Rod testing methods, devices and electronic equipment

Abstract

This disclosure relates to a method, apparatus, and electronic device for detecting poles. The method includes: acquiring a target image and a pole prediction result for the target image; the pole prediction result includes predicted coordinate information of two endpoints of the pole in the target image and a confidence score; determining the predicted height of the pole based on the predicted coordinate information of the two endpoints; determining the height range to which the pole belongs based on the predicted height and a pole height discrimination threshold; and determining the pole detection result of the target image based on the predicted coordinate information of the two endpoints, the confidence score, and the confidence threshold corresponding to the height range to which the pole belongs. Different height ranges correspond to different confidence thresholds, thereby setting different confidence thresholds so that a higher confidence threshold for a higher height range can favor the detection of taller poles, ensuring a higher detection rate for taller poles, and thus improving mapping and localization efficiency.

Description

Technical Field

[0001] This disclosure relates to the fields of autonomous driving and intelligent sensing technology, and in particular to a method, apparatus and electronic device for detecting poles. Background Technology

[0002] Currently, in autonomous driving systems, pole detection algorithms are key algorithms for visual perception-assisted localization. Their goal is to locate poles near the road, such as utility poles, light poles, and billboard poles, to assist autonomous vehicles in mapping and localization. In real-world autonomous driving scenarios, due to the diversity of road pole types, images captured by cameras typically contain multiple types of poles. During mapping and localization, the focus is usually on the taller poles among these, meaning a higher detection rate is required for taller poles.

[0003] Current pole detection algorithms use a pole detection model to obtain detection results for poles of all sizes in test images within an image test set. Based on these results, a general confidence threshold is determined, and the pole detection results for images captured by the camera are then used. However, the number of shorter poles in the test images far exceeds the number of taller poles. The general confidence threshold favors the detection of shorter poles, resulting in a poor detection rate for taller poles and consequently, poor mapping and localization efficiency. Summary of the Invention

[0004] This disclosure provides a method, apparatus, and electronic device for testing rods.

[0005] According to a first aspect of the present disclosure, a pole detection method is provided. The method includes: acquiring a target image and a pole prediction result of the target image; the pole prediction result includes predicted coordinate information of two endpoints of a pole in the target image and a confidence score of the pole; determining the predicted height of the pole based on the predicted coordinate information of the two endpoints of the pole; determining the height interval to which the pole belongs based on the predicted height of the pole and a pole height differentiation threshold; and determining the pole detection result of the target image based on the predicted coordinate information of the two endpoints of the pole, the confidence score, and the confidence threshold corresponding to the height interval to which the pole belongs; wherein different height intervals correspond to different confidence thresholds.

[0006] In one embodiment of this disclosure, the number of the member height differentiation thresholds is at least one; when the number of member height differentiation thresholds is one, the number of height intervals is two, namely a large member height interval and a small member height interval; the height in the large member height interval is greater than or equal to the member height differentiation threshold; the height in the small member height interval is less than the member height differentiation threshold; the confidence threshold corresponding to the large member height interval is less than the confidence threshold corresponding to the small member height interval.

[0007] In one embodiment of this disclosure, determining the pole detection result of the target image based on the predicted coordinate information of the two endpoints of the pole, the confidence score, and the confidence threshold corresponding to the height interval to which the pole belongs includes: identifying the pole as a target pole if the confidence score of the pole is greater than or equal to the confidence threshold corresponding to the height interval to which the pole belongs; and determining the pole detection result of the target image based on the predicted coordinate information of the two endpoints of the target pole.

[0008] In one embodiment of this disclosure, the method further includes: acquiring an image test set and pole prediction results for test images in the image test set; the image test set includes multiple test images and pole annotation results for the test images; the pole prediction results for the test images include predicted coordinate information of two endpoints of predicted poles in the test images and a confidence score of the predicted poles; the pole annotation results for the test images include labeled coordinate information of two endpoints of labeled poles in the test images; determining a pole height discrimination threshold based on the two endpoints of the labeled poles; determining at least two height intervals divided based on the two endpoints of the labeled poles and the pole height discrimination threshold, and a set of labeled poles corresponding to at least two height intervals; determining a set of predicted poles corresponding to at least two height intervals based on the predicted coordinate information of two endpoints of the predicted poles and the confidence scores of the predicted poles; and for each height interval, determining a confidence threshold corresponding to the height interval based on the set of labeled poles and the set of predicted poles corresponding to the height interval.

[0009] In one embodiment of this disclosure, determining the height distinction threshold of the multiple labeled rods based on the coordinate information of the two endpoints of the labeled rods includes: determining the labeled height of the multiple labeled rods based on the coordinate information of the two endpoints of the multiple labeled rods; performing clustering processing on the multiple labeled rods based on the labeled height of the multiple labeled rods to obtain at least two clusters; for each cluster, determining the average labeled height of the cluster based on the labeled height of the labeled rods in the cluster; and determining the height distinction threshold of the rods based on the average labeled height of two adjacent clusters in the at least two clusters.

[0010] In one embodiment of this disclosure, determining at least two height intervals and a set of labeled rods corresponding to at least two height intervals based on the coordinate information of the two endpoints of the plurality of labeled rods and the height differentiation threshold of the rods includes: dividing the height range based on the height differentiation threshold to obtain at least two height intervals; for each labeled rod, determining the height interval to which the labeled rod belongs based on the coordinate information of the two endpoints of the labeled rod; and for each height interval, determining the set of labeled rods corresponding to the height interval based on the coordinate information of the two endpoints of the labeled rods belonging to the height interval.

[0011] In one embodiment of this disclosure, determining the set of predicted poles corresponding to at least two height intervals based on the predicted coordinate information of the two endpoints of the plurality of predicted poles and the confidence scores of the plurality of predicted poles includes: for each predicted pole, determining the height interval to which the predicted pole belongs based on the predicted coordinate information of the two endpoints of the predicted pole; for each height interval, determining the set of predicted poles corresponding to the height interval based on the predicted coordinate information of the two endpoints of the predicted poles belonging to the height interval and the confidence scores of the predicted poles belonging to the height interval.

[0012] In one embodiment of this disclosure, determining the confidence threshold corresponding to each height interval based on the set of labeled poles and the set of predicted poles for that height interval includes: setting an initial confidence threshold for each height interval; filtering the set of predicted poles corresponding to the height interval based on the initial confidence threshold and the confidence scores of the predicted poles in the set of predicted poles corresponding to the height interval to obtain a filtered set of predicted poles; obtaining multiple target labeled poles in the set of labeled poles corresponding to the height interval, and multiple target predicted poles in the filtered set of predicted poles corresponding to the height interval; and so on. Based on the coordinate information of the two endpoints of multiple target labeled poles and the coordinate information of the two endpoints of multiple target predicted poles, the multiple target labeled poles and multiple target predicted poles are paired to obtain multiple sets of paired poles and at least one unpaired target labeled pole. Pole indicators are determined based on the number of paired pole groups, the number of unpaired labeled poles, the total number of target labeled poles, and the total number of target predicted poles. If the pole indicator does not meet the pole indicator conditions, the initial confidence threshold is adjusted until the pole indicator determined based on the adjusted confidence threshold meets the pole indicator conditions.

