Convolution-long short-term memory neural network article detection and identification method

Abstract

The invention relates to the technical field of security check, in particular to a convolution-long short-term memory neural network article detection and identification method, which comprises the following steps: converting a current millimeter wave image data frame into M parallel memory type millimeter wave image data frames, M being greater than or equal to 2; the M paths of parallel memory type millimeter wave image data frames are output to M space attention convolution modules in a one-to-one correspondence mode to be processed, and M first weighted feature maps are obtained; the M first weighted feature maps are output to M channel attention LSTM modules in a one-to-one correspondence mode to be processed, and M second weighted feature maps are obtained and used for follow-up forbidden article detection and recognition. According to the method, the attention mechanism, the CNN and the LSTM are fused, so that the prohibited articles in the millimeter wave image are automatically recognized, the recognition accuracy of the prohibited articles is improved, and the urgent demand of entry and exit site security check on a security check technology with high efficiency, safety and privacy protection is met.

Description

Technical Field

[0001] The present invention relates to the technical field of security inspection, and in particular to a convolution-long short-term memory neural network object detection and recognition method. Background Art

[0002] As an advanced security inspection method, millimeter wave human body imaging technology has a wide detection range. It can not only detect various metal objects, but also identify prohibited items such as weapons, knives, liquids, powders and fire sources, greatly enhancing the comprehensiveness and accuracy of security inspection. The high-resolution function of millimeter wave imaging can clearly show the location and shape of prohibited items, which helps security inspectors make quick judgments. At the same time, millimeter waves are safe electromagnetic waves that do not penetrate the human body and have no ionizing radiation. They are very safe for special groups such as children, the elderly, and pregnant women. Therefore, they have broad application prospects in the field of security inspection.

[0003] Traditional security inspection methods often rely on manual observation and judgment, which is not only inefficient, but also easily affected by human factors, resulting in missed detection or false alarms. Therefore, it is particularly important to develop an intelligent detection and identification method that can automatically, efficiently and accurately identify prohibited items in millimeter wave images. However, efficient and high-precision identification of prohibited items in millimeter wave images is still a difficult problem to be solved. In order to solve the problem of efficient identification of prohibited items in millimeter wave images, deep learning technology has been introduced into the field of security inspection. Deep learning is a computational model that simulates the neural network of the human brain. Through learning and training, it can realize automatic identification and classification of objects. In object recognition, convolutional neural network (CNN) is one of the most commonly used structures. CNN extracts features from images through multiple convolutional layers and pooling layers, and classifies them through fully connected layers. This structure enables neural networks to effectively process image data and have high recognition accuracy.

[0004] However, when processing complex and changeable millimeter-wave images, a single CNN model may not be able to fully capture the key information in the image, resulting in poor recognition results, which needs to be further improved. Summary of the invention

[0005] The purpose of the present invention is to propose a convolutional-long short-term memory neural network object detection and recognition method, which is used to improve the recognition performance by introducing a long short-term memory network (LSTM) and an attention mechanism. LSTM is a special recurrent neural network (RNN) that can process long sequence data and capture the temporal dependencies in the data. The attention mechanism enables the model to perform a dot product operation based on the current hidden state and the hidden state output by the encoder at each time step of the decoder to generate an attention score, thereby dynamically focusing on the importance of different parts of the input sequence.

[0006] The present invention proposes a convolution-long short-term memory neural network object detection and recognition method by integrating the attention mechanism with CNN and LSTM, realizing the automatic recognition of prohibited items in millimeter wave images, so as to improve the recognition accuracy of prohibited items, reduce the false alarm rate and missed detection rate, ensure the safety of security inspection, and avoid the exposure of passengers' millimeter wave scanned images to others, thus protecting passengers' personal privacy. The present invention mainly stems from the deep integration of millimeter wave imaging technology and deep learning technology, so as to meet the urgent demand for efficient, safe and privacy-protective security inspection technology at the entry and exit site.

[0007] To achieve the purpose of the present invention, the present invention adopts the following technical solutions:

[0008] A convolutional-long short-term memory neural network object detection and recognition method comprises the following steps:

[0009] Convert the current millimeter wave image data frame into M parallel memory-type millimeter wave image data frames, where M≥2;

[0010] Outputting M parallel memory-type millimeter wave image data frames to M spatial attention convolution modules one by one for processing to obtain M first weighted feature maps;

[0011] The M first weighted features Figure 1 One pair is output to M channel attention LSTM modules for processing to obtain M second weighted feature maps for subsequent detection and identification of prohibited items.

