according to the ratio of the point cloud attribute value corresponding to any frame of point cloud data to the sum of point cloud attribute values, the terminal can obtain the weight corresponding to any frame of point cloud data in multiple frames of point cloud data according to the ratio of the point cloud attribute value corresponding to any frame of point cloud data under each type of point cloud attribute to the sum of point cloud attribute values corresponding to the corresponding type of point cloud attribute. In this way, for each frame of point cloud data in multiple frames of point cloud data, based on the same method, the terminal can obtain the weight corresponding to each frame of point cloud data in multiple frames of point cloud data respectively.

[0090] The technical solution of this embodiment, when the point cloud attribute information includes point cloud attribute values corresponding to at least two types of point cloud attributes, by respectively obtaining the sum of the point cloud attribute values corresponding to each frame of point cloud data under each type of point cloud attribute to obtain the sum of point cloud attribute values corresponding to each type of point cloud attribute; and obtaining the weight corresponding to any frame of point cloud data in multiple frames of point cloud data according to the ratio of the point cloud attribute value corresponding to any frame of point cloud data under each type of point cloud attribute to the sum of point cloud attribute values corresponding to the corresponding type of point cloud attribute. In this way, for the point cloud attribute value of the point cloud data under each type of point cloud attribute, calculate its ratio to the sum of point cloud attribute values corresponding to the corresponding type of point cloud attribute, so as to realize the comprehensive point cloud attribute values corresponding to the point cloud data under multiple types of point cloud attributes, comprehensively and accurately evaluate the data reliability of the point cloud data, and then more accurately assign the weight corresponding to each frame of point cloud data in multiple frames of point cloud data respectively.

[0091] In one embodiment, according to the ratio of the point cloud attribute value corresponding to any frame of point cloud data under each point cloud attribute type to the sum of the point cloud attribute values corresponding to the corresponding point cloud attribute type, the weight corresponding to any frame of point cloud data in multiple frames of point cloud data is obtained, including: according to the ratio of the point cloud attribute value corresponding to any frame of point cloud data under each point cloud attribute type to the sum of the point cloud attribute values corresponding to the corresponding point cloud attribute type, the weight corresponding to any frame of point cloud data under each point cloud attribute type is obtained; according to the average value of the weights corresponding to any frame of point cloud data under each point cloud attribute type, the weight corresponding to any frame of point cloud data in multiple frames of point cloud data is obtained.

[0092] In a specific implementation, when the terminal obtains the weight corresponding to any frame of point cloud data in multiple frames of point cloud data according to the ratio of the point cloud attribute value corresponding to any frame of point cloud data under each point cloud attribute type to the sum of the point cloud attribute values corresponding to the corresponding point cloud attribute type, the terminal can obtain the weight corresponding to any frame of point cloud data under each point cloud attribute type according to the ratio of the point cloud attribute value corresponding to any frame of point cloud data under each point cloud attribute type to the sum of the point cloud attribute values corresponding to the corresponding point cloud attribute type. In this way, the terminal can obtain the weight corresponding to any frame of point cloud data in multiple frames of point cloud data according to the weight corresponding to any frame of point cloud data under each point cloud attribute type.

[0093] Furthermore, the terminal can obtain the weight corresponding to any frame of point cloud data in multiple frames of point cloud data based on the average value of the weights corresponding to any frame of point cloud data under each point cloud attribute type. In an optional embodiment of the present application, each point cloud attribute type may have a corresponding attribute weight, and the terminal can determine the weighted average value of the weights corresponding to any frame of point cloud data under each point cloud attribute type according to the attribute weight corresponding to each point cloud attribute type, so as to obtain the weight corresponding to any frame of point cloud data in multiple frames of point cloud data. In this way, for each frame of point cloud data in multiple frames of point cloud data, based on the same method, the terminal can obtain the weight corresponding to each frame of point cloud data in multiple frames of point cloud data respectively.