[0013] In one embodiment of this disclosure, the step of filtering the predicted pole set corresponding to the height interval based on the initial confidence threshold and the confidence scores of the predicted poles in the predicted pole set corresponding to the height interval to obtain a filtered predicted pole set includes: sequentially filtering each predicted pole in the predicted pole set corresponding to the height interval if the confidence score of the predicted pole is less than the initial confidence threshold, thereby obtaining a filtered predicted pole set.

[0014] In one embodiment of this disclosure, the step of pairing multiple target labeled rods and multiple target predicted rods based on the coordinate information of the two endpoints of multiple target labeled rods and the coordinate information of the two endpoints of multiple target predicted rods to obtain multiple sets of paired rods and at least one unpaired target labeled rod includes: sequentially, for each target labeled rod, determining the distance between the target labeled rod and the multiple target predicted rods based on the coordinate information of the two endpoints of the target labeled rod and the coordinate information of the two endpoints of the multiple target predicted rods; if the minimum distance among the multiple distances is less than or equal to a preset distance threshold, determining the target predicted rod corresponding to the minimum distance as a target predicted rod paired with the target labeled rod; if the minimum distance among the multiple distances is greater than the preset distance threshold, determining the target labeled rod as an unpaired target labeled rod.

[0015] In one embodiment of this disclosure, the lever metrics include lever recall and lever precision. Determining the lever metrics based on the number of paired lever groups, the number of unpaired labeled levers, the total number of target labeled levers, and the total number of target predicted levers includes: determining the lever recall as the ratio of the number of groups to the total number of target labeled levers; determining the difference between the total number of target predicted levers and the number of groups; and determining the lever precision as the ratio of the difference to the total number of target predicted levers.

[0016] In one embodiment of this disclosure, the pole metrics include: pole recall rate and pole precision rate; the pole metric condition is that the pole recall rate is greater than or equal to the recall rate threshold corresponding to the height interval, and the pole precision rate is greater than or equal to the precision rate threshold corresponding to the height interval.

[0017] According to a second aspect of the present disclosure, a pole detection device is also provided. The device includes: a first acquisition module, configured to acquire a target image and a pole prediction result of the target image; the pole prediction result includes predicted coordinate information of two endpoints of a pole in the target image and a confidence score of the pole; a first determination module, configured to determine the predicted height of the pole based on the predicted coordinate information of the two endpoints of the pole; a second determination module, configured to determine the height range to which the pole belongs based on the predicted height of the pole and a pole height differentiation threshold; and a third determination module, configured to determine the pole detection result of the target image based on the predicted coordinate information of the two endpoints of the pole, the confidence score, and a confidence threshold corresponding to the height range to which the pole belongs; wherein different height ranges correspond to different confidence thresholds.

[0018] According to a third aspect of the present disclosure, an electronic device is also provided, comprising: a processor; a memory for storing processor-executable instructions; wherein the processor is configured to: implement the steps of the rod detection method as described above.

[0019] According to a fourth aspect of the present disclosure, a vehicle is also provided, including the electronic device described above, or the vehicle is connected to the electronic device described above.

[0020] According to a fifth aspect of the present disclosure, a non-transitory computer-readable storage medium is also provided, which, when instructions in the storage medium are executed by a processor, enables the processor to perform the rod detection method as described above.

[0021] The technical solutions provided by the embodiments of this disclosure have at least the following beneficial effects:

[0022] The process involves acquiring a target image and predicting the poles within it. The prediction results include the predicted coordinates of the two endpoints of each pole and its confidence score. Based on these coordinates, the predicted height of the pole is determined. Then, based on the predicted height and a height discrimination threshold, the height range to which the pole belongs is determined. Finally, based on the predicted coordinates, confidence score, and the confidence threshold corresponding to the height range, the pole detection result for the target image is determined. Different height ranges correspond to different confidence thresholds. By setting different confidence thresholds, higher height ranges are biased towards detecting taller poles, thus ensuring a higher detection rate for taller poles and improving mapping and localization efficiency.

[0023] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit this disclosure. Attached Figure Description

[0024] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this disclosure and, together with the description, serve to explain the principles of this disclosure, and are not intended to unduly limit this disclosure.

[0025] Figure 1 This is a flowchart of a rod detection method according to an embodiment of the present disclosure;

[0026] Figure 2 This is a flowchart of a rod detection method according to another embodiment of the present disclosure;

[0027] Figure 3This is a schematic diagram illustrating the determination of the confidence thresholds for the height range of large members and the height range of small members.

[0028] Figure 4 A schematic diagram for determining the rod detection results of the target image;

[0029] Figure 5 This is a schematic diagram of the structure of a rod detection device according to an embodiment of the present disclosure;

[0030] Figure 6 This is a structural block diagram of an electronic device according to an exemplary embodiment of the present disclosure;

[0031] Figure 7 This is a block diagram illustrating a vehicle according to an exemplary embodiment of the present disclosure. Detailed Implementation

[0032] To enable those skilled in the art to better understand the technical solutions of this disclosure, the technical solutions in the embodiments of this disclosure will be clearly and completely described below with reference to the accompanying drawings.

[0033] It should be noted that the terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this disclosure are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this disclosure described herein can be implemented in orders other than those illustrated or described herein. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this disclosure. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this disclosure as detailed in the appended claims.

[0034] Currently, in autonomous driving systems, pole detection algorithms are key algorithms for visual perception-assisted localization. Their goal is to locate poles near the road, such as utility poles, light poles, and billboard poles, to assist autonomous vehicles in mapping and localization. In real-world autonomous driving scenarios, due to the diversity of road pole types, images captured by cameras typically contain multiple types of poles. During mapping and localization, the focus is usually on the taller poles among these, meaning a higher detection rate is required for taller poles.

[0035] Current pole detection algorithms use a pole detection model to obtain detection results for poles of all sizes in test images within an image test set. Based on these results, a general confidence threshold is determined, and the pole detection results for images captured by the camera are then used. However, the number of shorter poles in the test images far exceeds the number of taller poles. The general confidence threshold favors the detection of shorter poles, resulting in a poor detection rate for taller poles and consequently, poor mapping and localization efficiency.

[0036] Figure 1 This is a flowchart of a member detection method according to an embodiment of the present disclosure. It should be noted that the member detection method of this embodiment can be applied to a member detection device, which can be configured in an electronic device to enable the electronic device to perform member detection functions.

[0037] The electronic device can be any device with computing capabilities, such as a personal computer (PC), mobile terminal, server, etc. A mobile terminal can be a hardware device with various operating systems, touchscreens, and / or displays, such as an in-vehicle device, mobile phone, tablet computer, personal digital assistant, wearable device, etc. The following embodiments use an electronic device as an example for illustration.

[0038] like Figure 1 As shown, the method includes the following steps:

[0039] Step 101: Obtain the target image and the prediction results of the poles in the target image; the prediction results include the predicted coordinate information of the two endpoints of the poles in the target image and the confidence score of the poles.

[0040] In this embodiment of the disclosure, the target image can be an image captured by a camera on the vehicle. The process by which the electronic device obtains the pole prediction result of the target image can, for example, involve inputting the target image into a trained pole detection model and obtaining the pole prediction result of the target image output by the pole detection model.