[0012] A further improvement is that each of the spatial attention convolution modules includes a first convolution layer, a first maximum pooling layer, a spatial attention mechanism module and a first weighted fusion module.

[0013] A further improvement is that the specific method of outputting M parallel memory-type millimeter wave image data frames one by one to M spatial attention convolution modules for processing to obtain M first weighted feature maps includes:

[0014] Performing preliminary feature extraction on the input memory-type millimeter wave image data frame through the first convolution layer to generate a first original feature map;

[0015] The first original feature map is processed by a first maximum pooling layer to reduce the amount of calculation while retaining important features;

[0016] The first original feature map processed by the first maximum pooling layer is processed by the spatial attention mechanism module to generate a spatial attention map;

[0017] The spatial attention map is multiplied element by element with the first original feature map through the first weighted fusion module to achieve effective feature fusion and generate the first weighted feature map.

[0018] A further improvement is that each of the channel attention LSTM modules includes a first LSTM unit, a second maximum pooling layer, a channel attention mechanism module and a second weighted fusion module.

[0019] A further improvement is to use the M first weighted features Figure 1 The specific method of outputting one pair of pairs to M channel attention LSTM modules for processing to obtain M second weighted feature maps includes:

[0020] The input first weighted feature map is processed by the first LSTM unit and the second maximum pooling layer to further extract deep feature information to obtain a second original feature map;

[0021] Processing the second original feature map through the channel attention mechanism module to generate a channel attention map;

[0022] The channel attention map is multiplied element by element with the second original feature map output after being processed by the second maximum pooling layer through the second weighted fusion module to achieve effective feature fusion and generate a second weighted feature map.

[0023] A further improvement is that the object detection and identification method further comprises the following steps:

[0024] The M second weighted feature maps obtained by processing the M channel attention LSTM modules are cascaded through a cascade layer to form a weighted feature panorama covering M parallel memory-type millimeter wave image data frames;

[0025] The weighted feature panorama is input into the second LSTM unit and the third maximum pooling layer for deep-level prohibited items feature information extraction, where the second LSTM unit is used to capture the temporal dependency in the feature sequence, and the third maximum pooling layer is used to reduce the dimension of the data and retain the most important feature information.

[0026] A further improvement is that the object detection and identification method further comprises the following steps:

[0027] The extracted feature information of prohibited items is input into the fully connected layer for feature fusion and classification decision.

[0028] A further improvement is that the fully connected layer includes at least two layers of fully connected networks. The fully connected layer maps the feature information to the classification space through linear transformation and activation function, thereby realizing the detection and identification of prohibited items, and finally outputs the feature classification results to determine whether prohibited items are contained and the type of prohibited items.

[0029] A further improvement is that the object detection and identification method further includes the following steps: marking, locating and visually displaying classification information of detected prohibited objects.

[0030] A further improvement is that the conversion of the current millimeter wave image data frame into M-way parallel memory-type millimeter wave image data frame is achieved by using a time-delay neural network.

[0031] The beneficial effects of the present invention are:

[0032] The convolution-long short-term memory (CNN-LSTM) neural network object detection and recognition method proposed in the present invention shows significant technical advantages compared with the existing entry and exit security inspection millimeter wave imaging processing technology.

[0033] First, in the data preprocessing stage, the present invention converts the current millimeter-wave image data frame into M-way parallel memory-type millimeter-wave image data frame through a time-delay neural network. This innovative conversion method effectively improves the data processing speed and parallel processing capability, laying a solid foundation for subsequent efficient feature extraction.

[0034] Secondly, in terms of feature extraction, the present invention combines the spatial attention mechanism and the channel attention mechanism. Through M spatial attention convolution modules and M channel attention LSTM modules, M consecutive frames of millimeter wave image data frames are processed in parallel. The use of this dual attention mechanism enables the system to focus on the key information in the image more accurately, effectively suppressing the interference of irrelevant information, thereby improving the accuracy and robustness of prohibited item identification.

[0035] In addition, the present invention further mines the deep feature information in the millimeter wave image through deep feature extraction of LSTM layer and maximum pooling layer. This deep feature extraction method not only enhances the recognition ability of the system, but also improves the recognition accuracy of prohibited items in complex scenes.

