A background segmentation target recognition method under complex 3D background

Abstract

The present invention discloses a background segmentation target recognition method under a complex 3D background. A fixed plane coordinate system is established to perform preliminary sampling and recognition on the target. At the same time, AI recognition partitions the overall target pattern according to the texture and color of the target pattern, and the edges of each partition are traced point by point and a recognition anchor point is set at the center of the partition. The corresponding coordinates of each point in the traced line and the recognition anchor point are marked in the plane coordinate system. The scene is placed in the plane coordinate system and primary background segmentation is performed. Subsequently, the remaining area after the primary background segmentation is decomposed frame by frame, and a secondary comparison is performed with the texture and color of the target pattern. The edges of the area that meets the texture and color of the target pattern are secondary stroked and the corresponding coordinates of the secondary stroked lines are marked point by point. Subsequently, the target pattern area is accurately cut according to the new coordinates to finally obtain a target pattern with precise edges, and finally the recognition is completed. The present invention has the advantage of higher recognition accuracy for the target.

Description

Technical Field

[0001] The present invention relates to the field of AI recognition, and in particular to a background segmentation target recognition method under a complex 3D background. Background Art

[0002] With the development of science and technology, the research on autonomous unmanned intelligent vehicles has become one of the research hotspots of research institutions in various countries. Autonomous unmanned intelligent vehicles can effectively reduce the mortality rate of traffic accidents, complete dangerous environment operations without human conditions, and greatly improve the level of intelligence in human life. Environmental perception technology is one of the core technologies of autonomous intelligent vehicles. LiDAR and cameras are the core environmental perception sensors in current unmanned vehicle technology. The target recognition method in laser 3D point cloud data is a key technology to improve the environmental perception and simultaneous localization and mapping (SLAM) technology level of autonomous intelligent vehicles and autonomous navigation of autonomous intelligent vehicles.

[0003] Object recognition under 3D point cloud generally includes two stages: feature extraction and feature matching. At present, the object recognition methods under 3D point cloud can be mainly divided into the following two types according to the types of extracted features: recognition methods using global features and recognition methods using local features. Global features are mainly based on extracting the global contour features of the target. In previous experiments, they were too sensitive to factors such as target deformation and incomplete targets due to occlusion, and were gradually replaced by some stable local features with scale and rotation invariance. The method of relying on local features for target recognition is one of the most widely used methods. In the feature matching stage, feature matching methods can also be divided into two categories: direct feature point matching methods and indirect feature point matching methods. The direct feature point matching method directly calculates the matching degree of the extracted features on the model and the object to be identified, and identifies the target by counting the percentage of feature points that successfully match; while the indirect feature point matching method further combines the features in the model to achieve the description of global features through local features, and then matches these feature combinations between the model and the target to be matched to identify the target. Stable local features combined with appropriate direct feature point matching methods have achieved good results in some applications, but how to use stable local features to describe the global features of the target object is still one of the difficulties in achieving target recognition in complex scenes. The method based on local features and the bag-of-visual-words model solves the semantic gap of using local features to describe the global features of objects, and has achieved good results in the field of image processing. However, the traditional bag-of-visual-words model ignores the inherent spatial order logic of many local features in the process of expressing the global features of the target, which greatly reduces its description accuracy when describing the global features of the object, thus greatly reducing the recognition accuracy of this method. Summary of the invention

[0004] Purpose of the invention: The purpose of the present invention is to provide a background segmentation target recognition method under complex 3D background in view of the shortcomings of the prior art.

[0005] Technical solution: The background segmentation target recognition method under a complex 3D background described in the present invention comprises the following steps:

[0006] S1. Establish a fixed plane coordinate system, perform preliminary sampling and recognition of the target, and at the same time, AI recognition divides the target pattern into zones according to the texture and color of the target pattern, draws lines point by point on the edges of each zone, sets recognition anchor points at the center of the zone, and marks the corresponding coordinates of each point in the line drawing and the recognition anchor points in the plane coordinate system;

[0007] S2, placing the scene to be identified in which the complex 3D background and the target coexist in a plane coordinate system, and entering the corresponding coordinates obtained by preliminary sampling and identification into the plane coordinate system;

