Low-altitude target identification method and device

Abstract

The invention relates to the technical field of target recognition, and provides a low-altitude target recognition method and device. The method comprises the following steps: storing basic information of a plurality of preset targets in a target low-altitude region in a database in a classified manner; acquiring image data and sound data of a to-be-recognized target acquired by an aircraft; extracting target image features from the image data, and extracting target sound features from the sound data; inputting the target image features and the target sound features into a target detection model to obtain the probability that the to-be-recognized target exists in a database; and when the probability is greater than a probability threshold value, identifying a classification type and basic information corresponding to the to-be-identified target from a database. According to the method, the multi-dimensional features of the image and the sound serve as input, target recognition is performed through the historical multi-dimensional feature training model, the target recognition precision is greatly improved, and after it is recognized that the target exists in the database, the classification type to which the target belongs and the basic information of the target can be further recognized.

Description

Technical Field

[0001] The present application relates to the field of target recognition technology, and in particular to a method and device for low-altitude target recognition. Background Art

[0002] Low-altitude intelligent networked aircraft can complete inspection tasks over large areas in a short period of time, greatly shortening the inspection cycle and improving work efficiency. Taking drones as an example, during power inspections, they can quickly identify problems in transmission lines and substations, reduce the time and cost of manual inspections, avoid manual operations in dangerous environments such as high altitudes, high temperatures, and high pressures, and reduce safety risks for workers. Especially in emergency rescues, drones can quickly arrive at the scene of the accident, provide first-hand information, and improve rescue efficiency. In photovoltaic inspections, they are equipped with high-definition cameras and different mounting devices, which can transmit photovoltaic power station inspection images and data in real time, conduct timely and accurate analysis and management of inspection data, automatically identify and report abnormal situations, and improve operation and maintenance efficiency and safety.

[0003] However, traditional low-altitude intelligent networked aircraft have low recognition accuracy for inspection targets during patrols, and are prone to recognition errors or omissions, which cannot meet the accuracy requirements of current patrol tasks. Summary of the invention

[0004] The embodiments of the present application provide a low-altitude target recognition method and device to solve the technical problem that traditional low-altitude intelligent networked aircraft have low recognition accuracy for patrol targets during patrols, are prone to recognition errors or recognition omissions, and cannot meet the accuracy requirements of current patrol tasks.

[0005] In a first aspect, an embodiment of the present application provides a low-altitude target recognition method, comprising: Classify and store basic information of multiple preset targets in the target low-altitude area into a database; the basic information includes the identification, location and real-time status of the preset targets; Acquiring image data and sound data of the target to be identified collected by the aircraft when flying in the target low-altitude area; extracting target image features from the image data, and extracting target sound features from the sound data; Inputting the target image feature and the target sound feature into a target detection model to obtain the probability that the target to be identified outputted by the target detection model exists in the database; When the probability is greater than the probability threshold, identifying the classification type and basic information corresponding to the target to be identified from the database; The target detection model is obtained by training the historical target image features, the historical target sound data and the labels of the plurality of preset targets on the basis of the convolution function.

[0006] In one embodiment, the step of acquiring image data and sound data of the target to be identified collected by the aircraft when flying in the target low-altitude area includes: Acquiring light signals, sound signals and environmental pressure signals of the target to be identified collected by the aircraft when flying in the target low-altitude area; fusing the optical signal and the ambient pressure signal into a first electrical signal; fusing the sound signal and the environmental pressure signal into a second electrical signal; generating image data of the target to be identified based on the first electrical signal; The sound data of the target to be identified is generated based on the second electrical signal.

[0007] In one embodiment, extracting target image features from the image data includes: Standardizing and normalizing the image data to obtain preprocessed image data; extracting a plurality of image features from the preprocessed image data; A target image feature representative of the image data is screened out from the multiple image features.

