39results about How to "Good retrieval effect" patented technology

The invention discloses a self-adaptive personalized information retrieval system and method. For timely catching irregularly distributed dynamic retrieval requirements of a user, a retrieval module is timely updated through interaction of the user and a search engine. The system comprises a data input sub system, a parameter training and predicating sub system, a retrieval performing sub system and a data output sub system, wherein the data input sub system is used for combining historical inquiry information and historical click information to form a characteristic matrix according to the current inquiry information, and acquiring a training parameter predicating module according to the characteristic matrix; the parameter training and predicating sub system is used for training and applying the parameter predicating module to acquire the predicated parameters according to the characteristic matrix; the retrieval performing sub system is used for predicating the parameters to organize the current inquiry and the historical inquiry, and combining the user module and the inquiry module to form a personalized inquiry module; and the data output sub system is used for searching a document matched with the personalized inquiry from the document to be retrieved as a primary retrieved result, and sequencing the primary retrieved result according to the correlation to obtain the final retrieved result for outputting.

The invention relates to the field of data retrieval, and discloses a preference-based intelligent retrieval method and system. The method includes the steps that a subject preference model of a user is established on the basis of subject classification of data, user characteristics and operation logs; query expansion is performed through the subject preference model of the user and retrieval input of the user to obtain a primary retrieval result; subject preference scoring is performed on the data through the subject preference model of the user and distribution conditions of the data on all subjects, and personalized retrieval ranking is preformed on the primary retrieval result on the basis of subject preference; secondary feedback retrieval is performed on the ranked primary retrieval result through a comprehensive model of relevance feedback and pseudo relevance feedback to obtain a final retrieval result. According to the method and system, subject distribution of data resources is determined through the subject indexing technology, retrieval vectors better representing user requirements are established through the query expansion based on subjects, relevance feedback and other technologies, and the retrieval result meeting potential requirements of the user better is provided for the user.

The invention discloses an MPEG-7-based image retrieving system, which mainly comprises three key techniques, namely, characteristic extraction, content description and searching engine. The MPEG-7-based image retrieving system comprises (1) content-based image retrieval CBIR, (2) a multimedia content description interface MPEG-7 and (3) a database model and management. Specifically, the MPEG-7-based image retrieving system comprises a characteristic extraction module, a retrieval module, an output module and a query module, and realizes network function-oriented image content-based query, so that a user can quickly and effectively retrieve image resources. A descriptor specified by an MPEG-7 standard is improved so as to obtain remarkable retrieval effect and perform combined query on different visual characteristics, such as colors, textures and shapes. The image retrieving system is developed for digital libraries, medical application and trademark retrieval application, and an algorithm is optimized according to the MPEG-7 standard, so that the MPEG-7-based image retrieving system reaches a leading domestic level in terms of algorithm study and application. The MPEG-7-based image retrieving system has the characteristics of clear technical direction, smooth development and relatively stronger maturity.

The present invention discloses a method for retrieving atmospheric aerosol based on statistical segmentation. Firstly a multi-band remote sensing image including an apparent reflectance and an aerosol optical thickness look-up table corresponding to a retrieval band is obtained, then pixels are partitioned and screened according to apparent reflectance segments of a mid-infrared 2.1 micrometer band. After that the retained pixel sets are further partitioned and screened according to the apparent reflectance segments of the mid-infrared 1.6 micrometer band. Finally the obtained pixel sets are partitioned into two categories according to the pixel number, one category including pixels having more pixels, the other including those with less pixels.The category with more pixels is taken as the reference part for retrieval. Specifically, the pixel sets are first searched for the clean segment, then the ground surface reflectivity of the clean segment is taken as the ground surface reflectivity of the whole pixel set, thereby obtaining the aerosol thickness value through retrieval. After that these pixels are taken as references to perform retrieval on the other category.The present invention can improve accuracy and resolution of the retrieval result of the bright ground surface area, and is applied to a wider range.

