CN115131314A - Object detection method, device, device and storage medium - Google Patents

Abstract

The present disclosure provides a target detection method, apparatus, device and storage medium, which relate to the technical field of artificial intelligence, specifically to the technical fields of image processing, computer vision, virtual reality, deep learning, and the like, and can be applied to scenes such as 3D vision, smart cities, intelligent transportation, and the like. The method comprises the following steps: acquiring an image to be detected; dividing an image to be detected into at least two image blocks based on the depth information, wherein the at least two image blocks are overlapped; performing target detection on the image blocks aiming at each of the at least two image blocks to obtain detection frames corresponding to target objects in the image blocks; and in response to determining that the same target object corresponds to at least one detection frame, determining a target detection frame of the target object based on the position of the at least one detection frame. The target detection method provided by the disclosure improves the efficiency and accuracy of target detection.

Description

Object detection method, device, device and storage medium

technical field

The present disclosure relates to the technical field of artificial intelligence, in particular to the technical fields of image processing, computer vision, virtual reality, and deep learning, and can be applied to scenarios such as 3D vision, smart cities, and intelligent transportation.

Background technique

Monocular 3D Object Detection (Monocular 3D Object Detection) is one of the most widely used computer vision technologies, which can be applied to vehicle automatic driving systems, intelligent robots, intelligent transportation and other fields. At present, monocular 3D target detection strongly relies on a single model to estimate the 3D properties of the target object in the entire scene, and different models have different scene perception capabilities at different distances. The range of scenes is overfit, which affects the performance of 3D object detection.

SUMMARY OF THE INVENTION

The present disclosure provides a target detection method, apparatus, device, and storage medium.

According to a first aspect of the present disclosure, there is provided a target detection method, comprising: acquiring an image to be detected; dividing the image to be detected into at least two image blocks based on depth information, wherein there is overlap between the at least two image blocks ; For each image block in the at least two image blocks, perform target detection on the image block to obtain a detection frame corresponding to the target object in the image block; In response to determining that the same target object corresponds to at least one detection frame, based on at least one detection frame The position of the frame determines the target detection frame of the target object.

According to a second aspect of the present disclosure, there is provided a target detection apparatus, comprising: an acquisition module configured to acquire an image to be detected; a division module configured to divide the image to be detected into at least two image blocks based on depth information, Wherein, there is overlap between at least two image blocks; the detection module is configured to perform target detection on the image blocks for each image block in the at least two image blocks, and obtain a detection frame corresponding to the target object in the image blocks a determining module configured to, in response to determining that the same target object corresponds to at least one detection frame, determine a target detection frame of the target object based on the position of the at least one detection frame.

According to a third aspect of the present disclosure, there is provided an electronic device, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor. The at least one processor executes to enable the at least one processor to perform a method as described in any implementation of the first aspect.

According to a fourth aspect of the present disclosure, there is provided a non-transitory computer-readable storage medium storing computer instructions for causing a computer to perform a method as described in any of the implementations of the first aspect.

According to a fifth aspect of the present disclosure, there is provided a computer program product comprising a computer program which, when executed by a processor, implements the method as described in any one of the implementations of the first aspect.

It should be understood that what is described in this section is not intended to identify key or critical features of embodiments of the disclosure, nor is it intended to limit the scope of the disclosure. Other features of the present disclosure will become readily understood from the following description.

Description of drawings

The accompanying drawings are used for better understanding of the present solution, and do not constitute a limitation to the present disclosure. in:

FIG. 1 is an exemplary system architecture diagram to which the present disclosure may be applied;

FIG. 2 is a flowchart of an embodiment of a target detection method according to the present disclosure;

3 is a flowchart of another embodiment of a target detection method according to the present disclosure;

4 is a flow chart of yet another embodiment of a target detection method according to the present disclosure;

5 is an application scenario diagram of the target detection method according to the present disclosure;

6 is a schematic structural diagram of an embodiment of a target detection apparatus according to the present disclosure;

FIG. 7 is a block diagram of an electronic device used to implement the target detection method of an embodiment of the present disclosure.

Detailed ways

Exemplary embodiments of the present disclosure are described below with reference to the accompanying drawings, which include various details of the embodiments of the present disclosure to facilitate understanding and should be considered as exemplary only. Accordingly, those of ordinary skill in the art will recognize that various changes and modifications of the embodiments described herein can be made without departing from the scope and spirit of the present disclosure. Also, descriptions of well-known functions and constructions are omitted from the following description for clarity and conciseness.