[0008] In one embodiment, the step of selecting a target image feature representative of the image data from the plurality of image features includes: Randomly splicing the multiple image features to obtain multiple spliced ​​image features; Obtaining a distribution probability of the data of each of the stitched image features in the preprocessed image data to obtain a plurality of first probabilities; Obtaining the distribution probability of the data of each image feature in the data of the spliced ​​image feature to which it belongs, to obtain a plurality of second probabilities; Multiplying the corresponding first and second probabilities to obtain a plurality of third probabilities; The third probabilities are sorted from large to small, and the image features corresponding to the third probabilities with the highest sorting are determined as the target image features.

[0009] In one embodiment, extracting the target sound feature from the sound data includes: Improving the signal-to-noise ratio of the sound data in the high frequency part to obtain preprocessed sound data; Dividing the pre-processed sound data into a plurality of sound frames, and performing a windowing operation on each of the sound frames to obtain a plurality of pre-processed sound frames; Performing a fast Fourier transform on each of the pre-processed sound frames to obtain a sound spectrum graph; Converting the sound spectrogram into a Mel-spectrogram; Taking the logarithm of the Mel-spectrogram to obtain a logarithmic spectrum graph; Performing discrete cosine transform on the logarithmic spectrum to obtain Mel-frequency cepstrum coefficients; Low-order coefficients in the Mel-frequency cepstral coefficients are obtained, and sound data features corresponding to the low-order coefficients are determined as the target sound features.

[0010] In one embodiment, the target image features include color features, shape features, and texture features; The target sound features include pitch features, loudness features and timbre features.

[0011] In a second aspect, an embodiment of the present application provides a low-altitude target recognition device, comprising: An information classification module is used to: classify and store basic information of multiple preset targets in the target low-altitude area into a database; the basic information includes the identification, location and real-time status of the preset targets; The data acquisition module is used to: acquire image data and sound data of the target to be identified collected by the aircraft when flying in the target low-altitude area; A target feature extraction module, used to extract target image features from the image data and target sound features from the sound data; A first target recognition module is used to: input the target image feature and the target sound feature into a target detection model to obtain the probability that the target to be recognized output by the target detection model exists in the database; A second target recognition module is used to: when the probability is greater than the probability threshold, identify the classification type and basic information corresponding to the target to be recognized from the database; The target detection model is obtained by training the historical target image features, the historical target sound data and the labels of the plurality of preset targets on the basis of the convolution function.

[0012] In a third aspect, an embodiment of the present application provides an electronic device, comprising a processor and a memory storing a computer program, wherein when the processor executes the program, the steps of the low-altitude target identification method described in the first aspect are implemented.

[0013] In a fourth aspect, an embodiment of the present application provides a computer program product, including a computer program, which, when executed by a processor, implements the steps of the low-altitude target identification method described in the first aspect.

[0014] In a fifth aspect, an embodiment of the present application provides a non-transitory computer-readable storage medium, including a computer program, which, when executed by a processor, implements the steps of the low-altitude target identification method described in the first aspect.

[0015] The low-altitude target recognition method and device provided by the present application classify and store the basic information of multiple preset targets in the target low-altitude area into a database, wherein the basic information includes the identification, location and real-time status of the preset targets, obtains the image data and sound data of the target to be identified collected by the aircraft when flying in the target low-altitude area, extracts the target image features from the image data, and extracts the target sound features from the sound data, inputs the target image features and the target sound features into the target detection model, obtains the probability that the target to be identified output by the target detection model exists in the database, and when the probability is greater than the probability threshold, identifies the classification type and basic information corresponding to the target to be identified from the database, and the target detection model is obtained based on the convolution function through historical target image features, historical target sound data and label training of multiple preset targets. The present application simultaneously obtains the image data and sound data of the target to be identified during the aircraft inspection process, performs model recognition based on the target features, and determines whether the target to be identified exists in the database. That is, the multi-dimensional features of the image and sound are used as input, and the target is recognized by a model trained with historical multi-dimensional features. This can greatly improve the accuracy of target recognition. Moreover, since the basic information of the preset target has been classified in advance, after identifying that the target exists in the database, the classification type to which the target belongs and its corresponding basic information can be further identified. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the present application or the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0017] Figure 1 This is one of the flow charts of the low-altitude target recognition method provided in the embodiment of the present application; Figure 2 This is the second flow chart of the low-altitude target recognition method provided in the embodiment of the present application; Figure 3 This is the third flow chart of the low-altitude target recognition method provided in the embodiment of the present application; Figure 4 This is the fourth flow chart of the low-altitude target recognition method provided in the embodiment of the present application; Figure 5 It is a structural schematic diagram of a low-altitude target recognition device provided in an embodiment of the present application; Figure 6 It is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0018] In order to make the purpose, technical solutions and advantages of this application clearer, the technical solutions in this application will be clearly and completely described below in conjunction with the drawings in the embodiments of this application. Obviously, the described embodiments are part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0019] It should be noted that in the description of the embodiments of the present application, the terms "include", "comprise" or any other variant thereof are intended to cover non-exclusive inclusion, so that the process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "including one..." do not exclude the existence of other identical elements in the process, method, article or device including the elements. The orientation or position relationship indicated by the terms "upper", "lower" and the like is based on the orientation or position relationship shown in the drawings, which is only for the convenience of describing the present application and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present application. Unless otherwise clearly specified and limited, the terms "installed", "connected" and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be directly connected, or indirectly connected through an intermediate medium, or it can be a connection between the two elements. For those skilled in the art, the specific meanings of the above terms in this application can be understood according to specific circumstances.