The invention discloses a multi-feature fusion image retrieval method. The method comprises the steps of firstly, inputting an image I to be retrieved; secondly, constructing a color feature vector and a SIFT feature vector of the image I; thirdly, training the image in a query image library to obtain a color feature dictionary and a SIFI feature dictionary, and using a visual word for representing the image in the image library; fourthly, using the visual word for representing the image I, calling a candidate image set Q from the query image library according to the visual word, and calculating a similarity value score (Q,I); fifthly, selecting a local area Si with the visual saliency in the image I and repeatedly executing the step three and the step four to obtain a candidate image set K, and calculating a similarity value score<sal>(K,I); sixthly, using an overlapped image set of the two candidate fusion sets as D, fusing score<sal>(D,I) and score (D,I), and calculating a final similarity value score<*>(D,I); seventhly, using the image with the highest final similarity value as a retrieval result of the image I to be retrieved. The multi-feature fusion image retrieval method has the advantages of lowering image noise and improving the retrieval accuracy.

The invention provides a method for screening and optimizing an audio-frequency keyword template. The method comprises the steps of step1, extracting features of each audio-frequency keyword template sample, making extracted features through a deep neural network and calculating posterior probabilities of all phonemes on a given phoneme set; step2, calculating posterior probability stability scores, pronunciation reliability scores and neighborhood similarity scores of the template; step3, calculating a weighted average of the above three scores of each audio-frequency keyword template and denoting the weighted average as an average score; step4, sequencing according to the order of average scores from large to small and selecting the first L audio-frequency keyword templates as representative pronunciation templates; step5, processing each representative pronunciation template, adjusting the posterior probability of each pronunciation unit on each frame of the pronunciation sequence and minimizing the neighborhood similarity scores of the template; generating optimized L audio-frequency retrieval word templates.

The invention discloses a three-dimensional model cross-domain retrieval method and system based on a complex background image. The method comprises the steps: building a data set suitable for triplenetwork input, enabling a three-dimensional model to be expressed through multiple views, and unifying the size of image data; designing a cross-domain retrieval triple deep network; adopting an imageaccurate feature extraction network and a three-dimensional model grouping view feature extraction network for completing effective feature extraction of input data, constructing a feature joint embedding space, mapping the features of different domains into the same high-dimensional space, wherein the feature distance of data of the same category is small, and the feature distance of data of different categories is large; finally, measuring the similarity between the image and the three-dimensional model in the feature joint embedding space by adopting an Euclidean distance to finish cross-domain retrieval. According to the method, the corresponding three-dimensional model can be obtained through retrieval according to the input single RGB image with complex background information.

The invention discloses an image retrieval algorithm based on a distribution entropy gain loss function, and the algorithm employs a pre-training network for initialization, trains a network accordingto the demands of an image retrieval task, employs a self-designed distribution entropy gain loss function during the training of the network, and improves the accuracy of image retrieval. A distribution entropy gain loss function is combined with a comparison loss function and a relative entropy, so that the accuracy of image similarity measurement during network training is enhanced; the comparison loss function calculates the similarity between the features through the Euclidean distance, the relative entropy can be used for measuring the distribution difference between the feature vectors, and the relative entropy is supplemented to the comparison loss function to improve the similarity measurement of the feature vectors; a distribution entropy gain loss function is used for traininga network model, the network model more suitable for an image retrieval task is obtained by adjusting network parameters, and the trained network model obtains a better retrieval effect in an image retrieval experiment.