It should be noted that the embodiments of the present disclosure and the features of the embodiments may be combined with each other under the condition of no conflict. The present disclosure will be described in detail below with reference to the accompanying drawings and in conjunction with embodiments.

FIG. 1 illustrates an exemplary system architecture 100 to which embodiments of the object detection method or object detection apparatus of the present disclosure may be applied.

As shown in FIG. 1 , the system architecture 100 may include terminal devices 101 , 102 , and 103 , a network 104 and a server 105 . The network 104 is a medium used to provide a communication link between the terminal devices 101 , 102 , 103 and the server 105 . The network 104 may include various connection types, such as wired, wireless communication links, or fiber optic cables, among others.

The user can use the terminal devices 101, 102, 103 to interact with the server 105 through the network 104 to receive or send information and the like. Various client applications may be installed on the terminal devices 101 , 102 and 103 .

The terminal devices 101, 102, and 103 may be hardware or software. When the terminal devices 101, 102, and 103 are hardware, they can be various electronic devices, including but not limited to smart phones, tablet computers, laptop computers, desktop computers, and the like. When the terminal devices 101, 102, and 103 are software, they can be installed in the above-mentioned electronic devices. It can be implemented as a plurality of software or software modules, and can also be implemented as a single software or software module. There is no specific limitation here.

The server 105 may provide various services. For example, the server 105 may analyze and process the images to be detected obtained from the terminal devices 101 , 102 and 103 , and generate processing results (eg, target detection frames of the target objects).

It should be noted that the server 105 may be hardware or software. When the server 105 is hardware, it can be implemented as a distributed server cluster composed of multiple servers, or can be implemented as a single server. When the server 105 is software, it can be implemented as a plurality of software or software modules (for example, for providing distributed services), or can be implemented as a single software or software module. There is no specific limitation here.

It should be noted that the target detection method provided by the embodiment of the present disclosure is generally executed by the server 105 , and accordingly, the target detection apparatus is generally set in the server 105 .

It should be understood that the numbers of terminal devices, networks and servers in FIG. 1 are merely illustrative. There can be any number of terminal devices, networks and servers according to implementation needs.

Continuing to refer to FIG. 2, a flow 200 of one embodiment of a target detection method according to the present disclosure is shown. The target detection method includes the following steps:

Step 201, acquiring an image to be detected.

In this embodiment, the execution body of the target detection method (for example, the server 105 shown in FIG. 1 ) acquires the image to be detected. The image to be detected in this embodiment may be collected by a monocular camera, and the monocular camera may send the collected image to be detected to the above-mentioned execution subject in real time, or the monocular camera may also collect images at preset time intervals The to-be-detected image is sent to the above executive body.

Step 202: Divide the image to be detected into at least two image blocks based on the depth information.

In this embodiment, the above-mentioned execution body will divide the image to be detected into at least two image blocks based on the depth information, wherein there is overlap between the at least two image blocks. The depth information here refers to the depth range, which is the perception range of the image to be detected, such as 0-30 meters, 30-50 meters, and so on. For example, the above-mentioned executive body may divide the image to be detected based on a distance adaptive strategy to obtain multiple image blocks of different depth ranges, and the above-mentioned executive body will divide the to-be-detected image on the vertical axis of the image. Generally, the above executive body will divide the image to be detected according to three depth ranges of near, medium and far.

For example, if the above-mentioned executive body determines that the overall perception range of the image to be detected is 70 meters, then it will divide the entire space of the image to be detected into three categories: near, medium, and 0-30 meters, 30-50 meters, and 50-70 meters. There is a long-distance image block, and there will be overlap between the three obtained image blocks, so as to ensure that the edge information of the image will not be lost when the image is segmented, and the integrity of the image information will be ensured. It should be noted that since the objects contained in the short distance will occupy a larger pixel position in the image, the lower image blocks (that is, the image blocks with a minimum depth range of 0-30 meters) will be larger.

Optionally, in some scenarios, in order to ensure the consistency between the pixel ratio of the divided image block and the pixel ratio of the undivided image, the above-mentioned execution subject will divide the image to be detected into N*N image blocks, that is, in the image block. The image to be detected is divided into N blocks on the vertical axis, and then divided on the horizontal axis to obtain N*N image blocks, where N is a positive integer.

Step 203 , for each image block in the at least two image blocks, perform target detection on the image block to obtain a detection frame corresponding to the target object in the image block.