[0008] S3, as the scene moves, the area marked by the coordinates of each point in the traced line and the identified anchor point in the plane coordinate system changes continuously with the capture of the texture and color of the target pattern and identifies the rough area of ​​the target pattern in the scene, and then cuts the 3D background outside the rough area to complete the primary background segmentation;

[0009] S4. Decompose the remaining area after the primary background segmentation frame by frame, and compare it with the target pattern texture and color for the second time. Stroke the edge of the area that meets the target pattern texture and color for the second time and mark the corresponding coordinates of the second stroked lines point by point. Then, accurately cut the target pattern area according to the new coordinates, and finally obtain the target pattern with precise edges, and finally complete the recognition.

[0010] Preferably, after the second comparison and the second tracing are completed in S4 and the cropping is completed, a third comparison is performed. If the comparison is correct, the background segmentation is completed. Otherwise, three tracings and cropping are performed.

[0011] Preferably, the three-way comparison method in S4 combines AI recognition comparison with manual visual inspection.

[0012] Preferably, in S2, a central axis is set up in the plane coordinate system, and the scene moves along the central axis.

[0013] Compared with the prior art, the present invention has the following beneficial effects: compared with the traditional AI recognition method of directly capturing the content in the scene, the segmentation and cropping method can achieve accurate acquisition of the edge of the target pattern, improve the recognition accuracy, and effectively reduce the background recognition and analysis process, thereby improving the recognition efficiency. DETAILED DESCRIPTION

[0014] The technical solutions in the embodiments of the present invention are described clearly and completely below. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0015] In the description of the present invention, it is to be understood that the terms “center”, “longitudinal”, “lateral”, “up”, “down”, “front”, “back”, “left”, “right”, “vertical”, “horizontal”, “top”, “bottom”, “inside”, “outside”, “axial”, “radial”, “circumferential”, etc., indicating orientations or positional relationships, are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be understood as limiting the present invention.

[0016] In the present invention, unless otherwise clearly specified and limited, the terms "installed", "connected", "connected", "fixed" and the like should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integrated connection; it can be a mechanical connection, an electrical connection, or a communication connection; it can be a direct connection, or an indirect connection through an intermediate medium, it can be the internal connection of two elements or the interaction relationship between two elements, unless otherwise clearly defined. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0017] The technical solution of the present invention is described in detail with specific embodiments below. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described in detail in some embodiments.

[0018] Embodiment 1: A background segmentation target recognition method under a complex 3D background comprises the following steps:

[0019] S1. Establish a fixed plane coordinate system, perform preliminary sampling and recognition of the target, and at the same time, AI recognition divides the target pattern into zones according to the texture and color of the target pattern, draws lines point by point on the edges of each zone, sets recognition anchor points at the center of the zone, and marks the corresponding coordinates of each point in the line drawing and the recognition anchor points in the plane coordinate system;

[0020] S2, placing the scene to be identified in which the complex 3D background and the target coexist in a plane coordinate system, and entering the corresponding coordinates obtained by preliminary sampling and identification into the plane coordinate system;

[0021] S3, as the scene moves, the area marked by the coordinates of each point in the traced line and the identified anchor point in the plane coordinate system changes continuously with the capture of the texture and color of the target pattern and identifies the rough area of ​​the target pattern in the scene, and then cuts the 3D background outside the rough area to complete the primary background segmentation;

[0022] S4. Decompose the remaining area after the primary background segmentation frame by frame, and compare it with the target pattern texture and color for the second time. Stroke the edge of the area that meets the target pattern texture and color for the second time and mark the corresponding coordinates of the second stroked lines point by point. Then, accurately cut the target pattern area according to the new coordinates, and finally obtain the target pattern with precise edges, and finally complete the recognition.