[0020] The terms "first", "second", etc. in this application are used to distinguish similar objects, and are not used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present application can be implemented in an order other than those illustrated or described here, and the objects distinguished by "first", "second", etc. are generally of one type, and the number of objects is not limited. For example, the first object can be one or more. In addition, "and / or" represents at least one of the connected objects, and the character " / " generally indicates that the objects associated with each other are in an "or" relationship.

[0021] Figure 1This is one of the flow charts of the low-altitude target recognition method provided in the embodiment of the present application. Figure 1 , the present application embodiment provides a low-altitude target recognition method, which may include: 101. Classify and store basic information of multiple preset targets in the target low-altitude area into a database; Basic information includes the identification, location and real-time status of the preset target; 102. Acquire image data and sound data of the target to be identified collected when the aircraft flies in the target low-altitude area; 103. Extracting target image features from the image data, and extracting target sound features from the sound data; 104. Input the target image features and the target sound features into the target detection model to obtain the probability that the target to be identified output by the target detection model exists in the database; 105. When the probability is greater than the probability threshold, the classification type and basic information corresponding to the target to be identified are identified from the database.

[0022] The target detection model is trained based on the convolution function through historical target image features, historical target sound data and labels of multiple preset targets.

[0023] In step 101, any clustering method may be used to classify the basic information of the plurality of preset targets to obtain a plurality of clusters, and each cluster is stored in a database. In this embodiment, a k-means algorithm may be used to classify the basic information of the plurality of preset targets.

[0024] In step 102, the image data and sound data of the target to be identified may be collected by devices such as cameras, image sensors, and sound sensors installed on the aircraft.

[0025] In step 103, the target image features may include color features, shape features, and texture features, and the target sound features may include pitch features, loudness features, and timbre features.

[0026] In step 105, when the probability is greater than the probability threshold, it can be determined that the target to be identified exists in the database, so it can be further identified to which classification type it belongs and its corresponding basic information.

[0027] The low-altitude target recognition method provided in this embodiment classifies and stores basic information of multiple preset targets in a target low-altitude area into a database, where the basic information includes the identification, location and real-time status of the preset targets, obtains image data and sound data of the target to be recognized collected when the aircraft flies in the target low-altitude area, extracts target image features from the image data, and extracts target sound features from the sound data, inputs the target image features and the target sound features into a target detection model, obtains the probability that the target to be recognized output by the target detection model exists in the database, and when the probability is greater than the probability threshold, identifies the classification type and basic information corresponding to the target to be recognized from the database, and the target detection model is obtained based on the convolution function through historical target image features, historical target sound data and label training of multiple preset targets. This embodiment simultaneously obtains the image data and sound data of the target to be identified during the aircraft inspection process, performs model recognition based on the target features, and determines whether the target to be identified exists in the database. That is, the multi-dimensional features of the image and sound are used as input, and the target is recognized by a model trained with historical multi-dimensional features. This can greatly improve the accuracy of target recognition. Moreover, since the basic information of the preset target is classified in advance, after it is recognized that the target exists in the database, the classification type to which the target belongs and its corresponding basic information can be further identified.