In this embodiment, for each image block in the divided at least two image blocks, the execution subject will detect the target object in the image block, so as to obtain a detection frame corresponding to the target object, where the target object may be Arbitrary obstacles, such as vehicles, pedestrians, etc. For example, the above-mentioned executive body will first obtain the depth range information of each image block, and input the image block into the detection model corresponding to the depth range, so that the detection model can detect the target object in the image block, thereby obtaining The detection frame corresponding to the target object. Here, the corresponding detection model is pre-trained for each depth range, that is, for the depth range of 0-30 meters, the training data of the depth range will be collected, and the training data will be used for training to obtain the corresponding detection model. Due to the large distance range covered by the image scene to be detected, it is difficult to directly predict the 3D attributes of all objects in the scene with a single model, and the regression values are often not very accurate. Therefore, in this embodiment, the regression task is converted into a classification task, and the images to be detected are first classified and divided, and then different detection models are used to perform 3D detection respectively, thereby improving the accuracy of the detection results.

Step 204, in response to determining that the same target object corresponds to at least one detection frame, determine a target detection frame of the target object based on the position of the at least one detection frame.

In this embodiment, since there is overlap between different image blocks, and each image block uses a corresponding detection model for detection, there will be multiple different detection frames in the overlapping portion of different image blocks, That is, the same target object will correspond to multiple detection frames. Furthermore, even for the same target object in the same image patch, it may correspond to multiple detection boxes. Based on this, the above-mentioned execution subject determines the target detection frame from the plurality of detection frames for the target object based on the position information of the plurality of detection frames.

For example, if multiple detection frames corresponding to the same target object are in the same image block, the target detection frame may be determined based on the confidence value of the detection frame. If multiple detection frames corresponding to the same target object are in different image blocks, the above-mentioned execution subject will determine the target object from multiple detection frames based on the Non-Maximum Suppression (NMS) method. Object detection box. The NMS algorithm is widely used in target detection scenarios, and its purpose is to eliminate redundant candidate frames and find the best detection frame.

In the target detection method provided by the embodiment of the present disclosure, the image to be detected is first acquired; then the image to be detected is divided into at least two image blocks based on the depth information; and then for each image block in the at least two image blocks, the image block is Perform target detection to obtain a detection frame corresponding to the target object in the image block; finally, in response to determining that the same target object corresponds to at least one detection frame, the target detection frame of the target object is determined based on the position of the at least one detection frame. In the target detection method in this embodiment, the method first divides the image to be detected according to the depth range, and then performs detection in different distance ranges, thereby improving the accuracy of 3D target detection and the accuracy of 3D target detection results. .

In the technical solutions of the present disclosure, the collection, storage, use, processing, transmission, provision, and disclosure of the user's personal information involved are all in compliance with relevant laws and regulations, and do not violate public order and good customs.

Continuing to refer to FIG. 3 , FIG. 3 illustrates a flow 300 of another embodiment of a target detection method according to the present disclosure. The target detection method includes the following steps:

Step 301, acquiring an image to be detected.

Step 302: Divide the image to be detected into at least two image blocks based on the depth information.

Step 303 , for each image block in the at least two image blocks, perform target detection on the image block to obtain a detection frame corresponding to the target object in the image block.

Steps 301 to 303 are basically the same as steps 201 to 203 of the foregoing embodiment. For a specific implementation manner, reference may be made to the foregoing description of steps 201 to 203, which will not be repeated here.

Step 304, in response to determining that at least one detection frame corresponding to the same target object is in the same image block, determine the confidence of each detection frame, and use the detection frame with the highest confidence value as the target detection frame of the target object.

In this embodiment, if it is determined that at least one detection frame corresponding to the same target object is in the same image block, the execution body of the target detection method (for example, the server 105 shown in FIG. 1 ) will be based on the confidence level of each detection frame value to determine the target detection frame. That is, the above-mentioned executive body will obtain the confidence value of each detection frame, and then use the detection frame with the highest confidence value as the target detection frame of the target object. When the detection model is used to detect the image block, the detection model also outputs the confidence value of each detection frame, and the above-mentioned execution subject can directly obtain the confidence value of each detection frame, so as to determine the target detection frame based on the confidence value. Thereby ensuring the accuracy of the target detection frame.

Step 305 , in response to determining that at least one detection frame corresponding to the same target object is in different overlapping image blocks, determine a target detection frame of the target object from the at least one detection frame based on the non-maximum suppression method.

In this embodiment, if it is determined that at least one detection frame corresponding to the same target object is in different overlapping image blocks, the execution subject will determine the target object from multiple detection frames based on the non-maximum value suppression method target detection frame. The NMS algorithm is widely used in target detection scenarios, and its purpose is to eliminate redundant candidate frames and find the best detection frame. Thereby ensuring the accuracy of the target detection frame.