[0023] Embodiment 2: A background segmentation target recognition method under a complex 3D background comprises the following steps:

[0024] S1. Establish a fixed plane coordinate system, perform preliminary sampling and recognition of the target, and at the same time, AI recognition divides the target pattern into zones according to the texture and color of the target pattern, draws lines point by point on the edges of each zone, sets recognition anchor points at the center of the zone, and marks the corresponding coordinates of each point in the line drawing and the recognition anchor points in the plane coordinate system;

[0025] S2, placing the scene to be identified in which the complex 3D background and the target coexist in a plane coordinate system, setting up a central axis in the plane coordinate system, and moving the scene along the central axis, and entering the corresponding coordinates obtained by preliminary sampling and identification into the plane coordinate system;

[0026] S3, as the scene moves, the area marked by the coordinates of each point in the traced line and the identified anchor point in the plane coordinate system changes continuously with the capture of the texture and color of the target pattern and identifies the rough area of ​​the target pattern in the scene, and then cuts the 3D background outside the rough area to complete the primary background segmentation;

[0027] S4. Decompose the remaining area after the primary background segmentation frame by frame, and compare it with the target pattern texture and color for the second time, stroke the edge of the area that meets the target pattern texture and color for the second time, and mark the corresponding coordinates of the second stroked lines point by point, then accurately cut the target pattern area according to the new coordinates, and perform a third comparison after the second comparison and cutting after the second stroke. If the comparison is correct, the background segmentation is completed, otherwise, stroke and cut are performed three times. The three-way comparison method combines AI recognition and comparison with manual inspection to finally obtain the target pattern with precise edges and finally complete the recognition.

[0028] The advantage of adopting this technical solution is that compared with the traditional AI recognition method of directly capturing the content in the scene, the segmentation and cropping method can achieve accurate acquisition of the edges of the target pattern, improve recognition accuracy, and effectively reduce the background recognition and analysis process, thereby improving recognition efficiency.

[0029] In the present invention, unless otherwise clearly specified and limited, the first feature being "above" or "below" the second feature may mean that the first feature is in direct contact with the second feature, or the first feature and the second feature are in indirect contact through an intermediate medium. Moreover, the first feature being "above", "above" and "above" the second feature may mean that the first feature is directly above or obliquely above the second feature, or simply means that the first feature is higher in level than the second feature. The first feature being "below", "below" and "below" the second feature may mean that the first feature is directly below or obliquely below the second feature, or simply means that the first feature is lower in level than the second feature. In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example" or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic expressions of the above terms do not necessarily refer to the same embodiment or example.

[0030] Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art can combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples without contradicting each other.

[0031] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein by equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A background segmentation target recognition method under complex 3D background, characterized by: The following steps are involved: S1. Establish a fixed plane coordinate system, perform preliminary sampling and recognition of the target, and at the same time, AI recognition divides the target pattern into zones according to the texture and color of the target pattern, draws lines point by point on the edges of each zone, sets recognition anchor points at the center of the zone, and marks the corresponding coordinates of each point in the line drawing and the recognition anchor points in the plane coordinate system; S2, placing the scene to be identified in which the complex 3D background and the target coexist in a plane coordinate system, and entering the corresponding coordinates obtained by preliminary sampling and identification into the plane coordinate system; S3, as the scene moves, the area marked by the coordinates of each point in the traced line and the identified anchor point in the plane coordinate system changes continuously with the capture of the texture and color of the target pattern and identifies the rough area of ​​the target pattern in the scene, and then cuts the 3D background outside the rough area to complete the primary background segmentation; S4. Decompose the remaining area after the primary background segmentation frame by frame, and compare it with the target pattern texture and color for the second time. Stroke the edge of the area that meets the target pattern texture and color for the second time and mark the corresponding coordinates of the second stroked lines point by point. Then, accurately cut the target pattern area according to the new coordinates, and finally obtain the target pattern with precise edges, and finally complete the recognition.

2. The method for background segmentation and target recognition under a complex 3D background according to claim 1, characterized in that: After the second comparison and the second tracing are completed in S4, the third comparison is performed. If the comparison is correct, the background segmentation is completed. Otherwise, the third tracing and cutting are performed.

3. The method for background segmentation and target recognition under a complex 3D background according to claim 2, characterized in that: The three-way comparison method in S4 combines AI recognition and comparison with manual visual inspection.

4. The method for background segmentation and target recognition under a complex 3D background according to claim 1, characterized in that: In S2, a central axis is set up in the plane coordinate system, and the scene moves along the central axis.