[0094] For example, continuing with the above example, is the weight corresponding to the i-th frame of point cloud data in consecutive T frames of point cloud data when the point cloud attribute value is the point cloud signal-to-noise ratio. That is, is the weight corresponding to the i-th frame of point cloud data under the point cloud attribute type of point cloud signal-to-noise ratio; is the weight corresponding to the i-th frame of point cloud data in consecutive T frames of point cloud data when the point cloud attribute value is the point cloud quantity value. That is, is the weight corresponding to the i-th frame of point cloud data under the point cloud attribute type of the number of point clouds. Thus, when the point cloud attribute types include the number of point clouds and the signal-to-noise ratio of the point cloud, the weight corresponding to the i-th frame of point cloud data in the multi-frame point cloud data (expressed in

[0095] ), is the average value of the weights corresponding to the i-th frame under the above two cloud attribute types. The calculation method is as follows:

[0096]

[0097] In the technical solution of this embodiment, by obtaining the ratio of the point cloud attribute value corresponding to any frame of point cloud data under each point cloud attribute type to the sum of the point cloud attribute values corresponding to the corresponding point cloud attribute type, the weight corresponding to any frame of point cloud data under each point cloud attribute type is obtained; according to the average value of the weights corresponding to any frame of point cloud data under each point cloud attribute type, the weight corresponding to any frame of point cloud data in the multi-frame point cloud data is obtained. Thus, for the point cloud attribute value of the point cloud data under each point cloud attribute type, calculating the ratio of it to the sum of the point cloud attribute values corresponding to the corresponding point cloud attribute type can obtain the weight corresponding to the point cloud data under each point cloud attribute type, so that the weights corresponding to the point cloud data under each point cloud attribute type can be integrated, and the weights corresponding to each frame of point cloud data in the multi-frame point cloud data can be more accurately assigned. It realizes setting the weights corresponding to the point cloud data more comprehensively and accurately based on multiple point cloud attribute types.

[0098] In another embodiment, extracting the target feature data corresponding to the fused point cloud data includes: obtaining the features corresponding to the fused point cloud data under the discriminative feature type to obtain the discriminative features corresponding to the fused point cloud data; obtaining the target feature data corresponding to the fused point cloud data according to the discriminative features corresponding to the fused point cloud data.

[0099] Among them, the features corresponding to different categories of targets under the discriminative feature type satisfy the preset difference conditions. That is, the features corresponding to the discriminative feature type are the features used to distinguish targets belonging to different categories.

[0100] In specific implementation, during the process of the terminal extracting the target feature data corresponding to the fused point cloud data, the terminal can obtain the features corresponding to the fused point cloud data under the discriminative feature type to obtain the discriminative features corresponding to the fused point cloud data, and obtain the target feature data corresponding to the fused point cloud data according to the discriminative features corresponding to the fused point cloud data.

[0101] The technical solution of this embodiment obtains the discriminative features corresponding to the fused point cloud data under the discriminative feature type, so as to obtain the discriminative features corresponding to the fused point cloud data. Among them, the features corresponding to different categories of targets under the discriminative feature type satisfy the preset difference conditions. According to the discriminative features corresponding to the fused point cloud data, the target feature data corresponding to the fused point cloud data is obtained. In this way, the discriminative features of the fused point cloud data are extracted to obtain the target feature data for classification detection, which can further improve the feature discrimination of the point cloud data corresponding to the target to be recognized and also improve the accuracy of target recognition.

[0102] In one embodiment, the method further includes: under each feature type, performing histogram analysis on the fused sample point cloud data corresponding to sample targets belonging to different categories to obtain the histograms corresponding to each category under each feature type; and determining the feature type whose difference satisfies the preset difference condition as the discriminative feature type according to the differences of the histograms corresponding to each category under each feature type.

[0103] Among them, the sample targets belonging to different categories may include, but are not limited to, targets such as adults, children, pets, and fans.

[0104] Among them, each feature type includes at least one feature type, and each feature type includes, but is not limited to, at least one feature type such as height feature, size feature, and SNR feature.

[0105] In specific implementation, the terminal can perform histogram visualization analysis on the fused sample point cloud data corresponding to sample targets belonging to different categories under each preset feature type to obtain the histograms corresponding to each category under each feature type. Thus, the terminal can determine the feature type that is discriminative for different categories according to the histograms corresponding to each category under each feature type, and obtain the discriminative feature type.