In some optional implementations of this embodiment, step 305 includes: for each detection frame in the at least one detection frame, based on the depth information of the image block where the detection frame is located and the position information of the detection frame in the image block, determine The distance weight of the detection frame; the score of the detection frame is determined based on the confidence of the detection frame and the distance weight; the detection frame with the highest score in at least one detection frame is used as the target detection frame of the target object.

In this implementation manner, since multiple detection frames corresponding to the same target object are in multiple overlapping image blocks, the above-mentioned execution subject will first determine the depth information of the image block where each detection frame is located, and then determine the target object. The position information of the detection frame in the image block, and the distance weight is set for the detection frame based on the depth information and position information. Generally, since the perception ability of the near distance is relatively more accurate, a higher distance weight will be set for the detection frame that is close to the distance and accounts for a larger proportion. Then multiply the confidence value of the detection box by the distance weight to get the final score of the detection box. Finally, the detection frame with the highest score in at least one detection frame is used as the target detection frame of the target object. Therefore, it can be ensured that the determined target detection frame has a stronger detection capability.

As an example, suppose that the image Y is divided according to the depth information to obtain a close-range image block A and a medium-distance image block B, the close-range image block A and the intermediate-distance image block B overlap, and the target object X is in the close-range image block A. The detection frame in A1 is A1, the proportion of A1 in A is 60%, and the confidence level of A1 is 0.9. The detection frame of the target object X in the mid-range image block B is B1, the proportion of B1 in B is 30%, and the confidence level of B1 is 0.6. Since the image block where A1 is located is closer and A1 occupies a larger proportion, the above execution subject will set the distance weight of A1 to 1.2 and the distance weight of B1 to 1.0. Therefore, it can be calculated that the score of A1 is 1.2*0.9=1.08, and the score of B1 is 1.0*0.6=0.6. Therefore, the above executive body will use A1 as the target detection frame of the target object X.

As can be seen from FIG. 3 , compared with the embodiment corresponding to FIG. 2 , the target detection method in this embodiment highlights the process of determining the target detection frame, thereby ensuring that the determined target detection frame has better The detection ability improves the accuracy of target detection results.

Continuing to refer to FIG. 4 , FIG. 4 shows a flow 400 of yet another embodiment of the target detection method according to the present disclosure. The target detection method includes the following steps:

Step 401, acquiring an image to be detected.

Step 402: Input the image to be detected into the division model, and output at least two divided image blocks.

In this embodiment, the execution body of the target detection method (for example, the server 105 shown in FIG. 1 ) will input the image to be detected into the division model, and output at least two divided image blocks, wherein the division model is used for The image to be detected is divided into at least two image blocks according to different depth information. That is, in this embodiment, the division model is pre-trained, so that the division model can divide the image to be detected into multiple image blocks according to the depth information. Since the division model has the ability to divide the image according to the depth information interval, using the division model for division can improve the division efficiency and ensure the rationality of the division result.

In some optional implementations of this embodiment, the division model is obtained by training based on the following steps: acquiring a training sample set, wherein the training samples in the training sample set include sample images and sub-images corresponding to the sample images, and the sub-images are pairs of samples The image is obtained by dividing it according to the depth information; taking the sample image as input, and using the sub-image corresponding to the sample image as output, train the initial division model, and obtain the trained division model.

In this implementation manner, the above-mentioned execution body will first obtain a sample image and a plurality of sub-images corresponding to each sample image, and use the sample image and the sub-images corresponding to the sample image as training samples to obtain a training sample set. The sample image is divided into multiple sub-images according to different depth information. After that, the above-mentioned execution subject will use the sample image as input, and use the sub-image corresponding to the sample image as output to train the initial division model, thereby obtaining the trained division model. The division model trained in the above manner can quickly and accurately divide the image to be detected according to the depth range.

Step 403, for each image block in the at least two image blocks, determine the depth information of the image block.

In this embodiment, for each image block in the at least two image blocks, the above-mentioned execution body will first determine the depth information of the image block. Since when the image to be detected is divided, it is divided according to the depth information, and here the above-mentioned execution subject will obtain the depth information when each image block is divided.

Step 404: Input the image block into the detection model corresponding to the depth information, and output the detection frame corresponding to the target object in the image block.