[0041] In this embodiment of the disclosure, the target image may include at least one rod. Correspondingly, the rod prediction result of the target image may include the predicted coordinate information of the two endpoints of the at least one rod, and the confidence score of the at least one rod.

[0042] The predicted coordinate information of the two endpoints of the rod can be the predicted coordinate information of the two endpoints of the rod in the target image. For example, the predicted coordinate information of the two endpoints of the rod in the target image can be (x1, y1) and (x2, y2) respectively.

[0043] Step 102: Determine the predicted height of the rod based on the predicted coordinate information of the two endpoints of the rod.

[0044] In this embodiment of the disclosure, for each rod in the target image, the predicted height of the rod can be the height difference between two endpoints, for example, the absolute value of the difference between y2 and y1.

[0045] Step 103: Determine the height range to which the rod belongs based on the predicted height of the rod and the rod height differentiation threshold.

[0046] In this embodiment of the disclosure, the number of pole height differentiation thresholds is at least one, and the number of height intervals can be at least two; the height intervals are obtained based on the pole height differentiation thresholds. The number of height intervals can be the sum of the number of pole height differentiation thresholds and 1.

[0047] For example, when there is only one pole height differentiation threshold, there are two height intervals. The common boundary of the two height intervals is the pole height differentiation threshold. Alternatively, when there are two pole height differentiation thresholds, there can be three height intervals. The two pole height differentiation thresholds can serve as interval boundaries, resulting in one height interval; the height range above the higher pole height differentiation threshold can be considered as one height interval; and the height range below the lower pole height differentiation threshold can also be considered as one height interval.

[0048] When there is one threshold for distinguishing the height of a pole, there are two height intervals: a large pole height interval and a small pole height interval. The height in the large pole height interval is greater than or equal to the pole height distinguishing threshold. The height in the small pole height interval is less than the pole height distinguishing threshold. The confidence threshold corresponding to the large pole height interval is less than the confidence threshold corresponding to the small pole height interval.

[0049] In autonomous vehicle mapping and localization, the focus is typically on taller structural members (e.g., large members) among various structural members. A higher detection rate is required for these taller members. Therefore, the confidence threshold for the height range containing these taller members needs to be set relatively low to ensure a high detection rate. Conversely, for shorter members, higher accuracy is generally required. Therefore, the confidence threshold for the height range containing these shorter members can be set relatively high to ensure high accuracy.

[0050] In this embodiment of the disclosure, the electronic device may perform step 103 as follows: determine at least two height intervals based on the rod height differentiation threshold; determine the height interval containing the predicted height of the rod according to the predicted height of the rod; and determine the height interval containing the predicted height as the height interval to which the rod belongs.

[0051] Step 104: Determine the rod detection result of the target image based on the predicted coordinate information of the two endpoints of the rod, the confidence score, and the confidence threshold corresponding to the height interval to which the rod belongs; where different height intervals correspond to different confidence thresholds.

[0052] In this embodiment of the disclosure, the electronic device may perform step 104 as follows: if the confidence score of the pole is greater than or equal to the confidence threshold corresponding to the height interval to which the pole belongs, the pole is identified as a target pole; and the pole detection result of the target image is determined based on the predicted coordinate information of the two endpoints of the target pole.

[0053] Specifically, if the confidence score of a member is less than the confidence threshold corresponding to the height range to which the member belongs, the member will no longer be considered as a target member.

[0054] The target image may include at least one rod. Correspondingly, the electronic device may perform the operation of step 104 for each rod in the target image, thereby filtering out the target rod from the at least one rod in the target image; and combine the predicted coordinate information of the two endpoints of the at least one target rod to obtain the rod detection result of the target image.

[0055] In the pole detection method of this embodiment, a target image and pole prediction results of the target image are acquired. The pole prediction results include the predicted coordinate information of the two endpoints of the pole in the target image and the confidence score of the pole. The predicted height of the pole is determined based on the predicted coordinate information of the two endpoints of the pole. The height interval to which the pole belongs is determined based on the predicted height of the pole and the pole height discrimination threshold. The pole detection result of the target image is determined based on the predicted coordinate information of the two endpoints of the pole, the confidence score, and the confidence threshold corresponding to the height interval to which the pole belongs. Different height intervals correspond to different confidence thresholds. By setting different confidence thresholds, the confidence threshold corresponding to higher height intervals can be biased towards the detection of higher poles, thereby ensuring the detection rate of higher poles and improving mapping and localization efficiency.

[0056] Figure 2This is a flowchart illustrating a member detection method according to another embodiment of the present disclosure. It should be noted that the member detection method of this embodiment can be applied to a member detection device, which can be configured in an electronic device to enable the electronic device to perform member detection functions.

[0057] The electronic device can be any device with computing capabilities, such as a personal computer (PC), mobile terminal, server, etc. A mobile terminal can be a hardware device with various operating systems, touchscreens, and / or displays, such as an in-vehicle device, mobile phone, tablet computer, personal digital assistant, wearable device, etc. The following embodiments use an electronic device as an example for illustration.

[0058] like Figure 2 As shown, the method includes the following steps:

[0059] Step 201: Obtain an image test set and the pole prediction results of the test images in the image test set; the image test set includes multiple test images and the pole annotation results of the test images; the pole prediction results of the test images include the predicted coordinate information of the two endpoints of the predicted pole in the test image and the confidence score of the predicted pole; the pole annotation results of the test images include the labeled coordinate information of the two endpoints of the labeled pole in the test image.

[0060] In this embodiment of the disclosure, the pole prediction results of the test images in the image test set are determined by combining them with a trained pole detection model. The process by which the electronic device determines the pole prediction results of the test images in the image test set can, for example, involve inputting the test images from the image test set into the trained pole detection model and obtaining the pole prediction results of the test images output by the pole detection model.

[0061] The test image may contain at least one predicted member and at least one labeled member. The number of predicted members and the number of labeled members in the test image may be the same or different.

[0062] Step 202: Determine the height distinction threshold of the rods based on the coordinate information of the two endpoints of the multiple labeled rods.

[0063] In this embodiment of the disclosure, the electronic device may perform step 202 as follows: determine the annotation height of the multiple annotation rods based on the annotation coordinate information of the two endpoints of the multiple annotation rods; perform clustering processing on the multiple annotation rods based on the annotation height of the multiple annotation rods to obtain at least two clusters; for each cluster, determine the average annotation height of the cluster based on the annotation height of the annotation rods in the cluster; and determine the rod height differentiation threshold based on the average annotation height of two adjacent clusters in the at least two clusters.

[0064] The height of the labeled rod can be the height difference between the two endpoints of the labeled rod.

[0065] The electronic equipment performs clustering processing on multiple labeled rods. The clustering algorithm used, such as kmeans or kmeans++, can be set according to actual needs.

[0066] The process of determining the clustering of multiple labeled rods by electronic devices can be, for example, assuming the heights of multiple labeled rods can be denoted as {h1, h2, ..., hn}; where n represents the number of labeled rods; (1) randomly select a height from multiple heights as the first cluster center; (2) calculate the distance from all heights to their nearest cluster center, where the distance is the height difference; (3) the probability that all non-cluster center heights are selected as the next cluster center is related to the distance in step (2), that is, the farther the height is, the more likely it is to become the next cluster center; according to the probability, select a height from all non-cluster center heights as the next cluster center; (4) if the number of specified cluster centers is greater than 2, repeat steps (2) and (3) until the specified number of cluster centers is selected; (5) cluster based on these multiple cluster centers to obtain multiple clusters.