[0106] Among them, the method for obtaining the fused sample point cloud data corresponding to the sample target is the same as the principle of the method for obtaining the fused point cloud data corresponding to the target to be recognized described above, and will not be elaborated here. In practical applications, every 10 frames (which can also be other values, and no specific limitation is made here) of sample point cloud data can be continuously obtained and fused in each category of sample point cloud data to obtain the fused sample point cloud data corresponding to each category.

[0107] For the convenience of those skilled in the art to understand, FIG. 2(a) provides a schematic histogram of sample targets belonging to different categories under height features; FIG. 2(b) provides a schematic histogram of sample targets belonging to different categories under size features; FIG. 2(c) provides a schematic histogram of sample targets belonging to different categories under SNR features. Among them, the sample targets shown in FIG. 2(a), FIG. 2(b) and FIG. 2(c) are pets, children and adults respectively.

[0108] Among them, for the process of obtaining the histogram, since each point in the point cloud data contains information such as distance, x, y, z coordinates, speed, SNR, etc., taking the z coordinate (height information) as an example: after multi-frame accumulation, a large amount of z coordinate information can be obtained for each category, and then the histogram is calculated based on this z coordinate information (by the range of the maximum and minimum data of z, dividing the interval with this range, and then counting the number of times each z appears in this different interval to form the histogram). Similarly, the principle of obtaining the corresponding histogram under other feature types is similar and will not be elaborated here.

[0109] In this way, after obtaining the histograms corresponding to different categories under each feature type, the feature type whose difference satisfies the preset difference condition can be determined according to the differences between the histograms corresponding to different categories under each feature type, as the discriminative feature type. Specifically, for histograms belonging to different categories under the same feature type, if the difference in the data concentration regions of the histograms belonging to different categories under the same feature type is not significant (for example, the difference in the data concentration regions of the histograms belonging to different categories under the same feature type is less than the preset difference threshold), it means that this feature type is difficult to distinguish these categories; if the difference in the data concentration regions of the histograms belonging to different categories under the same feature type is relatively large (for example, the difference is greater than the preset difference threshold), it means that this feature type is discriminative and can be used as the discriminative feature type.

[0110] Exemplarily, it can be seen that in FIG. 2(a), FIG. 2(b) and FIG. 2(c), there are obvious differences in the data concentration regions of the histograms of adults, children, and pets under feature types such as height features, size features, and SNR features. Therefore, height features, size features, and SNR features can all be used as discriminative feature types. In an optional embodiment of the present application, the discriminative feature type may include at least one of height features, size features, and SNR features.

[0111] In the technical solution of this embodiment, by performing histogram analysis on the fused sample point cloud data corresponding to sample targets belonging to different categories under each feature type, histograms corresponding to each category under each feature type are obtained; according to the differences between the histograms corresponding to each category under each feature type, the feature types whose differences meet the preset difference conditions are determined as the discriminative feature types. In this way, the histograms of multiple feature types of sample targets of different categories are visualized using statistical analysis methods, and the differences and similarities between different feature types are compared, so that discriminative and differentiating feature types can be determined for classification and recognition, effectively improving the accuracy of target recognition.

[0112] In one embodiment, the discriminative feature includes a histogram feature; obtaining the feature corresponding to the fused point cloud data under the discriminative feature type to obtain the discriminative feature corresponding to the fused point cloud data includes: obtaining the histogram corresponding to the fused point cloud data under the discriminative feature type; according to the frequency information of the histogram, obtaining the histogram feature corresponding to the fused point cloud data under the discriminative feature type.

[0113] In specific implementation, after the terminal determines the discriminative feature type, the terminal can obtain the histogram corresponding to the fused point cloud data under the discriminative feature type; according to the frequency information of the histogram, obtain the histogram feature corresponding to the fused point cloud data under the discriminative feature type, and based on this histogram feature as the discriminative feature corresponding to the fused point cloud data.

[0114] Specifically, since the abscissa in the histogram represents different ranges and the ordinate represents the number of occurrences at different abscissa values, in practical applications, the array composed of the values of the ordinate corresponding to the abscissa can be directly used as the histogram feature.