In this embodiment, after determining the depth information of each image block, the above-mentioned execution body will determine the detection model corresponding to the depth information, and input the image block into the detection model, so as to obtain the corresponding target object in the image block. the detection frame. Here, multiple detection models corresponding to depth information will be pre-trained, so that each image block can be accurately detected and the accuracy of the detection results can be improved.

In some optional implementations of this embodiment, the detection model is obtained by training based on the following steps: acquiring a training data set, wherein the training data in the training data set includes the original image and the detection frame corresponding to the original image, and the training data in the training data set includes the original image and the detection frame corresponding to the original image. The data has the same depth information, and the detection frame is obtained by labeling the target object in the original image; the original image is used as the input, the detection frame corresponding to the original image is used as the output, the initial detection model is trained, and the trained detection model is obtained.

In this implementation, the above-mentioned execution subject will first obtain the original image and the detection frame corresponding to each original image, and use the original image and the detection frame corresponding to the original image as training data to obtain a training data set, where manual annotation can be used. method to annotate the target object in the original image, so as to obtain the corresponding detection frame. After that, the above-mentioned execution subject will use the original image as input, and use the detection frame corresponding to the original image as output to train the initial detection model, thereby obtaining the trained detection model. The detection model trained in the above manner can quickly and accurately detect the target object in the image block.

Step 405, in response to determining that at least one detection frame corresponding to the same target object is in the same image block, calculate the confidence of each detection frame respectively, and use the detection frame with the highest confidence value as the target detection frame of the target object.

Step 406 , in response to determining that at least one detection frame corresponding to the same target object is in different overlapping image blocks, determine a target detection frame of the target object from the at least one detection frame based on the non-maximum suppression method.

Steps 405-406 are basically the same as steps 304-305 in the foregoing embodiment. For a specific implementation manner, reference may be made to the foregoing description of steps 304-305, and details are not repeated here.

As can be seen from FIG. 4 , compared with the embodiment corresponding to FIG. 3 , the target detection method in this embodiment highlights the steps of dividing the image to be detected and determining the detection frame corresponding to the target object in the image block. , which further improves the efficiency and accuracy of 3D target detection.

Further referring to FIG. 5 , FIG. 5 shows an application scenario diagram of the target detection method according to the present disclosure. In this application scenario, the execution subject will input the image to be detected into the distance segmentation network, so that the distance segmentation network divides the image to be detected according to the depth information, and obtains divided 3*3 image blocks, these 3* The 3 image blocks include three image blocks at close range, three image blocks at medium distance, and three image blocks at long distance, and these 3*3 image blocks overlap each other, so as to ensure that the edge information of the image is not changed when dividing the image. will be lost, and it can be seen that the lower image blocks will be larger because the obstacles contained in the close distance will occupy a larger pixel position in the image.

After that, the above-mentioned executive body will input each image block into the corresponding perceptual network, so as to obtain the detection frame corresponding to the target object in each image block. For example, inputting three image blocks at a short distance into the short-range sensing network, inputting three image blocks at a medium distance into the middle-distance sensing network, and inputting three image blocks at a long distance into the long-distance sensing network.

Finally, the distance perception weighting is performed based on the NMS strategy, that is, since the near perception ability is relatively more accurate, a higher weight value will be set for the detection frame of the near distance, and then all the 3D detection frames obtained by the distance network will be used. NMS is performed together, that is, the final target detection frame is determined based on the weight and confidence value of each 3D detection frame.

Further referring to FIG. 6 , as an implementation of the methods shown in the above figures, the present disclosure provides an embodiment of a target detection apparatus, the apparatus embodiment corresponds to the method embodiment shown in FIG. 2 , and the apparatus may specifically Used in various electronic devices.

As shown in FIG. 6 , the target detection apparatus 600 in this embodiment includes: an acquisition module 601 , a division module 602 , a detection module 603 and a determination module 604 . Wherein, the acquiring module 601 is configured to acquire the image to be detected; the dividing module 602 is configured to divide the image to be detected into at least two image blocks based on the depth information, wherein there is overlap between the at least two image blocks; detecting The module 603 is configured to perform target detection on the image blocks for each image block in the at least two image blocks to obtain a detection frame corresponding to the target object in the image blocks; the determining module 604 is configured to respond to determining the same The target object corresponds to at least one detection frame, and the target detection frame of the target object is determined based on the position of the at least one detection frame.

In this embodiment, in the target detection device 600: the specific processing of the acquisition module 601, the division module 602, the detection module 603, and the determination module 604 and the technical effects brought about by the acquiring module 601, the division module 602, and the technical effect brought by them may refer to step 201 in the corresponding embodiment of FIG. 2, respectively. The related description of -204 will not be repeated here.

