39results about How to "Accurate Classification and Identification" patented technology

The invention discloses a product packaging surface defect detection and classification method based on machine vision. The method comprises the steps of 1, acquiring a high-definition color image of defect-free product packaging, making the high-definition color image into a standard image, conducting real-time shooting with a camera, conducting online acquisition of a high-definition color image of product packaging to be detected, and making the high-definition color image as an image to be detected; 2, conducting image matching on the image to be detected and the standard image based on SURF algorithm; 3, conducting difference image operation on the two images matched in step 2 to obtain a defective image; 4, conducting feature extraction on the defective image to obtain the geometrical features and color features of the defective image; 5, classifying product packaging surface defects by means of RBF neural network algorithm. Automatic defect detection and classification are conducted by means of a machine vision system, human factor interference can be avoided, labor cost is reduced greatly, and then huge hidden cost caused by training and management when artificial detection is adopted is avoided.

The invention discloses a method for classifying and identifying coal gangue and coal in raw coal, which includes collecting an image of coal gangue as a training sample set 1, labeling the image as coal gangue, collecting an image of coal as a training sample set 2, and labeling the image as coal; extracting a coal gangue main body part in the training sample set 1 and a coal main body part in the training sample set 2 through a segmentation algorithm, and separating coal gangue and coal main bodies from the sample set image background; performing image decomposition on the new image sample set obtained in the step 2 by using wavelet transform, and extracting features of each frequency domain sub-graph; and taking a decomposed image obtained after wavelet transformation as the input of the convolutional neural network, and optimizing and updating the network weight parameter through multiple times of training, thereby obtaining the neural network parameter with the highest accuracy, and obtaining an optimal coal gangue and coal identification and classification model. The method can be easily arranged on a conveying belt for coal production, and is simple and convenient to operateand relatively low in cost.

The invention relates to a BiLSTM-based composite characteristic optical fiber sensing disturbance signal mode identification method, which comprises the following steps that vibration signals including optical fiber sensing disturbance signals in different modes are collected, data is stored, and type labels are added; the time domain feature extraction unit is used for calculating short-time energy and a short-time over-level rate for the acquired vibration signals, setting thresholds of the short-time energy and the short-time over-level rate, and preliminarily judging intrusion disturbancesignals according to judgment conditions; the frequency domain feature extraction unit is used for carrying out four-layer wavelet packet decomposition on each vibration signal, solving 16 sub-band energy spectrum distribution, splicing short-time energy and short-time over-level rate to form a composite feature vector, carrying out normalization processing on the composite feature vector, and taking the normalized feature vector as an input feature vector; and a bidirectional LSTM network model is constrcuted, the normalized feature vector is taken as an input, the event label is taken as aclassification output result, and a classifier is trained by using the test sample to realize optical fiber sensing disturbance signal mode identification.

The invention discloses a radio frequency interference suppression and classification method based on a convolutional neural network. The method comprises: initializing a transmission signal parameter, a target signal parameter and an interference signal parameter of an SAR system, calculating a target signal and an interference signal, superposing the interference signal on the target signal to obtain affected echo signals, and marking an interference type of the interference signal in each echo signal; carrying out discrete Fourier transform on the echo signals to obtain frequency domain forms of the echo signals, and carrying out classification based on a proportion to obtain a training set and a testing set randomly; inputting the training set into a convolutional neural network VGG16for training to obtain a test network; and inputting the testing set in the test network, and verifying a classification result of the testing set by the test network. Therefore, the classification and recognition of the interference signals by the echo signals are realized under the circumstances that the synthetic aperture radar works normally and the key parameters such as the signal parameter,the imaging range and the resolution are not changed.

The invention discloses a biological chip, testing method and testing agent box to classify Tephritidae, which selects probe of general-purpose probe and atopic probe or subgeneric general purposed probe and composite group/sibling species atopic probe to identify Tephritidae rapidly and precisely.

The invention provides an intelligent identification method and an identification device for quality of a silicon material, and belongs to the technical field of silicon material production. The intelligent identification method comprises the following steps: placing silicon materials on a conveying device, detecting surface morphology of the silicon materials by using a detection device in the conveying process, collecting morphology information, calculating morphology information to be identified and a similarity value of a morphology information sample on the basis of the collected morphology information by using a preset deep learning model, recognizing the morphology information to be identified, conveying a conveying device by a control device to respectively convey the silicon materials into corresponding accomodating devices according to different classes. The identification device comprises a conveying device for conveying the silicon materials, a detection device, a control device and a plurality of accomodating devices for accomodating classified silicon materials, wherein the detection device is matched with the control device to classify the silicon materials. By adopting the identification method of the identification device, secondary pollution of artificial identification and classification upon the silicon materials is avoided, and meanwhile relatively accurate identification is achieved.