[0115] It can be understood that in addition to obtaining the discriminative feature by using the histogram feature extraction method, after determining the discriminative feature type, the terminal can also perform feature extraction under the discriminative feature type through feature extraction methods such as deep convolutional neural networks and machine learning to obtain the discriminative feature corresponding to the fused point cloud data.

[0116] In the technical solution of this embodiment, the discriminative feature includes a histogram feature; by obtaining the histogram corresponding to the fused point cloud data under the discriminative feature type; according to the frequency information of the histogram, obtaining the histogram feature corresponding to the fused point cloud data under the discriminative feature type. In this way, according to the frequency information of the histogram corresponding to the fused point cloud data under the discriminative feature type, the histogram feature corresponding to the fused point cloud data under the discriminative feature type can be determined efficiently and accurately, so that the discriminative feature corresponding to the fused point cloud data can be obtained efficiently according to the histogram feature corresponding to the fused point cloud data under the discriminative feature type.

[0117] In one embodiment, in the presence of at least two types of discriminative features, obtaining target feature data corresponding to the fused point cloud data according to the discriminative features corresponding to the fused point cloud data includes: fusing the histogram features corresponding to the fused point cloud data under each type of discriminative feature in a cascaded manner to obtain the target feature data.

[0118] Among them, the target feature data includes a target feature vector.

[0119] In a specific implementation, in the presence of at least two types of discriminative features, if the discriminative feature uses a histogram feature, during the process of the terminal obtaining the target feature data corresponding to the fused point cloud data according to the discriminative features corresponding to the fused point cloud data, the terminal can fuse the histogram features corresponding to the fused point cloud data under each type of discriminative feature in a cascaded manner to obtain a target feature vector, and the target feature data can be obtained according to the target feature vector.

[0120] Further, during the process of the terminal fusing the histogram features corresponding to the fused point cloud data under each type of discriminative feature in a cascaded manner to obtain a target feature vector, the terminal can directly connect the histogram features corresponding to different types of discriminative features to obtain the target feature vector. The dimension of the target feature vector is the sum of the dimensions of each histogram feature. For example, if the types of discriminative features include height feature and SNR feature, and the histogram feature under the height feature is 20-dimensional and the histogram feature under the SNR feature is 30-dimensional, the dimension of the target feature vector obtained by the cascaded method is 20 + 30 = 50 dimensions.

[0121] The technical solution of this embodiment, in the presence of at least two types of discriminative features and the discriminative feature using a histogram feature, fuses the histogram features corresponding to the fused point cloud data under each type of discriminative feature in a cascaded manner to obtain the target feature data. In this way, by fusing the histogram features belonging to different types of discriminative features for target classification, the discriminability of the discriminative features used for classification and the accuracy of target classification can be improved.

[0122] For the convenience of those skilled in the art to understand, Figure 3 a general architecture diagram of a target detection method is provided. As Figure 3As shown, the terminal can obtain the point cloud data of consecutive T frames, determine the weights corresponding to each frame of point cloud data in the point cloud data of the consecutive T frames, perform weighted fusion on the point cloud data of the consecutive T frames according to the weights corresponding to each frame of point cloud data to obtain the fused point cloud data, extract the histogram features corresponding to each discriminative feature type of the fused point cloud data, fuse the histogram features corresponding to each discriminative feature type to obtain the target feature data including the target feature vector, and input it into the trained target classification model to obtain the classification detection result of the target to be recognized.

[0123] In an optional embodiment of the present application, the method further includes: collecting multiple frames of sample point cloud data; the sample point cloud data is obtained by collecting point cloud data of sample targets belonging to different categories; fusing the sample point cloud data in each category according to a preset number of frames to obtain multiple fused sample point cloud data in each category; obtaining the target sample feature data corresponding to each fused sample point cloud data; using the target sample feature data corresponding to each fused sample point cloud data and the corresponding sample category label as a sample data to train the target classification model to be trained, and obtaining the trained target classification model.