In some optional implementations of this embodiment, the determining module includes: a first determining sub-module, configured to, in response to determining that at least one detection frame corresponding to the same target object is in the same image block, determine each detection frame The confidence level of the frame, the detection frame with the highest confidence value is used as the target detection frame of the target object; the second determination sub-module is configured to respond to determining that at least one detection frame corresponding to the same target object is in a different overlapping area In the image block, the target detection frame of the target object is determined from at least one detection frame based on the non-maximum value suppression method.

In some optional implementations of this embodiment, the second determination submodule is further configured to: for each detection frame in the at least one detection frame, based on the depth information of the image block where the detection frame is located and the detection frame in the image block Determine the distance weight of the detection frame based on the location information in the detection frame; determine the score of the detection frame based on the confidence of the detection frame and the distance weight; take the detection frame with the highest score in at least one detection frame as the target detection frame of the target object.

In some optional implementations of this embodiment, the division module includes: a division sub-module, configured to input the image to be detected into the division model, and output at least two divided image blocks, wherein the division model uses The image to be detected is divided into at least two image blocks according to different depth information.

In some optional implementations of this embodiment, the above-mentioned target detection apparatus 600 further includes a first training module for training the division model, and the first training module is configured to: acquire a training sample set, wherein the The training sample includes the sample image and the sub-image corresponding to the sample image. The sub-image is obtained by dividing the sample image according to the depth information; the sample image is used as the input, and the sub-image corresponding to the sample image is used as the output to train the initial division model, and the training is completed. division model.

In some optional implementations of this embodiment, the detection module includes: a third determination submodule, configured to determine depth information of the image block; and a detection submodule, configured to input the image block into the detection corresponding to the depth information In the model, the output is the detection frame corresponding to the target object in the image block.

In some optional implementations of this embodiment, the above-mentioned target detection apparatus 600 is used for training a second training module of the detection model, and the second training module is configured to: acquire a training data set, wherein the training data in the training data set Including the original image and the detection frame corresponding to the original image, the training data in the training data set has the same depth information, and the detection frame is obtained by marking the target object in the original image; taking the original image as input, the detection frame corresponding to the original image As the output, the initial detection model is trained, and the trained detection model is obtained.

According to embodiments of the present disclosure, the present disclosure also provides an electronic device, a readable storage medium, and a computer program product.

FIG. 7 shows a schematic block diagram of an example electronic device 700 that may be used to implement embodiments of the present disclosure. Electronic devices are intended to represent various forms of digital computers, such as laptops, desktops, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. Electronic devices may also represent various forms of mobile devices, such as personal digital processors, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are by way of example only, and are not intended to limit implementations of the disclosure described and/or claimed herein.

As shown in FIG. 7 , the device 700 includes a computing unit 701 that can be executed according to a computer program stored in a read only memory (ROM) 702 or loaded into a random access memory (RAM) 703 from a storage unit 708 Various appropriate actions and handling. In the RAM 703, various programs and data necessary for the operation of the device 700 can also be stored. The computing unit 701 , the ROM 702 , and the RAM 703 are connected to each other through a bus 704 . An input/output (I/O) interface 705 is also connected to bus 704 .

Various components in the device 700 are connected to the I/O interface 705, including: an input unit 706, such as a keyboard, mouse, etc.; an output unit 707, such as various types of displays, speakers, etc.; a storage unit 708, such as a magnetic disk, an optical disk, etc. ; and a communication unit 709, such as a network card, a modem, a wireless communication transceiver, and the like. The communication unit 709 allows the device 700 to exchange information/data with other devices through a computer network such as the Internet and/or various telecommunication networks.

Computing unit 701 may be various general-purpose and/or special-purpose processing components with processing and computing capabilities. Some examples of computing units 701 include, but are not limited to, central processing units (CPUs), graphics processing units (GPUs), various specialized artificial intelligence (AI) computing chips, various computing units that run machine learning model algorithms, digital signal processing processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 701 performs the various methods and processes described above, such as the target detection method. For example, in some embodiments, the object detection method may be implemented as a computer software program tangibly embodied on a machine-readable medium, such as storage unit 708 . In some embodiments, part or all of the computer program may be loaded and/or installed on device 700 via ROM 702 and/or communication unit 709 . When a computer program is loaded into RAM 703 and executed by computing unit 701, one or more steps of the object detection method described above may be performed. Alternatively, in other embodiments, the computing unit 701 may be configured to perform the object detection method by any other suitable means (eg, by means of firmware).