[0028] Figure 2 This is the second flow chart of the low-altitude target recognition method provided in the embodiment of the present application. Figure 2 In one embodiment, obtaining image data and sound data of a target to be identified collected by an aircraft when flying in a target low-altitude area may include: 201. Acquire the light signal, sound signal and environmental pressure signal of the target to be identified collected by the aircraft when flying in the target low-altitude area; 202. Fusing the optical signal and the environmental pressure signal into a first electrical signal; 203. Fusing the sound signal and the environmental pressure signal into a second electrical signal; 204. Generate image data of the target to be identified based on the first electrical signal; 205. Generate sound data of the target to be identified based on the second electrical signal.

[0029] In step 201, since the propagation of the optical signal and the sound signal are both affected by the environmental pressure signal, it is necessary to collect the environmental pressure signal of the target to be identified while collecting the optical signal and the sound signal of the target to be identified, so as to further process this influence later.

[0030] In step 202, the optical signal and the ambient pressure signal are fused and converted into a first electrical signal. On the one hand, the fusion process can perform error correction on the optical signal based on the ambient pressure signal to obtain a more accurate optical signal, which helps to obtain more accurate image data later. On the other hand, the converted first electrical signal is easier to process and transmit than the original optical signal and ambient pressure signal.

[0031] In step 203, the sound signal and the ambient pressure signal are fused and converted into a second electrical signal. On the one hand, the fusion process can perform error correction on the sound signal based on the ambient pressure signal to obtain a more accurate sound signal, which helps to obtain more accurate sound data later. On the other hand, the converted second electrical signal is easier to process and transmit than the original sound signal and ambient pressure signal.

[0032] In step 204 to step 205, the image data generated based on the first electrical signal and the sound data generated based on the second electrical signal are easier to control and measure than the original electrical signal.

[0033] This embodiment introduces an ambient pressure signal to correct the light signal and the sound signal, and converts the original signal into an electrical signal, and then generates image data and sound data based on the electrical signal, so that controllable, easy-to-process, easy-to-transmit and easy-to-measure data can be obtained based on accurate signals.

[0034] Figure 3 This is the third flow chart of the low-altitude target recognition method provided in the embodiment of the present application. Figure 3 In one embodiment, extracting target image features from image data may include: 301. Standardize and normalize the image data to obtain preprocessed image data; 302. Extracting a plurality of image features from the preprocessed image data; 303. Filter out target image features that are representative of the image data from multiple image features.

[0035] In step 301, by standardizing and normalizing the image data, the dimension of the image data can be eliminated and the image data can be placed within a specific data range, so that the obtained pre-processed image data is easier to process.

[0036] The specific steps of step 303 may be as follows: 303a. Randomly splicing multiple image features to obtain multiple spliced ​​image features; 303b, obtaining a distribution probability of data of each spliced ​​image feature in the pre-processed image data, and obtaining a plurality of first probabilities; 303c, obtaining a distribution probability of the data of each image feature in the data of the spliced ​​image feature to which it belongs, and obtaining a plurality of second probabilities; 303d, multiplying the corresponding first probability and second probability to obtain a plurality of third probabilities; 303e. Sort the third probabilities from large to small, and determine the image feature corresponding to the third probability with the highest sorting as the target image feature.

[0037] In step 303b to step 303d, the distribution probability of the data of each stitched image feature in the preprocessed image data is first obtained, and then the distribution probability of the data of each image feature in the data of the stitched image feature to which it belongs is obtained, and then the two corresponding distribution probabilities are multiplied to obtain the distribution probability of the data of each image feature in the preprocessed image data.

[0038] In step 303e, the larger the third probability is, the greater the distribution probability of the data corresponding to the image feature in the preprocessed image data is, and the more it can represent the image data, so these image features are determined as target image features.