[0124] Among them, the number of sample point cloud data in each category satisfies the preset quantity condition. For example, the number of sample point cloud data in each category is greater than 10 (the preset quantity threshold can also adopt other values, which are not specifically limited here).

[0125] In specific implementation, during the process of the terminal training the target classification model, the terminal can obtain multiple frames of sample point cloud data obtained by collecting point cloud data of sample targets belonging to different categories, and fuse the sample point cloud data in each category according to a preset number of frames to obtain multiple fused sample point cloud data in each category. For example, the preset number of frames can be 10, and the terminal can fuse every 10 frames of sample point cloud data in the multiple frames of sample point cloud data in each category to obtain multiple fused sample point cloud data in each category.

[0126] Then, the terminal can obtain the target sample feature data corresponding to each fused sample point cloud data; use the target sample feature data corresponding to each fused sample point cloud data and the corresponding sample category label as a sample data, comprehensively all the sample data of each category, and use a classification algorithm to train the target classification model to be trained to obtain the trained target classification model.

[0127] Among them, the method for obtaining the target sample feature data corresponding to each fused sample point cloud data is the same as the principle of the method for obtaining the target feature data corresponding to the fused point cloud data described above. For example, in the case where the discriminative feature types include height feature, size feature, and SNR feature, the terminal can cascade the histogram features corresponding to the height feature, size feature, and SNR feature for each fused sample point cloud data to form a feature vector, which is used as the target sample feature vector corresponding to the fused sample point cloud data, so as to obtain the target sample feature data.

[0128] In practical applications, the classification algorithm can be but is not limited to algorithms such as SVM (Support Vector Machine), convolutional neural network, and logistic regression.

[0129] In another embodiment, as Figure 4 shown, a target detection method is provided. Taking the application of this method to a terminal as an example, it includes the following steps:

[0130] Step S402: Obtain multiple frames of point cloud data obtained by scanning the target to be recognized in the target scene.

[0131] Step S404: When the point cloud attribute information includes point cloud attribute values corresponding to at least two point cloud attribute types, respectively obtain the sum of the point cloud attribute values corresponding to each frame of point cloud data under each point cloud attribute type, and obtain the point cloud attribute value sum corresponding to each point cloud attribute type.

[0132] Step S406: According to the ratio of the point cloud attribute value corresponding to any frame of point cloud data under each point cloud attribute type to the point cloud attribute value sum corresponding to the corresponding point cloud attribute type, obtain the weight corresponding to any frame of point cloud data under each point cloud attribute type.

[0133] Step S408: According to the average value of the weights corresponding to any frame of point cloud data under each point cloud attribute type, obtain the weight corresponding to any frame of point cloud data in the multiple frames of point cloud data.

[0134] Step S410: According to the weights corresponding to each frame of point cloud data, perform weighted fusion on the multiple frames of point cloud data to obtain fused point cloud data.

[0135] Step S412: Obtain the histogram corresponding to the fused point cloud data under the discriminative feature type.

[0136] Step S414: According to the frequency information of the histogram, obtain the histogram feature corresponding to the fused point cloud data under the discriminative feature type.

[0137] Step S416: In the case where there are at least two types of discriminative features, the histogram features corresponding to the fused point cloud data under each type of discriminative feature are fused in a cascaded manner to obtain target feature data.

[0138] Step S418: According to the target feature data, obtain the classification and detection result of the target to be recognized.

[0139] It should be noted that the specific limitations of the above steps can refer to the specific limitations of a target detection method described above.

[0140] It should be understood that although the steps in the flowcharts involved in the above-mentioned embodiments are shown in sequence according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless there is a clear description in this article, the execution of these steps has no strict order limitation, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments may include multiple steps or multiple stages. These steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be executed alternately or alternately with at least a part of other steps or steps or stages in other steps.

[0141] Based on the same inventive concept, the embodiments of the present application also provide a target detection device for implementing the above-mentioned target detection method. The implementation solutions provided by this device to solve problems are similar to the implementation solutions described in the above method. Therefore, the specific limitations in one or more embodiments of the target detection device provided below can refer to the limitations of the target detection method in the above text, and will not be repeated here.