Various implementations of the systems and techniques described herein above may be implemented in digital electronic circuitry, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips system (SOC), complex programmable logic device (CPLD), computer hardware, firmware, software, and/or combinations thereof. These various embodiments may include being implemented in one or more computer programs executable and/or interpretable on a programmable system including at least one programmable processor that The processor, which may be a special purpose or general-purpose programmable processor, may receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device an output device.

Program code for implementing the methods of the present disclosure may be written in any combination of one or more programming languages. These program codes may be provided to a processor or controller of a general purpose computer, special purpose computer or other programmable data processing apparatus, such that the program code, when executed by the processor or controller, performs the functions/functions specified in the flowcharts and/or block diagrams. Action is implemented. The program code may execute entirely on the machine, partly on the machine, partly on the machine and partly on a remote machine as a stand-alone software package or entirely on the remote machine or server.

In the context of the present disclosure, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in connection with the instruction execution system, apparatus or device. The machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. Machine-readable media may include, but are not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media would include one or more wire-based electrical connections, portable computer disks, hard disks, random access memory (RAM), read only memory (ROM), erasable programmable read only memory (EPROM or flash memory), fiber optics, compact disk read only memory (CD-ROM), optical storage, magnetic storage, or any suitable combination of the foregoing.

To provide interaction with a user, the systems and techniques described herein may be implemented on a computer having a display device (eg, a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user ); and a keyboard and pointing device (eg, a mouse or trackball) through which a user can provide input to the computer. Other kinds of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (eg, visual feedback, auditory feedback, or tactile feedback); and can be in any form (including acoustic input, voice input, or tactile input) to receive input from the user.

The systems and techniques described herein may be implemented on a computing system that includes back-end components (eg, as a data server), or a computing system that includes middleware components (eg, an application server), or a computing system that includes front-end components (eg, a user's computer having a graphical user interface or web browser through which a user may interact with implementations of the systems and techniques described herein), or including such backend components, middleware components, Or any combination of front-end components in a computing system. The components of the system may be interconnected by any form or medium of digital data communication (eg, a communication network). Examples of communication networks include: Local Area Networks (LANs), Wide Area Networks (WANs), and the Internet.

Cloud computing refers to accessing an elastically scalable shared physical or virtual resource pool through a network. Resources can include servers, operating systems, networks, software, applications or storage devices, etc. A technical system for deploying and managing resources in the form of services. Through cloud computing technology, it can provide efficient and powerful data processing capabilities for artificial intelligence, blockchain and other technical applications and model training.

A computer system can include clients and servers. Clients and servers are generally remote from each other and usually interact through a communication network. The relationship of client and server arises by computer programs running on the respective computers and having a client-server relationship to each other. The server can be a cloud server, a distributed system server, or a server combined with blockchain.

It should be understood that steps may be reordered, added or deleted using the various forms of flow shown above. For example, the steps described in the present disclosure can be executed in parallel, sequentially, or in different orders. As long as the desired results of the technical solutions disclosed in the present disclosure can be achieved, there is no limitation herein.

The above-mentioned specific embodiments do not constitute a limitation on the protection scope of the present disclosure. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions may occur depending on design requirements and other factors. Any modifications, equivalent replacements, and improvements made within the spirit and principles of the present disclosure should be included within the protection scope of the present disclosure.

Claims (17)

1. A method of target detection, comprising:

acquiring an image to be detected;

dividing the image to be detected into at least two image blocks based on the depth information, wherein the at least two image blocks are overlapped;

performing target detection on each image block of the at least two image blocks to obtain a detection frame corresponding to a target object in the image block;

in response to determining that the same target object corresponds to at least one detection frame, determining a target detection frame of the target object based on a position of the at least one detection frame.

2. The method of claim 1, wherein, in response to determining that the same target object corresponds to at least one detection box, determining the target detection box of the target object based on a position of the at least one detection box comprises:

in response to the fact that at least one detection frame corresponding to the same target object is located in the same image block, determining the confidence of each detection frame, and taking the detection frame with the highest confidence value as the target detection frame of the target object;

and in response to determining that at least one detection frame corresponding to the same target object is in different image blocks with overlapping, determining a target detection frame of the target object from the at least one detection frame based on a non-maximum suppression method.

3. The method of claim 2, wherein the determining a target detection box for the target object from the at least one detection box based on a non-maximum suppression method comprises:

for each detection frame in the at least one detection frame, determining a distance weight of the detection frame based on depth information of an image block where the detection frame is located and position information of the detection frame in the image block; determining a score for the detection box based on the confidence of the detection box and the distance weight;

and taking the detection frame with the highest score in the at least one detection frame as the target detection frame of the target object.