[0067] In this embodiment of the disclosure, for each cluster, the electronic device can sum and average the annotation heights of all labeled members in the cluster to obtain the average annotation height of the cluster. Furthermore, for at least two adjacent clusters, the electronic device can sum and average the annotation heights of the two adjacent clusters, using the result as a threshold for distinguishing member heights.

[0068] For example, if there are three clusters, namely cluster A, cluster B and cluster C, then by combining the average height of the labels in cluster A and cluster B, a threshold for distinguishing the height of a pole can be determined; similarly, by combining the average height of the labels in cluster B and cluster C, a threshold for distinguishing the height of a pole can be determined.

[0069] Step 203: Based on the coordinate information of the two endpoints of the multiple labeled rods and the rod height differentiation threshold, determine at least two height intervals obtained by dividing based on the rod height differentiation threshold, and the set of labeled rods corresponding to the at least two height intervals.

[0070] In this embodiment of the disclosure, the electronic device may perform step 203 as follows: divide the height range based on the rod height differentiation threshold to obtain at least two height intervals; for each marked rod, determine the height interval to which the marked rod belongs based on the coordinate information of the two endpoints of the marked rod; for each height interval, determine the set of marked rods corresponding to the height interval based on the coordinate information of the two endpoints of the marked rods belonging to the height interval.

[0071] Specifically, for each labeled rod, the electronic device can determine the labeled height of the rod based on the labeled coordinate information of the two endpoints of the rod; determine the height range containing the labeled height; and use the height range containing the labeled height as the height range to which the labeled rod belongs.

[0072] Step 204: Based on the predicted coordinate information of the two endpoints of the multiple predicted members and the confidence scores of the multiple predicted members, determine the set of predicted members corresponding to at least two height intervals.

[0073] In this embodiment of the disclosure, the electronic device may perform step 204 as follows: for each predicted rod, determine the height range to which the predicted rod belongs based on the predicted coordinate information of the two endpoints of the predicted rod; for each height range, determine the set of predicted rods corresponding to the height range based on the predicted coordinate information of the two endpoints of the predicted rods belonging to the height range and the confidence score of the predicted rods belonging to the height range.

[0074] Specifically, for each predicted rod, the electronic device can determine the predicted height of the predicted rod based on the predicted coordinate information of the two endpoints of the predicted rod; determine the height range containing the predicted height; and use the height range containing the predicted height as the height range to which the predicted rod belongs.

[0075] Step 205: For each height interval, determine the confidence threshold corresponding to the height interval based on the set of labeled poles and the set of predicted poles corresponding to the height interval.

[0076] In this embodiment of the disclosure, the electronic device performing step 205 may, for example, involve setting an initial confidence threshold for each height interval; filtering the predicted pole set corresponding to the height interval based on the initial confidence threshold and the confidence scores of the predicted poles in the predicted pole set corresponding to the height interval to obtain a filtered predicted pole set; acquiring multiple target labeled poles in the labeled pole set corresponding to the height interval, and multiple target predicted poles in the filtered predicted pole set corresponding to the height interval; pairing the multiple target labeled poles and multiple target predicted poles based on the two endpoint labeled coordinate information of the multiple target labeled poles and the two endpoint labeled coordinate information of the multiple target predicted poles to obtain multiple sets of paired poles and at least one unpaired target labeled pole; determining pole indicators based on the number of paired pole groups, the number of unpaired labeled poles, the total number of target labeled poles, and the total number of target predicted poles; and adjusting the initial confidence threshold if the pole indicators do not meet the pole indicator conditions, until the pole indicators obtained based on the adjusted confidence threshold meet the pole indicator conditions.

[0077] In this embodiment of the disclosure, the process by which the electronic device determines the filtered set of predicted poles can be, for example, by sequentially filtering each predicted pole in the set of predicted poles corresponding to the height interval if the confidence score of the predicted pole is less than the initial confidence threshold, thereby obtaining the filtered set of predicted poles.

[0078] The filtered set of predicted poles may include the predicted coordinates of the two endpoints of multiple predicted poles. The confidence score of each predicted pole in the filtered set is greater than or equal to the initial confidence threshold.

[0079] In this embodiment of the disclosure, the process by which the electronic device performs pairing processing on multiple target labeled rods and multiple target predicted rods based on the coordinate information of the two endpoints of multiple target labeled rods and the coordinate information of the two endpoints of multiple target predicted rods can be as follows: For each target labeled rod, the distance between the target labeled rod and multiple target predicted rods is determined based on the coordinate information of the two endpoints of the target labeled rod and the coordinate information of the two endpoints of the multiple target predicted rods; if the minimum distance among the multiple distances is less than or equal to a preset distance threshold, the target predicted rod corresponding to the minimum distance is determined as the target predicted rod paired with the target labeled rod; if the minimum distance among the multiple distances is greater than the preset distance threshold, the target labeled rod is determined to be an unpaired target labeled rod.

[0080] In this embodiment of the disclosure, the component metrics include component recall and component precision. Correspondingly, the electronic device determines the component metrics based on the number of paired component groups, the number of unpaired labeled components, the total number of target labeled components, and the total number of target predicted components. For example, this process can involve determining the component recall as the ratio of the number of groups to the total number of target labeled components; determining the difference between the total number of target predicted components and the number of groups; and determining the component precision as the ratio of the difference to the total number of target predicted components.

[0081] In this embodiment, the criteria for the rod indicator can be that the rod recall rate is greater than or equal to the recall rate threshold corresponding to the height interval, and the rod precision is greater than or equal to the precision threshold corresponding to the height interval. Specifically, when there are two height intervals—a large rod height interval and a small rod height interval—the recall rate threshold corresponding to the large rod height interval can be larger to ensure the detection rate of the large rod; conversely, the precision threshold corresponding to the small rod height interval can be larger to ensure the precision of the small rod.

[0082] Step 206: Obtain the target image and the prediction results of the poles in the target image; the prediction results include the predicted coordinate information of the two endpoints of the poles in the target image and the confidence score of the poles.

[0083] Step 207: Determine the predicted height of the rod based on the predicted coordinate information of the two endpoints of the rod.

[0084] Step 208: Determine the height range to which the rod belongs based on the predicted height of the rod and the rod height differentiation threshold.

[0085] Step 209: Determine the rod detection result of the target image based on the predicted coordinate information of the two endpoints of the rod, the confidence score, and the confidence threshold corresponding to the height interval to which the rod belongs; wherein, different height intervals correspond to different confidence thresholds.

[0086] It should be noted that for details of steps 206 to 209, please refer to [link / reference needed]. Figure 1 Steps 101 to 104 in the illustrated embodiment will not be described in detail here.

[0087] It should be noted that steps 201 to 205 can be executed only once; while steps 206 to 209 can be executed once for each target image. Additionally, steps 201 to 205 can be executed in advance.