[0142] In an exemplary embodiment, as Figure 5 shown, a target detection device is provided, including: an acquisition module 510, a determination module 520, a fusion module 530, and an identification module 540, where:

[0143] The acquisition module 510 is configured to acquire multiple frames of point cloud data obtained by scanning a target to be recognized in a target scene.

[0144] The determination module 520 is configured to determine the weights corresponding to each frame of the point cloud data in the multiple frames of point cloud data according to the point cloud attribute information of each frame of the point cloud data; wherein, the weight corresponding to the point cloud data is positively correlated with the data reliability corresponding to the point cloud data.

[0145] The fusion module 530 is configured to perform weighted fusion on the multiple frames of point cloud data according to the weights corresponding to each frame of the point cloud data to obtain fused point cloud data.

[0146] An identification module 540, configured to extract target feature data corresponding to the fused point cloud data, and obtain a classification detection result of the target to be identified according to the target feature data.

[0147] In one embodiment, the point cloud attribute information includes point cloud attribute values; the determination module 520 is specifically configured to obtain the sum of the point cloud attribute values corresponding to each frame of the point cloud data to obtain a point cloud attribute value sum; for any frame of the point cloud data among the multiple frames of point cloud data, according to the ratio of the point cloud attribute value corresponding to the any frame of point cloud data to the point cloud attribute value sum, obtain the weight corresponding to the any frame of point cloud data in the multiple frames of point cloud data.

[0148] In one embodiment, when the point cloud attribute information includes point cloud attribute values corresponding to at least two types of point cloud attribute types, the determination module 520 is specifically configured to respectively obtain the sum of the point cloud attribute values corresponding to each frame of the point cloud data under each point cloud attribute type to obtain the point cloud attribute value sums corresponding to each point cloud attribute type; according to the ratio of the point cloud attribute value corresponding to the any frame of point cloud data under each point cloud attribute type to the corresponding point cloud attribute value sum, obtain the weight corresponding to the any frame of point cloud data in the multiple frames of point cloud data.

[0149] In one embodiment, the determination module 520 is specifically configured to obtain the weight corresponding to the any frame of point cloud data under each point cloud attribute type according to the ratio of the point cloud attribute value corresponding to the any frame of point cloud data under each point cloud attribute type to the corresponding point cloud attribute value sum; according to the average value of the weights corresponding to the any frame of point cloud data under each point cloud attribute type, obtain the weight corresponding to the any frame of point cloud data in the multiple frames of point cloud data.

[0150] In one embodiment, the point cloud attribute value includes at least one of a point cloud signal-to-noise ratio value and a point cloud quantity value.

[0151] In one embodiment, the identification module 540 is specifically configured to obtain the features corresponding to the fused point cloud data under the discriminative feature type to obtain the discriminative features corresponding to the fused point cloud data; wherein, the features corresponding to different categories of targets under the discriminative feature type satisfy a preset difference condition; according to the discriminative features corresponding to the fused point cloud data, obtain the target feature data corresponding to the fused point cloud data.

[0152] In one embodiment, the device further includes: a feature type determination module, configured to perform histogram analysis on the fused sample point cloud data corresponding to sample targets belonging to different categories under each feature type, to obtain histograms corresponding to each of the categories under each of the feature types; and determine, according to differences between the histograms corresponding to each of the categories under each of the feature types, a feature type whose differences satisfy a preset difference condition as the discriminative feature type.

[0153] In one embodiment, the discriminative feature includes a histogram feature; the recognition module 540 is specifically configured to obtain a histogram corresponding to the fused point cloud data under the discriminative feature type; and obtain a histogram feature corresponding to the fused point cloud data under the discriminative feature type according to frequency information of the histogram.

[0154] In one embodiment, when there are at least two discriminative feature types, the recognition module 540 is specifically configured to fuse histogram features corresponding to the fused point cloud data under each of the discriminative feature types in a cascaded manner to obtain the target feature data.

[0155] Each module in the above target detection device can be implemented in whole or in part by software, hardware, and a combination thereof. Each of the above modules can be embedded in a processor in a computer device in a hardware form or be independent of the processor, or can be stored in a memory in the computer device in a software form, so that the processor can call and execute operations corresponding to each of the above modules.