The invention relates to the technical field of testing devices, in particular to an online testing automatic device. The online testing automatic device comprises a machine frame, a working table which is installed on the machine frame, a feed mechanism and a discharge mechanism which are installed on the two sides of the working table respectively, a storage mechanism which is located between the feed mechanism and the discharge mechanism, a material taking mechanism which is installed on the working table, a material taking mechanical hand which is installed on the surface of the working table, a good product material preparing platform and a reject product material storage platform which are installed on the working table, a conveying mechanism which is installed at the position, closeto the good product material preparing platform, of the working table, a testing, supporting and adjusting mechanism which is installed on the working table and connected with the conveying mechanism, and a testing and pressing mechanism which is installed on the working table and located on the upper side of the testing, supporting and adjusting mechanism. Through full-automatic feeding, discharging, conveying and testing, automatic classifying is conducted through the mechanical hand after testing, the problem that an existing semi-automatic device needs to be operated manually is solved, and the work efficiency and automation effect are greatly improved.

The invention discloses an image retrieval method based on a sketch. According to the method, domain migration learning from an image domain to a sketch domain is realized, and a network can output effective features which are adaptive to the sketch domain and have the distinguishing performance. The method comprises the following steps of extracting a Canny edge of an image to serve as network training data; carrying out network pre-training based on millions of images of ImageNet; sending the Canny edge of the image to the network for training, and realizing domain migration learning; sending the sketch and the Canny edge of the image to the trained network, and extracting the features respectively; and computing cosine distances between the extracted features and sorting, and realizingK nearest neighbor retrieval. According to the method, the defects of sketch data of the network in a training process are fully compensated, the effect of sketch training is improved, a cross-domaindifference between the sketch and the image is reduced, and a good retrieval effect is achieved.

The invention discloses an image retrieval method combining visual saliency and phrases. The image retrieval method comprises the following steps that a search image is input and decoded into a YUV color space, and K-means clustering is conducted on pixels of the YUV color space, so that the image is divided into multiple superpixel units; likelihood computation is conducted on each superpixel unit of the divided image, and four obtained likelihoods are fused, and then the saliency map of the superpixel precision is obtained; bilateral Gaussian filtering is conducted, an image saliency map of the pixel precision is obtained; self-adaptive threshold segmentation is conducted on the image saliency map, and a binary image with a prominent target portion is obtained; a dictionary is established, vision words of an image salient region are extracted, and image description is generated; the image similarity between the search image and each image in a image base is calculated. By the adoption of the image retrieval method, the salient region in the image can be more accurately reflected, the visual saliency and the phases are well combined, and the retrieval effect is good.

The invention discloses a material fetching device with material rake structures. The material fetching device comprises a main beam, a sliding device, material rake bodies and drawing structures. The slide device is mounted on the main beam and can slide along the main beam in a left-right reciprocating manner under the effect of a driving device, the material rake bodies are connected onto the slide device via a left hinge structure and a right hinge structure in such a manner that the material rake bodies can rotate around hinged shafts, and the drawing structures are arranged among the material rake bodies and the slide device and can drive the material rake bodies to rotate; two contact spherical surfaces are further fixedly arranged at the hinge structures, arc-shaped surfaces of the contact spherical surfaces face the ends of the hinged shafts and are butted to the ends of the hinged shafts, the contact spherical surface which corresponds to the left hinge structure is arranged at the right end of the left hinge structure, and the contact spherical surface which corresponds to the right hinge structure is arranged at the left end of the right hinge structure. The material rake bodies slightly deform under the effect of rightward resistance when moving leftwards, a small gap is formed between the left hinged shaft and the corresponding contact spherical surface, the right hinged shaft is tightly butted to the corresponding contact spherical surface, accordingly, the resistance can be transmitted to the slide device, and vice versa. The material fetching device has the advantages that stress on the material rake bodies can be improved, accordingly, the fatigue resistance of the material rake bodies can be enhanced, and the service lives of the material rake bodies can be prolonged.

The invention discloses a cross-modal retrieval method based on semantic constraint matrix decomposition hash, which comprises the following steps of: inputting original characteristic matrixes of different modals, decomposing each modal according to an established individual matrix model to obtain a potential semantic matrix, and constraining two modal potential semantic matrixes by utilizing a semantic similarity matrix to obtain a cross-modal retrieval matrix; generating a hash code according to the potential semantic matrix and the orthogonal rotation matrix, and finally obtaining a classlabel from the hash code by using the potential speech matrix. According to the method, the semantic similarity matrix is used for constraining the modal specific representation of each modal, the retrieval effect is superior to that of an existing matrix decomposition method, and a large number of experiments on three data sets show that the method has better retrieval performance.