[0039] It should be noted that the target image features may also be screened by other methods such as principal component analysis and random forest, which are not limited here.

[0040] In this embodiment, the image data is first standardized and normalized to make the pre-processed image data easier to process, and then a variety of image features are extracted therefrom, the multiple image features are randomly spliced, and the distribution probability of the spliced ​​image feature data in the pre-processed image data and the distribution probability of the image feature data in the spliced ​​image feature data are calculated, and then the two distribution probabilities are multiplied, so that the probability distribution of the image feature data in the pre-processed image data can be obtained across the spliced ​​image feature layer. On the one hand, the introduction of randomness can expand the types of spliced ​​image features, making the calculation of distribution probability more accurate and comprehensive. On the other hand, by using the spliced ​​image feature data as a bridge between the image feature data and the pre-processed image data, the accuracy of the distribution probability calculation can be finely controlled based on the step-by-step calculation and synthesis of the distribution probability, and finally the target image features obtained are more representative, thereby achieving the purpose of reducing the redundancy of image features and improving the effectiveness of image features.

[0041] Figure 4 This is the fourth flow chart of the low-altitude target recognition method provided in the embodiment of the present application. Figure 4 In one embodiment, extracting target sound features from sound data may include: 401. Improving the signal-to-noise ratio of the sound data in the high frequency part to obtain preprocessed sound data; 402. Divide the preprocessed sound data into a plurality of sound frames, and perform a windowing operation on each sound frame to obtain a plurality of preprocessed sound frames; 403. Perform fast Fourier transform on each pre-processed sound frame to obtain a sound spectrum diagram; 404. Convert the sound spectrum map into a Mel spectrum map; 405. Taking the logarithm of the Mel-frequency spectrum to obtain a logarithmic frequency spectrum; 406. Perform discrete cosine transform on the logarithmic spectrum to obtain Mel frequency cepstrum coefficients; 407. Obtain low-order coefficients in the Mel-frequency cepstrum coefficients, and determine the sound data features corresponding to the low-order coefficients as target sound features.

[0042] In step 401, the power spectrum of the sound data decreases with the increase of frequency, which will result in a lower signal-to-noise ratio in the high-frequency part. Therefore, a first-order high-pass filter can be used to improve the signal-to-noise ratio of the sound data in the high-frequency part to avoid various numerical problems in the subsequent fast Fourier transform, such as fence effect, frequency leakage, amplitude distortion, poor accuracy, etc.

[0043] In step 402, since the sound data is actually non-stationary, the pre-processed sound data is divided into multiple shorter sound frames to ensure that the data of the sound frames are basically stable, and then a windowing operation is performed on each sound frame to ensure a smooth transition between frames.

[0044] In step 403, the pre-processed sound frame is converted from the time domain to the frequency domain for further analysis.

[0045] In step 404, the sound spectrum may be input into a group of Mel filters. These Mel filters can simulate the nonlinear perception characteristics of human ears to frequency and convert the sound spectrum into a Mel spectrum.

[0046] In step 405, the logarithm of the Mel spectrum is taken to simulate the human ear's perception of sound intensity.

[0047] In step 406 to step 407, the discrete cosine transform can map the high-dimensional logarithmic spectrum to a low-dimensional space, and then select low-order coefficients therefrom to achieve dimensionality reduction of the logarithmic spectrum. Since the sound data features corresponding to the low-order coefficients are the main components in the sound data features, they are determined as the target sound features.

[0048] This embodiment can highlight the main components in the sound data features by processing the high-frequency part of the sound data, dividing the sound frame and adding windows, fast Fourier transform, converting the spectrum graph to take the logarithm and discrete cosine transform, so as to accurately obtain the target sound features and achieve the purpose of reducing the redundancy of sound features and improving the effectiveness of sound features.

[0049] The following is a description of a low-altitude target recognition device provided in an embodiment of the present application. The low-altitude target recognition device described below and the low-altitude target recognition method described above can be referenced to each other.