[0156] In an exemplary embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as Figure 6As shown in the figure. The computer device includes a processor, a memory, an input / output interface, a communication interface, a display unit, and an input device. Among them, the processor, the memory, and the input / output interface are connected through a system bus, and the communication interface, the display unit, and the input device are connected to the system bus through the input / output interface. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and external devices. The communication interface of the computer device is used to communicate with external terminals in a wired or wireless manner, and the wireless manner can be achieved through WIFI, a mobile cellular network, NFC (Near Field Communication), or other technologies. The computer program, when executed by the processor, implements a target detection method. The display unit of the computer device is used to form a visually visible picture, which can be a display screen, a projection device, or a virtual reality imaging device. The display screen can be a liquid crystal display screen or an electronic ink display screen. The input device of the computer device can be a touch layer covering the display screen, or a button, a trackball, or a touchpad provided on the housing of the computer device, or an external keyboard, touchpad, or mouse, etc.

[0157] Those skilled in the art can understand that Figure 6 the structure shown in the figure is only a block diagram of some structures related to the solution of this application, and does not constitute a limitation on the computer device to which the solution of this application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine some components, or have different component arrangements.

[0158] In one embodiment, a computer device is further provided, including a memory and a processor. A computer program is stored in the memory, and when the processor executes the computer program, the steps in the above method embodiments are implemented.

[0159] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by the processor, the steps in the above method embodiments are implemented.

[0160] In one embodiment, a computer program product is provided, including a computer program. When the computer program is executed by the processor, the steps in the above method embodiments are implemented.

[0161] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in this application are all information and data that have been authorized by the user or fully authorized by all parties, and the collection, use, and processing of relevant data need to comply with relevant regulations.

[0162] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, database, or other medium used in the various embodiments provided in this application can include at least one of non-volatile and volatile memories. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc. The databases involved in the various embodiments provided in this application can include at least one of relational databases and non-relational databases. Non-relational databases can include distributed databases based on blockchain, etc., without limitation. The processors involved in the various embodiments provided in this application can be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, data processing logics based on quantum computing, etc., without limitation.

[0163] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity of description, not all possible combinations of the various technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope described in this specification.

[0164] The above-described embodiments merely represent several implementation manners of the present application. The description thereof is relatively specific and detailed, but it should not be construed as a limitation on the patent scope of the present application. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present application, several modifications and improvements can still be made, and these all fall within the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the appended claims.

Claims

1. A target detection method, characterized in that, The method includes: Obtaining multiple frames of point cloud data obtained by scanning a target to be recognized in a target scene; Determining the weights corresponding to each frame of the point cloud data in the multiple frames of point cloud data according to the point cloud attribute information of each frame of the point cloud data; wherein, the weight corresponding to the point cloud data is positively correlated with the data reliability corresponding to the point cloud data; Performing weighted fusion on the multiple frames of point cloud data according to the weights corresponding to each frame of the point cloud data to obtain fused point cloud data; Extracting target feature data corresponding to the fused point cloud data, and obtaining a classification detection result of the target to be recognized according to the target feature data.

2. The method according to claim 1, wherein The point cloud attribute information includes point cloud attribute values; the determining the weights corresponding to each frame of the point cloud data in the multiple frames of point cloud data according to the point cloud attribute information of each frame of the point cloud data includes: Obtaining the sum of the point cloud attribute values corresponding to each frame of the point cloud data to obtain a sum of point cloud attribute values; For any frame of point cloud data in the multiple frames of point cloud data, obtaining the weight corresponding to the any frame of point cloud data in the multiple frames of point cloud data according to the ratio of the point cloud attribute value corresponding to the any frame of point cloud data to the sum of point cloud attribute values.