The invention relates to an ejection type edge coiling bending mold on the side of a floating mandrel and a usage method thereof. The ejection type edge coiling bending mold comprises a puncher pin bearing seat and a lower mold bearing seat, wherein a forming puncher pin is arranged on the lower side face of the puncher pin bearing seat, a press plate is arranged on the rear side of the forming puncher pin, a forming lower die is arranged on the upper side face of the lower die bearing seat and is matched with the forming puncher pin, a discharging frame is arranged beside the forming lower die, a guiding block is arranged on the rear side of the forming lower die, a guiding groove is formed in the guiding block, a floating block is arranged in the guiding groove, a cylinder body is arranged at the upper end of the floating block, the mandrel is horizontally arranged at the front portion of the cylinder body and is located between the forming lower die and the forming puncher pin, a counter bored hole is formed in the rear portion of the cylinder body, a piston is arranged in the counter bored hole, an air cylinder cover is arranged at the rear end of the counter bored hole, horizontal pin holes are evenly distributed in the front end of the cylinder body and are located in the periphery of the mandrel, ejection pins penetrate into the horizontal pin holes, and the rear ends of the ejection pins are connected with the piston. By applying the floating mandrel, the problem that offsets or twisting easily occurs to cause scrap during edge coiling of a workpiece is solved, and a material taking effect is good by arranging lateral ejection structures.

The invention discloses a fine-grained image-text retrieval method based on a Transform model. The fine-grained image-text retrieval method comprises the following steps: acquiring a region vector group of an image target region and a word vector group of a text; each target area of the image is optimized by using a trained self-attention Transform model, so that each target area obtains effective information of other target areas around the target area; determining the meaning information of the current target word according to the semantic information of the text by using a trained self-attention Transform model; performing cross-modal information processing by using a trained mutual attention Transform model, so that the region vector group comprises key information, and the word vector group comprises detail information; and respectively calculating the similarity of each finally obtained region vector and each word vector, obtaining a fine-grained semantic similarity matrix of the input image and the input text, and obtaining a retrieval result. And the retrieval performance and accuracy are improved.

The invention belongs to the technical field of traffic pedestrian image processing, and discloses a pedestrian retrieval method and system based on sketch image semi-coupling measurement identification dictionary learning. The method comprises the following steps of establishing an own heterogeneous pedestrian retrieval database, extracting features from images, and marking different persons withcorresponding colors; processing the extracted sample characteristics, introducing a dictionary learning method, and learning the dictionary pairs of the heterogeneous data; learning a mapping matrixfrom a sketch image set and a normal image set; introducing the identification algorithm learning measurement. The method has the advantages that the problem of lacking heterogeneous pedestrian datasets in the field of heterogeneous pedestrian retrieval is solved, and the semi-coupling metric identification dictionary learning (SMD2L) technology is put forward for the first time. According to the technology, a semi-coupling mapping matrix can be learned from heterogeneous samples and dictionary pairs, so that the difference between the heterogeneous data is reduced to a certain extent, and an ideal retrieval effect is achieved on a new SINPID data set.

The invention relates to the field of fire-fighting equipment, in particular to automatic material pickup equipment for a small-hole stamping die. The equipment comprises a bottom plate; a base is arranged at the upper side of the bottom plate; the base is connected with the bottom plate through a horizontal driving assembly; a fixing base is fixedly arranged at the upper side of the base; a firstsliding block is arranged at one side of the fixing base and connected with the fixing base through a vertical driving assembly; a moving arm is fixedly arranged on the sliding block; a switching rodis rotationally arranged at one end of the moving arm; a first power assembly for driving the switching rod to rotate is arranged in the moving arm; a gripper assembly and a suction cup assembly arearranged at the two ends of the switching rod correspondingly; a blowing mechanism is arranged at the upper side of the bottom plate; an arc-shaped plate matched with the blowing mechanism is arrangedat one side of the switching rod; two bending rods are fixedly arranged on the lower side of the arc-shaped plate; and a collecting box is arranged at the lower end of the arc-shaped plate. The automatic material pickup equipment can be used for automatic material pickup of the small-hole stamping die, materials with different thicknesses can be picked up and placed, the working efficiency is high, meanwhile, a certain sweeping function is achieved on waste generated after stamping, and the waste can be collected.