[0088] In the pole detection method of this disclosure embodiment, an image test set and pole prediction results of test images in the image test set are obtained. The image test set includes multiple test images and pole annotation results of the test images. The pole prediction results of the test images include the predicted coordinate information of the two endpoints of the predicted poles in the test images and the confidence score of the predicted poles. The pole annotation results of the test images include the labeled coordinate information of the two endpoints of the labeled poles in the test images. Based on the labeled coordinate information of the two endpoints of the multiple labeled poles, a pole height discrimination threshold is determined. Based on the labeled coordinate information of the two endpoints of the multiple labeled poles and the pole height discrimination threshold, at least two height intervals are determined, and a set of labeled poles corresponding to the at least two height intervals are determined. Based on the predicted coordinate information of the two endpoints of the multiple predicted poles and the confidence score of the multiple predicted poles, a set of predicted poles corresponding to the at least two height intervals is determined. For each height interval, based on the height... The system uses a set of labeled and predicted poles corresponding to different height intervals to determine the confidence threshold for each interval. It then acquires the target image and the pole prediction results for that image. The prediction results include the predicted coordinates of the two endpoints of each pole in the target image and the pole's confidence score. Based on the predicted coordinates of the two endpoints, the predicted height of each pole is determined. Based on the predicted height and the pole height discrimination threshold, the height interval to which the pole belongs is determined. Finally, based on the predicted coordinates of the two endpoints, the confidence score, and the confidence threshold corresponding to the height interval to which the pole belongs, the pole detection result for the target image is determined. Different height intervals correspond to different confidence thresholds. By setting different confidence thresholds, higher height intervals are biased towards detecting taller poles, thus ensuring a higher detection rate for taller poles and improving mapping and localization efficiency.

[0089] The following example illustrates this. Assume there is one threshold value for distinguishing the height of the pole, and two height intervals: one for large poles and one for small poles. For example... Figure 3 The diagram shows the determination of the confidence thresholds for the height range of large members and the height range of small members.

[0090] exist Figure 3The process may include the following steps: (1) Obtain the pole test set (image test set). (2) Calculate the height of each pole in the pole test set, that is, the labeled height of each labeled pole in the test image of the image test set. (3) Cluster the pole heights and automatically divide them into two intervals, namely the height interval of large poles and the height interval of small poles. (4) Obtain the test images (test images) in the image test set. The execution order of steps (4) and (1) is not important. (5) The pole detection results (pole prediction results) of the test images output by the model (pole detection model) include endpoint coordinates (predicted coordinate information of the two endpoints of the predicted pole) and confidence score (confidence score of the predicted pole). (6) Calculate the height of each detection result (predicted pole), that is, the predicted height of each predicted pole. (7) Combine the two intervals obtained in step (3) to determine the height interval to which each detection result belongs; if the height interval to which it belongs is the height interval of large poles, execute step (8); if the height interval to which it belongs is the height interval of small poles, execute step (10). (8) Calculate the recall and precision of large members separately, that is, the recall and precision of members corresponding to the height range of large members. (9) Optimize the threshold th_b (the confidence threshold corresponding to the height range of large members) to ensure a high recall; then execute step (12). (10) Calculate the recall and precision of small members separately, that is, the recall and precision of members corresponding to the height range of small members. (11) Optimize the threshold th_s (the confidence threshold corresponding to the height range of small members) to ensure a high precision; then execute step (12). (12) Save the confidence threshold th_b of large members and the confidence threshold th_s of small members.

[0091] The following example illustrates this. Assume there is one threshold value for distinguishing the height of the pole, and two height intervals: one for large poles and one for small poles. For example... Figure 4 The diagram shown illustrates the results of rod detection in a target image.

[0092] exist Figure 4The process may include the following steps: (1) The camera captures the input image (target image). (2) The model (rod detection model) outputs the rod detection result (rod prediction result) of the image, including the endpoint coordinates (predicted coordinate information of the two endpoints of the rod) and the confidence score (confidence score of the rod). (3) Calculate the height of each detection result, that is, the predicted height of the rod. (4) Determine the height range to which each detection result (rod) belongs; if it is a large rod height range, execute step (5); if it is a small rod height range, execute step (8). (5) Whether the confidence score of the rod is greater than the confidence threshold th_b; if it is greater, execute step (7); otherwise, execute step (6). (6) Discard the false detection results. (7) Output the detection results. (8) Whether the confidence score of the rod is greater than the confidence threshold th_s; if it is greater, execute step (7); otherwise, execute step (9). (9) Discard the false detection results.

[0093] Figure 5 This is a schematic diagram of the structure of a rod detection device according to an embodiment of the present disclosure.

[0094] like Figure 5 As shown, the rod detection device may include: a first acquisition module 501, a first determination module 502, a second determination module 503, and a third determination module 504.

[0095] The first acquisition module 501 is used to acquire a target image and a rod prediction result of the target image; the rod prediction result includes the predicted coordinate information of the two endpoints of the rod in the target image and the confidence score of the rod.

[0096] The first determining module 502 is used to determine the predicted height of the rod based on the predicted coordinate information of the two endpoints of the rod.

[0097] The second determining module 503 is used to determine the height range to which the rod belongs based on the predicted height of the rod and the rod height differentiation threshold.

[0098] The third determining module 504 is used to determine the rod detection result of the target image based on the predicted coordinate information of the two endpoints of the rod, the confidence score, and the confidence threshold corresponding to the height interval to which the rod belongs; wherein, different height intervals correspond to different confidence thresholds.

[0099] In one embodiment of this disclosure, the number of the member height differentiation thresholds is at least one; when the number of member height differentiation thresholds is one, the number of height intervals is two, namely a large member height interval and a small member height interval; the height in the large member height interval is greater than or equal to the member height differentiation threshold; the height in the small member height interval is less than the member height differentiation threshold; the confidence threshold corresponding to the large member height interval is less than the confidence threshold corresponding to the small member height interval.

[0100] In one embodiment of this disclosure, the third determining module 504 is specifically used to determine the pole as a target pole when the confidence score of the pole is greater than or equal to the confidence threshold corresponding to the height interval to which the pole belongs; and to determine the pole detection result of the target image based on the predicted coordinate information of the two endpoints of the target pole.

[0101] In one embodiment of this disclosure, the apparatus further includes: a second acquisition module, a fourth determination module, a fifth determination module, a sixth determination module, and a seventh determination module; the second acquisition module is configured to acquire an image test set and the pole prediction results of test images in the image test set; the image test set includes multiple test images and pole annotation results of the test images; the pole prediction results of the test images include the predicted coordinate information of the two endpoints of the predicted poles in the test images and the confidence score of the predicted poles; the pole annotation results of the test images include the labeled coordinate information of the two endpoints of the labeled poles in the test images; the fourth determination module is configured to determine the coordinate information of the two endpoints of the labeled poles based on the coordinate information of the two endpoints of the multiple labeled poles. The fifth determining module is used to determine at least two height intervals and at least two sets of labeled poles corresponding to the height distinction threshold, based on the coordinate information of the two endpoints of the labeled poles and the height distinction threshold. The sixth determining module is used to determine at least two sets of predicted poles corresponding to the height intervals, based on the predicted coordinate information of the two endpoints of the predicted poles and the confidence scores of the predicted poles. The seventh determining module is used to determine the confidence threshold corresponding to each height interval based on the set of labeled poles and the set of predicted poles.