[0036] Finally, the present invention realizes intelligent detection and discrimination of prohibited items by fusing and classifying the features of prohibited items through a fully connected layer. This intelligent detection method not only improves the security inspection efficiency, but also greatly reduces the workload of security inspectors, providing strong support for improving the security inspection efficiency and accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 This is a flow chart of a convolution-long short-term memory neural network object detection and recognition method of the present invention;

[0038] Figure 2 This is a structural functional block diagram of a convolutional-long short-term memory neural network object detection and recognition method of the present invention;

[0039] Figure 3 Schematic diagram of the conversion of memory-type millimeter-wave image data frames in the input preprocessing layer;

[0040] Figure 4 This is the structural functional block diagram of the spatial attention convolution module;

[0041] Figure 5 This is the structural functional block diagram of the channel attention LSTM module. DETAILED DESCRIPTION

[0042] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0043] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily 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 invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0044] The following is an explanation of the technical terms involved in the present invention:

[0045] Millimeter wave security imaging: This is a technology that uses millimeter waves (a type of electromagnetic wave) to scan the human body or objects, which can generate X-ray-like photos to check whether prohibited items are carried.

[0046] Convolutional Neural Network (CNN): A special type of deep learning model that is particularly good at processing image data. It automatically learns features from images to perform classification or recognition tasks by simulating the working method of the human brain.

[0047] Long Short-Term Memory (LSTM) is a special type of recurrent neural network (RNN) that can process long sequence data and capture temporal dependencies in the data.

[0048] Attention mechanism: This is a technique that mimics human vision, allowing the model to "focus" on certain specific areas when processing an image, while ignoring other unimportant parts.

[0049] At the entry and exit site, the millimeter wave security inspection imaging system plays a vital role, but traditional methods often face challenges in efficiency and accuracy when dealing with complex scenes and identifying a variety of prohibited items. To solve this problem, an embodiment of the present invention proposes a convolution-long short-term memory (CNN-LSTM) neural network object detection and recognition method.

[0050] Please refer to the attached Figure 1 -Attached Figure 5 The embodiment of the present invention proposes a convolution-long short-term memory (CNN-LSTM) neural network object detection and identification method, which can be applied to the millimeter wave security imaging system in public places such as entry and exit, airports, and stations to achieve efficient and accurate detection and identification of prohibited items. Figure 1 As shown, the object detection and identification method includes the following steps:

[0051] Step S1: Convert the current millimeter wave image data frame x(n) into M parallel memory-type millimeter wave image data frames through the input preprocessing layer, where M≥2.

[0052] Specifically, in step S1, the current millimeter wave image data frame is converted into an M-way parallel memory-type millimeter wave image data frame by using a time-delay neural network (TDNN). The TDNN is used to convert a single data frame into a parallel data frame with M-way outputs. Each output data frame is a different representation of the original data, which enhances the parallelism and memory of the data and provides more perspectives for subsequent feature extraction. In simple terms, an original millimeter wave image is converted into M different image versions, each of which can provide information about different perspectives or angles of the original image, and can provide more diverse input data for subsequent analysis steps.

[0053] It is understandable that since millimeter wave images may be fuzzy or complex, especially when multiple objects appear in the image at the same time, it is difficult to directly identify specific prohibited items from the image. Therefore, the purpose of step S1 of the present invention is to increase the parallelism and memory of data, and provide a richer information basis for subsequent feature extraction and recognition, so as to solve the above problems.

[0054] Step S2: output the M parallel memory-type millimeter wave image data frames one by one to M spatial attention convolution modules for processing to obtain M first weighted feature maps.

[0055] It can be understood that by assigning M parallel memory-type millimeter-wave image data frames to M spatial attention convolution modules respectively, each spatial attention convolution module is responsible for processing one frame of millimeter-wave image data frame, that is, spatial attention convolution module 1, spatial attention convolution module 2, ... spatial attention convolution module M are respectively used to perform convolution processing on M parallel memory-type millimeter-wave image data frames to extract the spatial feature information of the millimeter-wave images in parallel, so as to process these image data more efficiently and accurately.

[0056] Step S3: M first weighted features Figure 1 One pair is output to M channel attention LSTM modules for processing to obtain M second weighted feature maps for subsequent detection and identification of prohibited items.

[0057] It can be understood that by assigning M first weighted feature maps to M channel attention LSTM modules respectively, each channel attention LSTM module is responsible for processing one first weighted feature map respectively, that is, channel attention LSTM module 1, channel attention LSTM module 2,...channel attention LSTM module M are used to process the M first weighted feature maps respectively, so as to further deeply extract the spatial feature information of the millimeter wave image.