The invention relates to an automatic electrical testing machine for a port element of a connector, and the automatic electrical testing machine comprises a rack. The rack is provided with a power distribution control box and an operating display panel, which are connected with each other. The right part above the rack is provided with a conveying device, and the left side of the conveying deviceis provided with a turnover device which is in butt joint with the conveying device. The turnover device cooperates with a detection device through a material rotating device. The turnover device comprises a turnover seat disposed on the rack, and the turnover seat is provided with a turnover frame in a hinged manner, wherein the turnover frame is connected with a pneumatic motor disposed on the turnover seat. The turnover frame is provided with an absorption plate cooperating with the conveying device, and the absorption plate is provided with a binding post and a groove cooperating with an interface. According to the invention, the automatic electrical testing machine employs the conveying device for conveying a connector port, turns the interface of the connector port downwards throughthe turnover device, and then enables the connector port to be inserted into an electrical testing device through the material rotating device, thereby achieving the continuous automatic detection, reducing the labor of a worker, and greatly improving the detection efficiency.

The present invention discloses a method for retrieving atmospheric aerosol based on statistical segmentation. Firstly a multi-band remote sensing image including an apparent reflectance and an aerosol optical thickness look-up table corresponding to a retrieval band is obtained, then pixels are partitioned and screened according to apparent reflectance segments of a mid-infrared 2.1 micrometer band. After that the retained pixel sets are further partitioned and screened according to the apparent reflectance segments of the mid-infrared 1.6 micrometer band. Finally the obtained pixel sets are partitioned into two categories according to the pixel number, one category including pixels having more pixels, the other including those with less pixels.The category with more pixels is taken as the reference part for retrieval. Specifically, the pixel sets are first searched for the clean segment, then the ground surface reflectivity of the clean segment is taken as the ground surface reflectivity of the whole pixel set, thereby obtaining the aerosol thickness value through retrieval. After that these pixels are taken as references to perform retrieval on the other category.The present invention can improve accuracy and resolution of the retrieval result of the bright ground surface area, and is applied to a wider range.

The invention belongs to the technical field of traffic pedestrian image processing, discloses a pedestrian retrieval method and system based on semi-coupling metric discrimination dictionary learning of sketch images, establishes its own heterogeneous pedestrian retrieval database, and then performs feature extraction to extract features from images And mark different people with corresponding colors; process the extracted sample features, introduce dictionary learning method, learn dictionary pairs of heterogeneous data; learn mapping matrix from sketch image set and regular image set; introduce discriminant algorithm to learn metrics. The present invention has the advantages of solving the problem of lack of heterogeneous pedestrian data sets in the field of heterogeneous pedestrian retrieval and proposing semi-coupled metric discrimination dictionary learning (SMD) for the first time. 2 L) Technology. This technique can learn a semi-coupled mapping matrix from heterogeneous samples and dictionary pairs, which can reduce the differences between heterogeneous data to a certain extent. The ideal retrieval effect has been achieved on the new SINPID dataset.