[0050] Figure 5 Schematic diagram of the structure of the low-altitude target recognition device provided in the embodiment of the present application. Figure 5 , the embodiment of the present application provides a low-altitude target recognition device, which may include: The information classification module 501 is used to classify and store basic information of multiple preset targets in the target low-altitude area into a database; the basic information includes the identification, location and real-time status of the preset targets; The data acquisition module 502 is used to: acquire image data and sound data of the target to be identified collected by the aircraft when flying in the target low-altitude area; A target feature extraction module 503 is used to extract target image features from the image data and target sound features from the sound data; The first target recognition module 504 is used to: input the target image feature and the target sound feature into a target detection model to obtain the probability that the target to be recognized output by the target detection model exists in the database; The second target recognition module 505 is used to: when the probability is greater than the probability threshold, identify the classification type and basic information corresponding to the target to be recognized from the database; The target detection model is obtained by training the historical target image features, the historical target sound data and the labels of the plurality of preset targets on the basis of the convolution function.

[0051] The low-altitude target recognition device provided in this embodiment classifies and stores basic information of multiple preset targets in a target low-altitude area into a database, where the basic information includes the identification, location and real-time status of the preset targets, obtains image data and sound data of the target to be recognized collected when the aircraft flies in the target low-altitude area, extracts target image features from the image data, and extracts target sound features from the sound data, inputs the target image features and the target sound features into a target detection model, obtains the probability that the target to be recognized output by the target detection model exists in the database, and when the probability is greater than the probability threshold, identifies the classification type and basic information corresponding to the target to be recognized from the database, and the target detection model is obtained based on the convolution function through historical target image features, historical target sound data and label training of multiple preset targets. This embodiment simultaneously obtains the image data and sound data of the target to be identified during the aircraft inspection process, performs model recognition based on the target features, and determines whether the target to be identified exists in the database. That is, the multi-dimensional features of the image and sound are used as input, and the target is recognized by a model trained with historical multi-dimensional features. This can greatly improve the accuracy of target recognition. Moreover, since the basic information of the preset target is classified in advance, after it is recognized that the target exists in the database, the classification type to which the target belongs and its corresponding basic information can be further identified.

[0052] In one embodiment, the data collection module 502 is specifically used to: Acquiring light signals, sound signals and environmental pressure signals of the target to be identified collected by the aircraft when flying in the target low-altitude area; fusing the optical signal and the ambient pressure signal into a first electrical signal; fusing the sound signal and the environmental pressure signal into a second electrical signal; generating image data of the target to be identified based on the first electrical signal; The sound data of the target to be identified is generated based on the second electrical signal.

[0053] In one embodiment, the target feature extraction module 503 is specifically used to: Standardizing and normalizing the image data to obtain preprocessed image data; extracting a plurality of image features from the preprocessed image data; A target image feature representative of the image data is screened out from the multiple image features.

[0054] In one embodiment, the target feature extraction module 503 is specifically used to: Randomly splicing the multiple image features to obtain multiple spliced ​​image features; Obtaining a distribution probability of the data of each of the stitched image features in the preprocessed image data to obtain a plurality of first probabilities; Obtaining the distribution probability of the data of each image feature in the data of the spliced ​​image feature to which it belongs, to obtain a plurality of second probabilities; Multiplying the corresponding first and second probabilities to obtain a plurality of third probabilities; The third probabilities are sorted from large to small, and the image features corresponding to the third probabilities with the highest sorting are determined as the target image features.

[0055] In one embodiment, the target feature extraction module 503 is specifically used to: Improving the signal-to-noise ratio of the sound data in the high frequency part to obtain preprocessed sound data; Dividing the pre-processed sound data into a plurality of sound frames, and performing a windowing operation on each of the sound frames to obtain a plurality of pre-processed sound frames; Performing a fast Fourier transform on each of the pre-processed sound frames to obtain a sound spectrum graph; Converting the sound spectrogram into a Mel-spectrogram; Taking the logarithm of the Mel-spectrogram to obtain a logarithmic spectrum graph; Performing discrete cosine transform on the logarithmic spectrum to obtain Mel-frequency cepstrum coefficients; Low-order coefficients in the Mel-frequency cepstral coefficients are obtained, and sound data features corresponding to the low-order coefficients are determined as the target sound features.