[0058] In a preferred solution of an embodiment of the present invention, each of the spatial attention convolution modules includes a first convolution layer, a first maximum pooling layer, a spatial attention mechanism module and a first weighted fusion module.

[0059] In step S2, the specific method of outputting M parallel memory-type millimeter wave image data frames to M spatial attention convolution modules for processing to obtain M first weighted feature maps includes:

[0060] Step S21: Perform preliminary feature extraction on the input memory-type millimeter wave image data frame through the first convolution layer to generate a first original feature map.

[0061] It can be understood that each spatial attention convolution module first performs preliminary feature extraction on the input millimeter wave image data frame through the first convolution layer. This process generates a set of original feature maps, which contain local patterns and structural information in the input image, such as edges, textures, etc.

[0062] Step S22: Processing the first original feature map through the first maximum pooling layer to reduce the amount of calculation while retaining important features.

[0063] It can be understood that the image size is reduced by the first maximum pooling layer, which reduces the amount of calculation while retaining important features. This step does not change the content of the original feature map, but only reduces its resolution. The addition of the first maximum pooling layer is used to avoid overfitting and increase the generalization ability of the model.

[0064] Step S23: The first original feature map processed by the first maximum pooling layer is processed by the spatial attention mechanism module to generate a spatial attention map.

[0065] It can be understood that the spatial attention map reflects the importance of different regions in the image, enabling the model to automatically learn and focus on more important regions in the image, thereby improving the accuracy of feature extraction.

[0066] Step S24: The spatial attention map is multiplied element by element with the first original feature map through the first weighted fusion module to achieve effective feature fusion and generate a first weighted feature map.

[0067] It can be understood that by multiplying the spatial attention map element-by-element with the original feature map to generate the first weighted feature map, this process enhances the feature representation of the key area, allowing the model to focus more on the parts that may contain prohibited items.

[0068] In general, the purpose of steps S21 to S24 in the present invention is to extract spatial features in the image, and enhance important areas and weaken irrelevant parts through the spatial attention mechanism. Improve the quality of feature representation so that subsequent processing can focus more on key information. Specifically, through the spatial attention mechanism, the model can automatically learn and pay attention to more important areas in the image, thereby improving the accuracy of feature extraction. Suppress irrelevant information and weaken irrelevant or interfering background information so that the model can focus more on the target object. Improve the quality of feature representation, extract more representative spatial features through convolution and pooling operations, and provide high-quality input for subsequent processing.

[0069] In a preferred scheme of an embodiment of the present invention, each of the channel attention LSTM modules includes a first LSTM unit, a second maximum pooling layer, a channel attention mechanism module and a second weighted fusion module.

[0070] Specifically, in step S3, the M first weighted features Figure 1 The specific method of outputting one pair of pairs to M channel attention LSTM modules for processing to obtain M second weighted feature maps includes:

[0071] Step S31: Process the input first weighted feature map through the first LSTM unit and the second maximum pooling layer to further extract deep feature information to obtain a second original feature map.

[0072] It can be understood that the first LSTM unit can effectively handle the long-term dependency problem in the sequence data. LSTM is used to capture the feature changes between different time points, especially when the object moves or deforms in the millimeter wave image, which is very important for object detection in dynamic scenes. The second maximum pooling layer reduces the size of the feature map, reducing the computational burden of subsequent processing while maintaining key features.

[0073] Step S32: Process the second original feature map through the channel attention mechanism module to generate a channel attention map.

[0074] It can be understood that the channel attention mechanism module calculates the average and maximum value of each channel on the entire feature map, and predicts the weight of each channel through a multi-layer perceptron (MLP). These weights are then applied to the second original feature map to generate a channel attention map, which reflects the importance of different channels for the current task.

[0075] The introduction of the channel attention mechanism has the following benefits: It enhances the model's ability to understand the relationship between feature channels, thereby improving the accuracy of feature extraction. Specifically, it enhances important channels: the channel attention map can highlight those channels that are more important to the current task, thereby improving the relevance and effectiveness of feature representation. It suppresses irrelevant channels: it weakens those channels that are less important to the current task, reducing the impact of noise.

[0076] Step S33: The channel attention map is multiplied element by element with the second original feature map output after being processed by the second maximum pooling layer through the second weighted fusion module to achieve effective feature fusion and generate a second weighted feature map.