The invention provides a reclaiming mechanical arm, and relates to the technical field of industrial mechanical arm equipment. The reclaiming mechanism comprises a rotating seat, a moving oil cylinder and a four-bar linkage, the four-bar linkage is connected with the rotating seat through a pin shaft, an operation head is driven to move up and down, the rotating seat is arranged on a rotating shaft, the rotating shaft is connected with a rotating mechanism in a rotational mode, and the moving oil cylinder is fixedly connected with the rotating seat and drives vertical bars of the four-bar linkage to move forwards and backwards. Machining and reclaiming of pipe members or thin arm pieces is achieved, reclaiming can be conducted inside the pipe members so that external damages of the pipe members are avoided, and the product pass rate is greatly improved; and the reclaiming mechanical arm is safe and reliable, economical and practical, easy to use, good in reclaiming effect and suitable for wide promotion.

The invention relates to a dust mask effect demonstration machine capable of realizing concentration regulation of different dust particles. The dust mask effect demonstration machine comprises a demonstration cabinet and a dust particle cabinet, wherein a communicating hole matched with the dust particle cabinet is formed in the left side wall of the demonstration cabinet, a dust mask is in fit in the communicating hole and is clamped by a mask clamping mechanism in the demonstration cabinet, a dust particle picking and placing device is arranged at a part, located below the communicating hole, of the dust particle cabinet, a base plate of the dust particle picking and placing device is a ventilation plate, a dust blowing air port is formed in the lower side of the dust particle cabinet,the dust blowing air port is communicated with a dust blowing fan, a concentration regulating air cylinder which is a vertical trend and penetrates the interior of the dust particle cabinet is arranged at the upper part of the dust particle cabinet, and a concentration regulating block in seal fit with the dust particle cabinet is connected below the concentration regulating air cylinder. The dustmask effect demonstration machine provided by the invention has the advantages that the concentration regulating block is controlled to go up and down by virtue of the concentration regulating air cylinder, further volume of a dust particle cavity is regulated, the effect of regulating concentrations of the dust particles is achieved, the condition that quantity of the dust particles is too largeand beyond a blowing range of the fan cannot occur, and then the good and convenient dust particle concentration regulating effect is realized.

The invention discloses a robot grabbing system for multi-type cloths. An adjustment assembly is movably connected with a truss; a clamping assembly is fixedly connected with the adjustment assembly,and located at one end away from the truss, of the adjustment assembly; a first slide element is slidably connected with the truss, and located in a first slide groove; one end of a first locking rodis fixedly connected with the first slide element; the other end of the first locking rod sequentially penetrates through a vertical plate and a first fixed plate, and is in threaded connection with afirst locking element; a third slide groove is formed in the vertical plate; one end of a second locking rod is fixedly connected with the vertical plate; and the other end of the second locking rodsequentially through the third slide groove and a second fixed plate, and is in threaded connection with a second locking element. The purpose of more ideal taking effect of the robot grabbing systemfor the multi-type cloths is achieved.

The invention relates to the field of fire-fighting equipment, in particular to automatic material pickup equipment for a small hole stamping die. The equipment comprises a bottom plate; a base is arranged at the upper side of the bottom plate; the base is connected with the bottom plate through a horizontal driving assembly; a fixing base is fixedly arranged at the upper side of the base; a firstsliding block is arranged at one side of the fixing base and connected with the fixing base through a vertical driving assembly; a moving arm is fixedly arranged on the sliding block; a switching rodis rotationally arranged at one end of the moving arm; a first power assembly for driving the switching rod to rotate is arranged in the moving arm; a gripper assembly and a suction cup assembly arearranged at the two ends of the switching rod correspondingly; a blowing mechanism is arranged at the upper side of the bottom plate; an arc-shaped plate matched with the blowing mechanism is arrangedat one side of the switching rod; two bending rods are fixedly arranged at the lower side of the arc-shaped plate; and a collecting box is arranged at the lower end of the arc-shaped plate. The automatic material pickup equipment can be used for automatic material pickup of the small hole stamping die, materials with different thicknesses can be picked up and placed, the working efficiency is high, meanwhile, a certain sweeping function is achieved on waste generated after stamping, and the waste can be collected.