[0102] In one embodiment of this disclosure, the fourth determining module is specifically used to: determine the labeling height of the plurality of labeling rods based on the labeling coordinate information of the two endpoints of the plurality of labeling rods; perform clustering processing on the plurality of labeling rods based on the labeling height of the plurality of labeling rods to obtain at least two clusters; for each cluster, determine the average labeling height of the cluster based on the labeling height of the labeling rods in the cluster; and determine the rod height differentiation threshold based on the average labeling height of two adjacent clusters in the at least two clusters.

[0103] In one embodiment of this disclosure, the fifth determining module is specifically used to: divide the height range based on the rod height differentiation threshold to obtain at least two height intervals; for each marked rod, determine the height interval to which the marked rod belongs based on the coordinate information of the two endpoints of the marked rod; and for each height interval, determine the set of marked rods corresponding to the height interval based on the coordinate information of the two endpoints of the marked rods belonging to the height interval.

[0104] In one embodiment of this disclosure, the sixth determining module is specifically used to: for each predicted rod, determine the height interval to which the predicted rod belongs based on the predicted coordinate information of the two endpoints of the predicted rod; and for each height interval, determine the set of predicted rods corresponding to the height interval based on the predicted coordinate information of the two endpoints of the predicted rods belonging to the height interval and the confidence score of the predicted rods belonging to the height interval.

[0105] In one embodiment of this disclosure, the seventh determining module includes: a setting unit, a filtering processing unit, an acquisition unit, a pairing processing unit, a determining unit, and an adjustment processing unit; the setting unit is used to set an initial confidence threshold corresponding to each height interval; the filtering processing unit is used to filter the predicted pole set corresponding to the height interval based on the initial confidence threshold and the confidence scores of the predicted poles in the predicted pole set corresponding to the height interval, to obtain a filtered predicted pole set; the acquisition unit is used to acquire multiple target labeled poles in the labeled pole set corresponding to the height interval, and multiple target predicted poles in the filtered predicted pole set corresponding to the height interval; the pairing processing unit... The first unit is used to perform pairing processing on multiple target labeled rods and multiple target predicted rods based on the coordinate information of the two endpoints of the multiple target labeled rods and the coordinate information of the two endpoints of the multiple target predicted rods, to obtain multiple sets of paired rods and at least one unpaired target labeled rod; the second unit is used to determine the rod index based on the number of paired rods, the number of unpaired labeled rods, the total number of target labeled rods, and the total number of target predicted rods; the third unit is used to adjust the initial confidence threshold when the rod index does not meet the rod index condition, until the rod index determined based on the adjusted confidence threshold meets the rod index condition.

[0106] In one embodiment of this disclosure, the filtering processing unit is specifically used to sequentially perform filtering processing on each predicted pole in the predicted pole set corresponding to the height interval, provided that the confidence score of the predicted pole is less than the initial confidence threshold, to obtain a filtered predicted pole set.

[0107] In one embodiment of this disclosure, the pairing processing unit is specifically configured to, sequentially for each target labeled rod, determine the distance between the target labeled rod and the multiple target predicted rods based on the coordinate information of the two endpoints of the target labeled rod and the coordinate information of the two endpoints of the multiple target predicted rods; if the minimum distance among the multiple distances is less than or equal to a preset distance threshold, determine the target predicted rod corresponding to the minimum distance as a target predicted rod paired with the target labeled rod; if the minimum distance among the multiple distances is greater than the preset distance threshold, determine the target labeled rod as an unpaired target labeled rod.

[0108] In one embodiment of this disclosure, the pole metrics include pole recall and pole accuracy. The determining unit is specifically configured to: determine the pole recall as the ratio of the number of groups to the total number of target labeled poles; determine the difference between the total number of target predicted poles and the number of groups; and determine the pole accuracy as the ratio of the difference to the total number of target predicted poles.

[0109] In one embodiment of this disclosure, the pole metrics include: pole recall rate and pole precision rate; the pole metric condition is that the pole recall rate is greater than or equal to the recall rate threshold corresponding to the height interval, and the pole precision rate is greater than or equal to the precision rate threshold corresponding to the height interval.

[0110] In the pole detection device of this embodiment, a target image and pole prediction results of the target image are acquired. The pole prediction results include the predicted coordinate information of the two endpoints of the pole in the target image and the confidence score of the pole. The predicted height of the pole is determined based on the predicted coordinate information of the two endpoints of the pole. The height range to which the pole belongs is determined based on the predicted height of the pole and the pole height differentiation threshold. The pole detection result of the target image is determined based on the predicted coordinate information of the two endpoints of the pole, the confidence score, and the confidence threshold corresponding to the height range to which the pole belongs. Different height ranges correspond to different confidence thresholds. By setting different confidence thresholds, the confidence threshold corresponding to higher height ranges can be biased towards the detection of higher poles, thereby ensuring the detection rate of higher poles and improving mapping and positioning efficiency.

[0111] According to a third aspect of the present disclosure, an electronic device is also provided, comprising: a processor; and a memory for storing processor-executable instructions, wherein the processor is configured to implement the rod detection method as described above.

[0112] To implement the above embodiments, this disclosure also proposes a storage medium.

[0113] When the instructions in the storage medium are executed by the processor, the processor is able to perform the rod detection method as described above.

[0114] To implement the above embodiments, this disclosure also provides a computer program product.

[0115] When the computer program product is executed by the processor of the electronic device, it enables the electronic device to perform the above-described method.

[0116] Figure 6 This is a structural block diagram of an electronic device according to an exemplary embodiment. Figure 6The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of the embodiments disclosed herein.

[0117] like Figure 6 As shown, the electronic device 1000 includes a processor 111, which can perform various appropriate actions and processes according to a program stored in read-only memory (ROM) 112 or a program loaded from memory 116 into random access memory (RAM) 113. The RAM 113 also stores various programs and data required for the operation of the electronic device 1000. The processor 111, ROM 112, and RAM 113 are interconnected via a bus 114. An input / output (I / O) interface 115 is also connected to the bus 114.

[0118] The following components are connected to I / O interface 115: memory 116 including hard disks, etc.; and communication section 117 including network interface cards such as local area network (LAN) cards, modems, etc., communication section 117 performs communication processing via a network such as the Internet; and driver 118 is also connected to I / O interface 115 as needed.

[0119] In particular, according to embodiments of this disclosure, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of this disclosure include a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication section 117. When the computer program is executed by processor 111, it performs the functions defined in the methods of this disclosure.

[0120] In an exemplary embodiment, a storage medium including instructions is also provided, such as a memory including instructions, which can be executed by the processor 111 of the electronic device 1000 to perform the above-described method. Optionally, the storage medium may be a non-transitory computer-readable storage medium, such as a ROM, random access memory (RAM), CD-ROM, magnetic tape, floppy disk, and optical data storage device.

[0121] In this disclosure, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in connection with an instruction execution system, apparatus, or device. In this disclosure, a computer-readable signal medium can include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium can also be any computer-readable medium other than a computer-readable storage medium, which can transmit, propagate, or transfer a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to: wireless, wireline, optical fiber, RF, etc., or any suitable combination thereof.