[0056] In one embodiment, the target image features include color features, shape features, and texture features; The target sound features include pitch features, loudness features and timbre features.

[0057] FIG6 is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application. Figure 6 As shown, the electronic device may include: a processor 610, a communication interface 620, a memory 630 and a communication bus 640, wherein the processor 610, the communication interface 620 and the memory 630 communicate with each other through the communication bus 640. The processor 610 may call a computer program in the memory 630 to execute the steps of the low-altitude target recognition method, for example including: Classify and store basic information of multiple preset targets in the target low-altitude area into a database; the basic information includes the identification, location and real-time status of the preset targets; Acquiring image data and sound data of the target to be identified collected by the aircraft when flying in the target low-altitude area; extracting target image features from the image data, and extracting target sound features from the sound data; Inputting the target image feature and the target sound feature into a target detection model to obtain the probability that the target to be identified outputted by the target detection model exists in the database; When the probability is greater than the probability threshold, identifying the classification type and basic information corresponding to the target to be identified from the database; The target detection model is obtained by training the historical target image features, the historical target sound data and the labels of the plurality of preset targets on the basis of the convolution function.

[0058] In addition, the logic instructions in the above-mentioned memory 630 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on this understanding, the technical solution of the present application can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0059] On the other hand, an embodiment of the present application further provides a computer program product, the computer program product comprising a computer program, the computer program may be stored on a non-transitory computer-readable storage medium, and when the computer program is executed by a processor, the computer can perform the steps of the low-altitude target recognition method provided in the above embodiments, for example, including: Classify and store basic information of multiple preset targets in the target low-altitude area into a database; the basic information includes the identification, location and real-time status of the preset targets; Acquiring image data and sound data of the target to be identified collected by the aircraft when flying in the target low-altitude area; extracting target image features from the image data, and extracting target sound features from the sound data; Inputting the target image feature and the target sound feature into a target detection model to obtain the probability that the target to be identified outputted by the target detection model exists in the database; When the probability is greater than the probability threshold, identifying the classification type and basic information corresponding to the target to be identified from the database; The target detection model is obtained by training the historical target image features, the historical target sound data and the labels of the plurality of preset targets on the basis of the convolution function.

[0060] On the other hand, an embodiment of the present application further provides a non-transitory computer-readable storage medium on which a computer program is stored, wherein the computer program is used to enable a processor to execute the steps of the low-altitude target recognition method provided in the above embodiments, for example, including: Classify and store basic information of multiple preset targets in the target low-altitude area into a database; the basic information includes the identification, location and real-time status of the preset targets; Acquiring image data and sound data of the target to be identified collected by the aircraft when flying in the target low-altitude area; extracting target image features from the image data, and extracting target sound features from the sound data; Inputting the target image feature and the target sound feature into a target detection model to obtain the probability that the target to be identified outputted by the target detection model exists in the database; When the probability is greater than the probability threshold, identifying the classification type and basic information corresponding to the target to be identified from the database; The target detection model is obtained by training the historical target image features, the historical target sound data and the labels of the plurality of preset targets on the basis of the convolution function.

[0061] The non-transitory computer-readable storage medium can be any available medium or data storage device that can be accessed by the processor, including but not limited to magnetic storage (such as floppy disks, hard disks, magnetic tapes, magneto-optical disks (MO), etc.), optical storage (such as CD, DVD, BD, HVD, etc.), and semiconductor storage (such as ROM, EPROM, EEPROM, non-volatile memory (NAND FLASH), solid-state drive (SSD)), etc.

[0062] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0063] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0064] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit it. Although the present application 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 make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A low-altitude target recognition method, characterized in that: include: Classify and store basic information of multiple preset targets in the target low-altitude area into a database; The basic information includes the identification, location and real-time status of the preset target; Acquiring image data and sound data of the target to be identified collected by the aircraft when flying in the target low-altitude area; extracting target image features from the image data, and extracting target sound features from the sound data; Inputting the target image feature and the target sound feature into a target detection model to obtain the probability that the target to be identified outputted by the target detection model exists in the database; When the probability is greater than the probability threshold, identifying the classification type and basic information corresponding to the target to be identified from the database; The target detection model is obtained by training the historical target image features, the historical target sound data and the labels of the plurality of preset targets on the basis of the convolution function.