The invention belongs to the technical field of cosmetics, in particular to an emulsification apparatus convenient for reclaiming and used for producing cosmetics. For the problem that the existing emulsification apparatus for producing the cosmetics is inconvenient for reclaiming,the invention provides the technical scheme as follows: the emulsification apparatus is characterized bycomprisinges an emulsification box body, wherein the outer wall of the top center of the emulsification box body is connected with a rotary motor by virtue of a bolt, an output shaft of the rotary motor is connected with a first rotating shaft which is vertically arranged by virtue of a bolt, and the circumferential top outer wall of the first rotating shaft is sleeved with a supporting ring by virtue of a bolt. The emulsification apparatus convenient for reclaiming and used for producing the cosmetics is provided with the rotary motor, and a supporting frame and a supporting column are driven by the rotarymotor to rotate, so that the emulsification of cosmetics production raw materials can be realized; and meanwhile, by arranging a circulating guide pipe and a first liquid suctionpump, the cosmetics raw materials which are insufficiently emulsified on the bottom of the apparatus can be circularly sufficiently emulsified, so that a sufficient and effective emulsification effect can be realized.

The invention discloses a name plate feeding mechanism for a solenoid valve member assembler. The name plate feeding mechanism for the solenoid valve member assembler comprises a name plate feeding frame; a name plate supplying device is mounted on the name plate feeding frame; a name plate feeding mechanical arm is mounted on the name plate feeding frame and above the name plate supplying device; the name plate supplying device comprises a feedstock mounting base plate mounted on the name plate feeding frame; the two sides of the upper plane of the feedstock mounting base plate are each equipped with a pair of displacement guide column mounting plates vertical to the feedstock mounting base plate; the upper parts of the side surfaces of each pair of displacement guide column mounting plates are each equipped with a name plate displacement guide column through a guide column support; the two name plate displacement guide columns are parallel to each other; each name plate displacement guide column is sleeved with a pair of linear sleeves; a first name plate mold mounting plate is fixedly arranged at the corresponding linear bearing on the corresponding name plate displacement guide column. Through the way, the name plate feeding mechanism disclosed by the invention is capable of realizing convenient feeding, high in work efficiency and long in service life.

The invention provides an airport storehouse luggage retrieval method based on an improved convolutional network. The method is characterized in that firstly, a network structure is improved, a new Asymbiotic and Mopping Convolution Block is provided, the structure extracts features through a plurality of branches to enhance the expression ability of a network convolution layer, and then the classification accuracy and robustness of the network to images are improved. The network training is carried out by using metric learning loss, so that the discrimination of luggage features is enhanced. Experimental results show that the provided luggage classification model has a good classification effect. In the image retrieval step, in order to enrich information contained in features, a multi-layer feature fusion method based on spatial pyramid pooling features and full connection layer features is provided, and a deep hash model capable of being used for luggage image retrieval is established in combination with a mixed loss function containing classification loss and quantization loss. Experiments show that the provided model has a better retrieval effect, the highest accuracy rate can reach 96.1%, and meanwhile time and space expenses are greatly reduced.

The invention discloses a self-adaptive personalized information retrieval system and method. For timely catching irregularly distributed dynamic retrieval requirements of a user, a retrieval module is timely updated through interaction of the user and a search engine. The system comprises a data input sub system, a parameter training and predicating sub system, a retrieval performing sub system and a data output sub system, wherein the data input sub system is used for combining historical inquiry information and historical click information to form a characteristic matrix according to the current inquiry information, and acquiring a training parameter predicating module according to the characteristic matrix; the parameter training and predicating sub system is used for training and applying the parameter predicating module to acquire the predicated parameters according to the characteristic matrix; the retrieval performing sub system is used for predicating the parameters to organize the current inquiry and the historical inquiry, and combining the user module and the inquiry module to form a personalized inquiry module; and the data output sub system is used for searching a document matched with the personalized inquiry from the document to be retrieved as a primary retrieved result, and sequencing the primary retrieved result according to the correlation to obtain the final retrieved result for outputting.