[0122] Figure 7 This is a block diagram illustrating a vehicle 700 according to an exemplary embodiment of the present disclosure. For example, vehicle 700 may be a hybrid vehicle, a non-hybrid vehicle, an electric vehicle, a fuel cell vehicle, or other types of vehicle. Vehicle 700 may be an autonomous vehicle, a semi-autonomous vehicle, or a non-autonomous vehicle.

[0123] Reference Figure 7 The vehicle 700 may include various subsystems, such as an infotainment system 710, a perception system 720, a decision control system 730, a drive system 740, and a computing platform 750. The vehicle 700 may also include more or fewer subsystems, and each subsystem may include multiple components. Furthermore, each subsystem and each component of the vehicle 700 can be interconnected via wired or wireless means.

[0124] In one example, the computing platform 750 can be as follows: Figure 6The electronic device shown can acquire a target image and the prediction results of the poles in the target image. The prediction results include the predicted coordinates of the two endpoints of the pole in the target image and the confidence score of the pole. Based on the predicted coordinates of the two endpoints, the predicted height of the pole is determined. Based on the predicted height and a height discrimination threshold, the height range to which the pole belongs is determined. Based on the predicted coordinates of the two endpoints, the confidence score, and the confidence threshold corresponding to the height range to which the pole belongs, the pole detection result of the target image is determined. Different height ranges correspond to different confidence thresholds. By setting different confidence thresholds, the confidence threshold corresponding to higher height ranges can be biased towards the detection of taller poles, thereby ensuring the detection rate of taller poles and improving mapping and localization efficiency. In another example, the computing platform 750 can be connected to... Figure 6 The electronic device shown is connected to obtain the pole detection results of the target image processed by the electronic device, which is used for mapping and localization of autonomous vehicles.

[0125] In some embodiments, the infotainment system 710 may include a communication system, an entertainment system, and a navigation system, etc.

[0126] The perception system 720 may include several sensors for sensing information about the environment surrounding the vehicle 700. For example, the perception system 720 may include a global positioning system (which may be GPS, BeiDou, or other positioning systems), an inertial measurement unit (IMU), lidar, millimeter-wave radar, ultrasonic radar, and a camera device.

[0127] The decision control system 730 may include a computing system, a vehicle controller, a steering system, a throttle, and a braking system.

[0128] The drive system 740 may include components that provide powered motion to the vehicle 700. In one embodiment, the drive system 740 may include an engine, an energy source, a transmission system, and wheels. The engine may be one or a combination of internal combustion engines, electric motors, and compressed air engines. The engine is capable of converting energy provided by the energy source into mechanical energy.

[0129] Some or all of the functions of vehicle 700 are controlled by computing platform 750. Computing platform 750 may include at least one processor 751 and memory 752, and processor 751 may execute instructions 753 stored in memory 752.

[0130] Processor 751 can be any conventional processor, such as a commercially available CPU. The processor may also include, for example, a Graphics Processing Unit (GPU), a Field Programmable Gate Array (FPGA), a System on Chip (SOC), an Application Specific Integrated Circuit (ASIC), or a combination thereof.

[0131] The memory 752 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk or optical disk.

[0132] In addition to instruction 753, memory 752 can also store data, such as road maps, route information, vehicle position, direction, speed, and other data. The data stored in memory 752 can be used by computing platform 750.

[0133] In this embodiment of the disclosure, the processor 751 may execute instruction 753 to complete all or part of the steps of the above-described rod detection method.

[0134] Furthermore, the term “exemplary” is used herein to mean serving as an example, instance, or illustration. Any aspect or design described herein as “exemplary” is not necessarily to be construed as advantageous compared to other aspects or designs. Rather, the use of the term “exemplary” is intended to present the concept in a concrete manner. As used herein, the term “or” is intended to mean an inclusive “or” rather than an exclusive “or.” That is, unless otherwise specified or clear from the context, “X applies A or B” is intended to mean any of the natural inclusive arrangements. That is, “X applies A or B” satisfies any of the foregoing instances if X applies A; X applies B; or both X applies A and B. Additionally, unless otherwise specified or clear from the context to refer to the singular form, the articles “a” and “an” as used in this application and the appended claims are generally understood to mean “one or more.”

[0135] Similarly, although this disclosure has been shown and described with respect to one or more implementations, equivalent variations and modifications will occur to those skilled in the art upon reading and understanding the specification and drawings. This disclosure includes all such modifications and variations and is limited only by the scope of the claims. In particular, with respect to the various functions performed by the components described above (e.g., elements, resources, etc.), unless otherwise indicated, the terminology used to describe such components is intended to correspond to any component (functionally equivalent) that performs the specific function of the described component, even if structurally not equivalent to the disclosed structure. Furthermore, although specific features of this disclosure may have been disclosed with respect to only one of several implementations, such features may be combined with one or more other features of other implementations, as may be desired and advantageous to any given or particular application. Moreover, with regard to the terms “comprising,” “owning,” “having,” “having,” or variations thereof as used in the detailed description or claims, such terms are intended to be inclusive in a manner similar to the term “including.”

[0136] Other embodiments of this disclosure will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of this disclosure that follow the general principles of this disclosure and include common knowledge or customary techniques in the art not disclosed herein. The specification and examples are to be considered exemplary only, and the true scope and spirit of this disclosure are indicated by the following claims.

[0137] It should be understood that this disclosure is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this disclosure is limited only by the appended claims.

Claims

1. A rod member inspection method characterized by comprising: The method includes: Acquire a target image and the pole prediction results of the target image; the pole prediction results include the predicted coordinate information of the two endpoints of the pole in the target image, and the confidence score of the pole; The predicted height of the rod is determined based on the predicted coordinate information of the two endpoints of the rod. The height range to which the rod belongs is determined based on the predicted height of the rod and the rod height differentiation threshold; Based on the predicted coordinate information of the two endpoints of the rod, the confidence score, and the confidence threshold corresponding to the height interval to which the rod belongs, the rod detection result of the target image is determined; wherein, different height intervals correspond to different confidence thresholds; The step of determining the pole detection result of the target image based on the predicted coordinate information of the two endpoints of the pole, the confidence score, and the confidence threshold corresponding to the height interval to which the pole belongs includes: determining the pole as a target pole when the confidence score of the pole is greater than or equal to the confidence threshold corresponding to the height interval to which the pole belongs; and determining the pole detection result of the target image based on the predicted coordinate information of the two endpoints of the target pole.

2. The method of claim 1, wherein, The number of the rod height differentiation thresholds is at least one; When the number of the rod height differentiation threshold is one, the number of the height intervals is two, namely the large rod height interval and the small rod height interval; the height in the large rod height interval is greater than or equal to the rod height differentiation threshold; the height in the small rod height interval is less than the rod height differentiation threshold. The confidence threshold corresponding to the height range of the large rod is less than the confidence threshold corresponding to the height range of the small rod.