4. The method of claim 1, wherein the dividing the image to be detected into at least two image blocks based on the depth information comprises:

and inputting the image to be detected into a division model, and outputting to obtain at least two divided image blocks, wherein the division model is used for dividing the image to be detected into the at least two image blocks according to different depth information.

5. The method of claim 4, wherein the partition model is trained based on the following steps:

acquiring a training sample set, wherein training samples in the training sample set comprise sample images and sub-images corresponding to the sample images, and the sub-images are obtained by dividing the sample images according to depth information;

and taking the sample image as input, taking the sub-image corresponding to the sample image as output, training an initial division model, and obtaining a division model after training.

6. The method according to claim 1, wherein the performing target detection on the image block to obtain a detection frame corresponding to a target object in the image block includes:

determining depth information of the image block;

and inputting the image block into a detection model corresponding to the depth information, and outputting to obtain a detection frame corresponding to a target object in the image block.

7. The method of claim 6, wherein the detection model is trained based on:

acquiring a training data set, wherein training data in the training data set comprises an original image and a detection frame corresponding to the original image, the training data in the training data set has the same depth information, and the detection frame is obtained by labeling a target object in the original image;

and taking the original image as input, taking a detection frame corresponding to the original image as output, training an initial detection model, and obtaining a trained detection model.

8. An object detection device comprising:

an acquisition module configured to acquire an image to be detected;

a dividing module configured to divide the image to be detected into at least two image blocks based on depth information, wherein there is an overlap between the at least two image blocks;

the detection module is configured to perform target detection on each image block of the at least two image blocks to obtain a detection frame corresponding to a target object in the image block;

the determining module is configured to determine a target detection frame of the target object based on a position of at least one detection frame in response to determining that the same target object corresponds to the at least one detection frame.

9. The apparatus of claim 8, wherein the means for determining comprises:

the first determining sub-module is configured to determine the confidence of each detection frame in response to determining that at least one detection frame corresponding to the same target object is in the same image block, and take the detection frame with the highest confidence value as the target detection frame of the target object;

and the second determining submodule is configured to determine a target detection frame of the target object from at least one detection frame corresponding to the same target object based on a non-maximum suppression method in response to determining that the at least one detection frame is located in different image blocks with overlapping.

10. The apparatus of claim 9, wherein the second determination submodule is further configured to:

for each detection frame in the at least one detection frame, determining a distance weight of the detection frame based on depth information of an image block where the detection frame is located and position information of the detection frame in the image block; determining a score for the detection box based on the confidence of the detection box and the distance weight;

and taking the detection frame with the highest score in the at least one detection frame as the target detection frame of the target object.

11. The apparatus of claim 8, wherein the means for dividing comprises:

and the division submodule is configured to input the image to be detected into a division model and output the image to be detected to obtain at least two divided image blocks, wherein the division model is used for dividing the image to be detected into the at least two image blocks according to different depth information.

12. The apparatus of claim 11, wherein the apparatus further comprises a first training module for training a partitioning model, the first training module configured to:

acquiring a training sample set, wherein training samples in the training sample set comprise sample images and sub-images corresponding to the sample images, and the sub-images are obtained by dividing the sample images according to depth information;

and taking the sample image as input, taking the sub-image corresponding to the sample image as output, and training an initial division model to obtain a trained division model.

13. The apparatus of claim 8, wherein the detection module comprises:

a third determining sub-module configured to determine depth information of the image block;

and the detection sub-module is configured to input the image block into a detection model corresponding to the depth information, and output a detection frame corresponding to a target object in the image block.

14. The apparatus of claim 13, wherein the apparatus further comprises a second training module for training a detection model, the second training module configured to:

acquiring a training data set, wherein training data in the training data set comprises an original image and a detection frame corresponding to the original image, the training data in the training data set has the same depth information, and the detection frame is obtained by labeling a target object in the original image;

and taking the original image as input, taking a detection frame corresponding to the original image as output, training an initial detection model, and obtaining a trained detection model.

15. An electronic device, comprising:

at least one processor; and

a memory communicatively coupled to the at least one processor; wherein,

the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the method of any one of claims 1-7.

16. A non-transitory computer readable storage medium having stored thereon computer instructions for causing the computer to perform the method of any one of claims 1-7.

17. A computer program product comprising a computer program which, when executed by a processor, implements the method according to any one of claims 1-7.

Images