2. The low-altitude target recognition method according to claim 1, characterized in that: The step of acquiring the image data and sound data of the target to be identified collected by the aircraft when flying in the target low-altitude area includes: Acquiring light signals, sound signals and environmental pressure signals of the target to be identified collected by the aircraft when flying in the target low-altitude area; fusing the optical signal and the ambient pressure signal into a first electrical signal; fusing the sound signal and the environmental pressure signal into a second electrical signal; generating image data of the target to be identified based on the first electrical signal; The sound data of the target to be identified is generated based on the second electrical signal.

3. The low-altitude target recognition method according to claim 1, characterized in that: The step of extracting target image features from the image data comprises: Standardizing and normalizing the image data to obtain preprocessed image data; extracting a plurality of image features from the preprocessed image data; A target image feature representative of the image data is screened out from the multiple image features.

4. The low-altitude target recognition method according to claim 3, characterized in that: The step of selecting a target image feature representative of the image data from the plurality of image features includes: Randomly splicing the multiple image features to obtain multiple spliced ​​image features; Obtaining a distribution probability of the data of each of the stitched image features in the preprocessed image data to obtain a plurality of first probabilities; Obtaining the distribution probability of the data of each image feature in the data of the spliced ​​image feature to which it belongs, to obtain a plurality of second probabilities; Multiplying the corresponding first and second probabilities to obtain a plurality of third probabilities; The third probabilities are sorted from large to small, and the image features corresponding to the third probabilities with the highest sorting are determined as the target image features.

5. The low-altitude target recognition method according to claim 1, characterized in that: The step of extracting target sound features from the sound data comprises: Improving the signal-to-noise ratio of the sound data in the high frequency part to obtain preprocessed sound data; Dividing the pre-processed sound data into a plurality of sound frames, and performing a windowing operation on each of the sound frames to obtain a plurality of pre-processed sound frames; Performing a fast Fourier transform on each of the pre-processed sound frames to obtain a sound spectrum graph; Converting the sound spectrogram into a Mel-spectrogram; Taking the logarithm of the Mel-spectrogram to obtain a logarithmic spectrum graph; Performing discrete cosine transform on the logarithmic spectrum to obtain Mel-frequency cepstrum coefficients; Low-order coefficients in the Mel-frequency cepstral coefficients are obtained, and sound data features corresponding to the low-order coefficients are determined as the target sound features.

6. The low-altitude target recognition method according to claim 1, characterized in that: The target image features include color features, shape features and texture features; The target sound features include pitch features, loudness features and timbre features.

7. A low-altitude target recognition device, characterized in that: include: An information classification module is used to classify and store basic information of multiple preset targets in the target low-altitude area into a database; The basic information includes the identification, location and real-time status of the preset target; The data acquisition module is used to: acquire image data and sound data of the target to be identified collected by the aircraft when flying in the target low-altitude area; A target feature extraction module, used to extract target image features from the image data and target sound features from the sound data; A first target recognition module is used to: input the target image feature and the target sound feature into a target detection model to obtain the probability that the target to be recognized output by the target detection model exists in the database; A second target recognition module is used to: when the probability is greater than the probability threshold, identify the classification type and basic information corresponding to the target to be recognized from the database; The target detection model is obtained by training the historical target image features, the historical target sound data and the labels of the plurality of preset targets on the basis of the convolution function.

8. An electronic device comprising a processor and a memory storing a computer program, characterized in that: When the processor executes the computer program, the steps of the low-altitude target recognition method described in any one of claims 1 to 6 are implemented.

9. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the low-altitude target recognition method according to any one of claims 1 to 6 are implemented.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the low-altitude target recognition method according to any one of claims 1 to 6 are implemented.

Images