[0077] It is understandable that the purpose of step S33 is to adjust the strength of each feature map according to the importance of the channel, so as to achieve effective feature fusion. The second weighted feature map finally generated contains both the spatial information of the original feature map and the important channel information emphasized by the channel attention map. By combining the attention mechanism of the two dimensions of space and channel, the generated second weighted feature map focuses more on the key features of the target object, improving the accuracy of detection and recognition. Since the attention mechanism can automatically adapt to different inputs, the model can still maintain high performance even in the presence of interference or noise.

[0078] In general, after the processing of step S31 to step S33, not only the deep feature information is further extracted, but also the quality of the feature map is optimized, so that the finally generated second weighted feature map is more suitable for the detection and identification of prohibited items, which effectively improves the accuracy and robustness of the system and provides strong support for practical applications.

[0079] In a preferred solution of this embodiment, the object detection and identification method further includes the following steps:

[0080] Step S4: The M second weighted feature maps obtained by processing the M channel attention LSTM modules are cascaded through a cascade layer to form a weighted feature panorama covering M parallel memory-type millimeter wave image data frames.

[0081] It is understandable that the cascade layer cascades the M second weighted feature maps in a specific order (for example, according to the time order of their corresponding millimeter wave image data frames). The key to this step is to integrate the information of multiple feature maps into a larger feature map to form a weighted feature panorama covering M parallel memory-type millimeter wave image data frames. Cascading can be achieved by simple splicing or by more complex fusion strategies, such as learning a weight matrix to dynamically adjust the contribution of each feature map.

[0082] Step S5: Input the weighted feature panorama into the second LSTM unit and the third maximum pooling layer for deep-level prohibited items feature information extraction, where the second LSTM unit is used to capture the temporal dependency in the feature sequence, and the third maximum pooling layer is used to reduce the dimension of the data and retain the most important feature information.

[0083] It can be understood that deep-level feature information of prohibited items is extracted from the weighted feature panorama, providing strong support for the final detection and identification. After the information of the weighted feature panorama is integrated, the long-term dependencies between these information can be better understood through the second LSTM unit. The weighted feature panorama processed by the second LSTM unit is downsampled again through the third maximum pooling layer. This process helps to reduce the amount of data while retaining the most significant features, allowing the model to focus on the most important information, thereby improving computational efficiency and model generalization ability.

[0084] In a preferred solution of this embodiment, the object detection and identification method further includes the following steps:

[0085] Step S6: Input the extracted feature information of prohibited items into the fully connected layer for feature fusion and classification decision.

[0086] Specifically, the fully connected layer includes at least two layers of fully connected networks (a first fully connected layer and a second fully connected layer). The fully connected layer maps the feature information to the classification space through linear transformation and activation function, thereby realizing the detection and identification of prohibited items, and finally outputs the feature classification results to determine whether prohibited items are contained and the type of prohibited items.

[0087] It can be understood that the fully connected layer is able to capture the complex relationship between input features and effectively fuse features from different sources or levels through multiple layers of linear transformation and nonlinear activation.

[0088] In a preferred solution of this embodiment, the object detection and identification method further includes the following steps: marking, locating and visually displaying classification information of detected prohibited items to facilitate security personnel to make quick judgments and responses.

[0089] Specifically, it can be intuitively presented to security personnel through a user interface, assisting them in making quick and accurate judgments.

[0090] In addition, the network structure parameters such as the convolutional layer, the maximum pooling layer, and the fully connected layer in the present invention can be optimized and adjusted according to the specific application scenario to improve the detection accuracy and recognition efficiency.

[0091] The convolution-long short-term memory (CNN-LSTM) neural network object detection and recognition method proposed in the present invention shows significant technical advantages compared with the existing entry and exit security inspection millimeter wave imaging processing technology.

[0092] First, in the data preprocessing stage, the present invention converts the current millimeter-wave image data frame into M-way parallel memory-type millimeter-wave image data frame through a time-delay neural network. This innovative conversion method effectively improves the data processing speed and parallel processing capability, laying a solid foundation for subsequent efficient feature extraction.

[0093] Secondly, in terms of feature extraction, the present invention combines the spatial attention mechanism and the channel attention mechanism. Through M spatial attention convolution modules and M channel attention LSTM modules, M consecutive frames of millimeter wave image data frames are processed in parallel. The use of this dual attention mechanism enables the system to focus on the key information in the image more accurately, effectively suppressing the interference of irrelevant information, thereby improving the accuracy and robustness of prohibited item identification.

[0094] In addition, the present invention further mines the deep feature information in the millimeter wave image through deep feature extraction of LSTM layer and maximum pooling layer. This deep feature extraction method not only enhances the recognition ability of the system, but also improves the recognition accuracy of prohibited items in complex scenes.