3. The method according to claim 2, wherein When the point cloud attribute information includes point cloud attribute values corresponding to at least two types of point cloud attributes, the obtaining the sum of the point cloud attribute values corresponding to each frame of the point cloud data to obtain a sum of point cloud attribute values includes: Respectively obtaining the sum of the point cloud attribute values corresponding to each frame of the point cloud data under each type of point cloud attribute to obtain the sum of point cloud attribute values corresponding to each type of point cloud attribute; The obtaining the weight corresponding to the any frame of point cloud data in the multiple frames of point cloud data according to the ratio of the point cloud attribute value corresponding to the any frame of point cloud data to the sum of point cloud attribute values includes: Obtaining the weight corresponding to the any frame of point cloud data in the multiple frames of point cloud data according to the ratio of the point cloud attribute value corresponding to the any frame of point cloud data under each type of point cloud attribute to the sum of point cloud attribute values corresponding to the corresponding type of point cloud attribute.

4. The method according to claim 3, wherein The obtaining the weight corresponding to the any frame of point cloud data in the multiple frames of point cloud data according to the ratio of the point cloud attribute value corresponding to the any frame of point cloud data under each type of point cloud attribute to the sum of point cloud attribute values corresponding to the corresponding type of point cloud attribute includes: Obtaining the weight corresponding to the any frame of point cloud data under each type of point cloud attribute according to the ratio of the point cloud attribute value corresponding to the any frame of point cloud data under each type of point cloud attribute to the sum of point cloud attribute values corresponding to the corresponding type of point cloud attribute; Obtaining the weight corresponding to the any frame of point cloud data in the multiple frames of point cloud data according to the average value of the weights corresponding to the any frame of point cloud data under each type of point cloud attribute.

5. The method according to claim 2, wherein The point cloud attribute value includes at least one of a point cloud signal-to-noise ratio value and a point cloud quantity value.

6. The method according to any one of claims 1-5, characterized in that, The extracting the target feature data corresponding to the fused point cloud data includes: Obtain the features corresponding to the fused point cloud data under the discriminative feature type to obtain the discriminative features corresponding to the fused point cloud data; wherein, the features corresponding to different categories of targets under the discriminative feature type satisfy preset difference conditions; Obtain the target feature data corresponding to the fused point cloud data according to the discriminative features corresponding to the fused point cloud data.

7. The method according to claim 6, wherein The method further includes: Under each feature type, perform histogram analysis on the fused sample point cloud data corresponding to sample targets of different categories to obtain the histograms corresponding to each category under each feature type; Determine the feature type whose difference satisfies the preset difference condition according to the differences between the histograms corresponding to each category under each feature type as the discriminative feature type.

8. The method according to claim 6, wherein The discriminative features include histogram features; The obtaining the features corresponding to the fused point cloud data under the discriminative feature type to obtain the discriminative features corresponding to the fused point cloud data includes: Obtain the histogram corresponding to the fused point cloud data under the discriminative feature type; Obtain the histogram features corresponding to the fused point cloud data under the discriminative feature type according to the frequency information of the histogram.

9. The method according to claim 8, wherein In the case where there are at least two discriminative feature types, the obtaining the target feature data corresponding to the fused point cloud data according to the discriminative features corresponding to the fused point cloud data includes: Fuse the histogram features corresponding to the fused point cloud data under each discriminative feature type in a cascaded manner to obtain the target feature data.

10. A target detection device, characterized in that, The device includes: An acquisition module, configured to acquire multiple frames of point cloud data obtained by scanning a target to be recognized in a target scene; A determination module, configured to determine the weights corresponding to each frame of the point cloud data in the multiple frames of point cloud data according to the point cloud attribute information of each frame of the point cloud data; wherein, the weight corresponding to the point cloud data is positively correlated with the data reliability corresponding to the point cloud data; A fusion module, configured to perform weighted fusion on the multiple frames of point cloud data according to the weights corresponding to each frame of the point cloud data to obtain fused point cloud data; An identification module, configured to extract the target feature data corresponding to the fused point cloud data and obtain a classification detection result of the target to be recognized according to the target feature data.

11. A computer device, comprising a memory and a processor, the memory storing a computer program, characterized in that, When the processor executes the computer program, the steps of the method according to any one of claims 1 to 9 are implemented.

12. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, the steps of the method according to any one of claims 1 to 9 are implemented.

Images