3. The method of claim 1, wherein, The method further includes: Obtain an image test set and the pole prediction results of the test images in the image test set; the image test set includes multiple test images and the pole annotation results of the test images; the pole prediction results of the test images include the predicted coordinate information of the two endpoints of the predicted pole in the test image and the confidence score of the predicted pole; the pole annotation results of the test images include the labeled coordinate information of the two endpoints of the labeled pole in the test image. Based on the coordinate information of the two endpoints of the multiple labeled rods, determine the rod height differentiation threshold; Based on the coordinate information of the two endpoints of the multiple labeled rods and the rod height differentiation threshold, at least two height intervals are determined based on the rod height differentiation threshold, and a set of labeled rods corresponding to at least two height intervals is determined. Based on the predicted coordinate information of the two endpoints of the multiple predicted rods and the confidence scores of the multiple predicted rods, determine at least two sets of predicted rods corresponding to the height intervals; For each height interval, a confidence threshold is determined based on the set of labeled poles and the set of predicted poles corresponding to the height interval.

4. The method of claim 3, wherein, The step of determining the height distinction threshold of the rods based on the coordinate information of the two endpoints of the multiple labeled rods includes: The annotation height of the multiple annotation rods is determined based on the annotation coordinate information of the two endpoints of the multiple annotation rods; Based on the annotation height of the multiple annotation rods, the multiple annotation rods are clustered to obtain at least two clusters; For each cluster, the average height of the labels in the cluster is determined based on the height of the labels on the labeled members in the cluster. The rod height differentiation threshold is determined based on the average height of the labels of two adjacent clusters in at least two of the said clusters.

5. The method of claim 3, wherein, The step of determining at least two height intervals divided based on the height distinction threshold and the set of labeled rods corresponding to the at least two height intervals, based on the coordinate information of the two endpoints of the multiple labeled rods and the height distinction threshold of the rods, includes: The height range is divided based on the height differentiation threshold of the rod to obtain at least two height intervals; For each marked rod, the height range to which the marked rod belongs is determined based on the coordinate information of the two endpoints of the marked rod. For each height interval, the set of labeled rods corresponding to the height interval is determined based on the coordinate information of the two endpoints of the labeled rods belonging to the height interval.

6. The method of claim 3, wherein, The step of determining a set of predicted poles corresponding to at least two height intervals based on the predicted coordinate information of the two endpoints of the plurality of predicted poles and the confidence scores of the plurality of predicted poles includes: For each predicted member, the height range to which the predicted member belongs is determined based on the predicted coordinate information of the two endpoints of the predicted member. For each height interval, the set of predicted poles corresponding to the height interval is determined based on the predicted coordinate information of the two endpoints of the predicted poles belonging to the height interval and the confidence score of the predicted poles belonging to the height interval.

7. The method of claim 3, wherein, For each height interval, determining the confidence threshold corresponding to the height interval based on the set of labeled poles and the set of predicted poles corresponding to the height interval includes: For each height interval, set an initial confidence threshold corresponding to the height interval; Based on the initial confidence threshold and the confidence scores of the predicted poles in the predicted pole set corresponding to the height interval, the predicted pole set corresponding to the height interval is filtered to obtain the filtered predicted pole set. Obtain multiple target labeled poles from the set of labeled poles corresponding to the height interval, and multiple target predicted poles from the set of filtered predicted poles corresponding to the height interval; Based on the coordinate information of the two endpoints of the multiple target labeled rods and the coordinate information of the two endpoints of the multiple target predicted rods, the multiple target labeled rods and the multiple target predicted rods are paired to obtain multiple sets of paired rods and at least one unpaired target labeled rod. The pole index is determined based on the number of paired pole groups, the number of unpaired target labeled poles, the total number of target labeled poles, and the total number of target predicted poles. If the member index does not meet the member index conditions, the initial confidence threshold is adjusted until the member index determined based on the adjusted confidence threshold meets the member index conditions.

8. The method of claim 7, wherein, The step involves filtering the set of predicted poles corresponding to the height interval based on the initial confidence threshold and the confidence scores of the predicted poles in the set of predicted poles corresponding to the height interval, to obtain a filtered set of predicted poles, including: For each predicted pole in the predicted pole set corresponding to the height interval, if the confidence score of the predicted pole is less than the initial confidence threshold, the predicted pole in the predicted pole set corresponding to the height interval is filtered to obtain a filtered predicted pole set.

9. The method of claim 7, wherein, The step of pairing multiple target labeled rods and multiple target predicted rods based on the coordinate information of the two endpoints of the multiple target labeled rods and the coordinate information of the two endpoints of the multiple target predicted rods to obtain multiple sets of paired rods and at least one unpaired target labeled rod includes: For each target labeled rod, the distance between the target labeled rod and the multiple target predicted rods is determined based on the coordinate information of the two endpoints of the target labeled rod and the coordinate information of the two endpoints of the multiple target predicted rods. If the minimum distance among the multiple distances is less than or equal to a preset distance threshold, the target predicted pole corresponding to the minimum distance is determined as the target predicted pole paired with the target labeled pole; If the minimum distance among the multiple distances is greater than the preset distance threshold, the target marker is determined to be an unpaired target marker.

10. The method of claim 7, wherein, The lever metrics include lever recall and lever precision. The lever metrics are determined based on the number of paired lever groups, the number of unpaired target labeled levers, the total number of target labeled levers, and the total number of target predicted levers, including: The ratio of the number of groups to the total number of target labeled rods is determined as the rod recall rate; Determine the difference between the total number of the target predicted members and the number of groups; The ratio of the difference to the total number of the target predicted members is determined as the member accuracy.

11. The method according to claim 7 or 10, characterized in that, The pole component metrics include: pole component recall rate and pole component accuracy rate; The criteria for the rod indicator are that the recall rate of the rod is greater than or equal to the recall rate threshold corresponding to the height interval, and the accuracy of the rod is greater than or equal to the accuracy threshold corresponding to the height interval.

12. A rod member detecting device characterized by comprising: The device includes: The first acquisition module is used to acquire a target image and a rod prediction result of the target image; the rod prediction result includes the predicted coordinate information of the two endpoints of the rod in the target image, and the confidence score of the rod. The first determining module is used to determine the predicted height of the rod based on the predicted coordinate information of the two endpoints of the rod. The second determining module is used to determine the height range to which the rod belongs based on the predicted height of the rod and the rod height differentiation threshold. The third determining module is used to determine the rod detection result of the target image based on the predicted coordinate information of the two endpoints of the rod, the confidence score, and the confidence threshold corresponding to the height interval to which the rod belongs; wherein, different height intervals correspond to different confidence thresholds; Specifically, the third determining module is used to determine the pole as a target pole when the confidence score of the pole is greater than or equal to the confidence threshold corresponding to the height interval to which the pole belongs; and to determine the pole detection result of the target image based on the predicted coordinate information of the two endpoints of the target pole.

13. An electronic device, comprising: include: processor; Memory used to store the processor's executable instructions; The processor is configured as follows: The steps of implementing the rod detection method as described in any one of claims 1 to 11.

14. A vehicle characterized by comprising: Includes the electronic device as claimed in claim 13, or the vehicle is connected to the electronic device as claimed in claim 13.

15. A non-transitory computer-readable storage medium, wherein instructions in the storage medium, when executed by a processor, enable the processor to perform the rod detection method as described in any one of claims 1 to 11.

Images