[0095] Finally, the present invention realizes intelligent detection and discrimination of prohibited items by fusing and classifying the features of prohibited items through a fully connected layer. This intelligent detection method not only improves the security inspection efficiency, but also greatly reduces the workload of security inspectors, providing strong support for improving the security inspection efficiency and accuracy.

[0096] The embodiments described above are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, a person skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such 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, and should all be included in the protection scope of the present application.

Claims

1. A convolutional-long short-term memory neural network object detection and recognition method, characterized in that: The following steps are involved: Convert the current millimeter wave image data frame into M parallel memory-type millimeter wave image data frames, where M≥2; Outputting M parallel memory-type millimeter wave image data frames to M spatial attention convolution modules one by one for processing to obtain M first weighted feature maps; The M first weighted feature maps are output one by one to M channel attention LSTM modules for processing to obtain M second weighted feature maps for subsequent detection and identification of prohibited items.

2. The method for object detection and recognition using a convolutional-long short-term memory neural network according to claim 1, characterized in that: Each of the spatial attention convolution modules includes a first convolution layer, a first maximum pooling layer, a spatial attention mechanism module and a first weighted fusion module.

3. The method for object detection and recognition using a convolutional-long short-term memory neural network according to claim 2, characterized in that: The specific method of outputting M parallel memory-type millimeter wave image data frames to M spatial attention convolution modules one by one for processing to obtain M first weighted feature maps includes: Performing preliminary feature extraction on the input memory-type millimeter wave image data frame through the first convolution layer to generate a first original feature map; The first original feature map is processed by a first maximum pooling layer to reduce the amount of calculation while retaining important features; The first original feature map processed by the first maximum pooling layer is processed by the spatial attention mechanism module to generate a spatial attention map; The spatial attention map is multiplied element by element with the first original feature map through the first weighted fusion module to achieve effective feature fusion and generate the first weighted feature map.

4. The method for object detection and recognition using a convolutional-long short-term memory neural network according to claim 1, characterized in that: Each of the channel attention LSTM modules includes a first LSTM unit, a second maximum pooling layer, a channel attention mechanism module and a second weighted fusion module.

5. The method for object detection and recognition using a convolutional-long short-term memory neural network according to claim 4, characterized in that: The specific method of outputting the M first weighted feature maps one by one to the M channel attention LSTM modules for processing to obtain the M second weighted feature maps includes: The input first weighted feature map is processed by the first LSTM unit and the second maximum pooling layer to further extract deep feature information to obtain a second original feature map; Processing the second original feature map through the channel attention mechanism module to generate a channel attention map; The channel attention map is multiplied element by element with the second original feature map output after being processed by the second maximum pooling layer through the second weighted fusion module to achieve effective feature fusion and generate a second weighted feature map.

6. A convolutional-long short-term memory neural network object detection and recognition method according to any one of claims 1 to 5, characterized in that: The object detection and identification method further comprises the following steps: The M second weighted feature maps obtained by processing the M channel attention LSTM modules are cascaded through a cascade layer to form a weighted feature panorama covering M parallel memory-type millimeter wave image data frames; The weighted feature panorama is input into the second LSTM unit and the third maximum pooling layer for deep-level prohibited items feature information extraction, where the second LSTM unit is used to capture the temporal dependency in the feature sequence, and the third maximum pooling layer is used to reduce the dimension of the data and retain the most important feature information.

7. The method for object detection and recognition using a convolutional-long short-term memory neural network according to claim 6, characterized in that: The object detection and identification method further comprises the following steps: The extracted feature information of prohibited items is input into the fully connected layer for feature fusion and classification decision.

8. The method for object detection and recognition using a convolutional-long short-term memory neural network according to claim 7, characterized in that: The fully connected layer includes at least two layers of fully connected networks. The fully connected layer maps the feature information to the classification space through linear transformation and activation function, thereby realizing the detection and identification of prohibited items, and finally outputs the feature classification result to determine whether it contains prohibited items and the type of prohibited items.

9. The method for object detection and recognition using a convolutional-long short-term memory neural network according to claim 8, characterized in that: The object detection and identification method also includes the following steps: marking, locating and visually displaying the classification information of the detected prohibited objects.

10. The method for object detection and recognition using a convolutional-long short-term memory neural network according to claim 1, characterized in that: The conversion of the current millimeter wave image data frame into M-way parallel memory-type millimeter wave image data frames is achieved by using a time-delay neural network.

Images