The invention discloses a fine-grained image classification method, system and device, and a storage medium. The method comprises the following steps: obtaining to-be-classified image information; andinputting the image information into an identification model trained by adopting a local area to carry out fine-grained identification, and outputting an identification classification result. According to the method, the local area is adopted to carry out enhanced model training, so that the model is more suitable for fine-grained identification, and when two commodities are highly similar, the identification model can more effectively grasp key distinguishing information in the commodity images, so that the commodities can be rapidly and accurately identified and classified, and the method can be widely applied to the field of image data processing.

The invention discloses an SAR (Synthetic Aperture Radar) automatic target recognition method based on a variable convolutional neural network, which is applied to the field of radar target recognition and aims to solve the problems that scattering characteristics and morphological characteristics of an SAR image target in the prior art are not fully mined and utilized by a network and the overallidentification performance is not greatly improved. According to the invention, through data expansion and based on the idea of adding spatial sampling positions of extra offsets in the modules, sampling points of a convolution kernel in an input characteristic spectrum are deviated and concentrated in an interested area or target, the influence of different azimuth angles of an SAR target is overcome, the problem of insufficient SAR image sample size is alleviated, the condition of insufficient characteristic extraction is avoided, the generalization ability is high, the specific form and scattering information of the SAR target can be effectively utilized, and the target can be accurately identified and classified.

The invention provides a grain quality real-time detection device and method, and belongs to the field of agricultural intelligent equipment. The real-time detection device comprises a flowing grain image capture device, an embedded processor and an interactive display screen; a grain flow channel bottom plate in the flowing grain image capture device is fixed to an outlet of a grain conveyor through a fixing iron wire; the grain flow channel bottom plate is further wound with tightening bolts on the two sides of a grain collecting bin through adjustable iron wires; and therefore, the installation of the flowing grain image capture device is realized. In a real-time detection process, grains in the grain conveyor flow out, part of the grains are intercepted by the grain flow channel lowerbottom plate; a CCD industrial camera exposes the interior of a flow channel in a fixed period to capture images and transmits the images to an embedded processor; and the processor processes the images through the OPENCV and the XGBoost algorithm to obtain the real-time quality of the grains. The device and method are suitable for real-time detection of grains of different varieties in various application scenes.

The invention relates to a reverse dot blot (RDB) detection membrane strip, a detection method and a kit for identifying insects. The membrane strip comprises a universal primer, a solid phase carrier, a probe and a PCR positive quality control plasmid, wherein the probe is fixed on the carrier; the universal primer is a degenerate primer pair TF/TR; the solid phase carrier is a Biodyne C membrane; the probe fixed on the carrier is a group of arrays consisting of a quality control probe and an RDB detection probe; the quality control probe comprises a color development point probe, a positive comparison probe and a negative comparison probe; and the RDB detection probe is a section of 15 to 19nt oligonucleotide of which the 5' end is marked with amino (-NH2). The membrane strip can rapidly, correctly and vividly realize correct classified identification of adults, larvae, eggs, pupae and the like of insects and solves the problems of the need of professional classification knowledge, type specimen comparison, a large number of insect types and difficult identification existing in the conventional morphological insect identification, particularly the problem that the larvae, the eggs and the pupae of the insects and the adults of which the identification characteristics are damaged cannot be identified effectively.

The invention discloses a garbage recognition and classification method based on deep learning, and the method comprises the following steps: 1, carrying out the preliminary classification of garbage;2, respectively acquiring the preliminarily classified garbage data based on a data acquisition module. Through preliminary classification of garbage, metal and nonmetal in the garbage are separated,recognition is easier during classification and recognition, the influence of the nonmetal garbage on metal garbage classification is reduced, and then recyclable metal garbage classification is moreaccurate. Through data detection of the non-metal garbage, kitchen garbage and household garbage contained in the non-metal garbage can be conveniently recognized and separated, then only recyclablenon-metal garbage such as plastic, glass and paper is left, and when the recyclable non-metal garbage is classified, the influence of the kitchen garbage and the household garbage is reduced as well;and thus, the garbage can be identified and classified more accurately based on deep learning.

The invention relates to the technical field of peripheral blood cell detection, and discloses a disease association method of a peripheral blood cell morphology automatic detection system. The methodcomprises the following steps: scanning a peripheral blood cell smear; establishing a normal peripheral blood cell morphological standard and an abnormal peripheral blood cell morphological standard;acquiring a detection range set by a human-computer interaction operating system; performing feature extraction on the peripheral blood cell morphology of the scanned picture by using a convolutionalneural network method in a detection range; comparing, classifying and counting the extracted peripheral blood cell characteristics, and calculating sampling counting data; establishing an existing disease association database, associating the finally calculated data with the existing diseases, performing association analysis, and combining to obtain a final association result. According to the invention, the morphology of each cell in the peripheral blood cells can be accurately identified, and classified counting is carried out, so that the disease association result accuracy of the peripheral blood cell morphology automatic detection system is higher.

The invention discloses a mask wearing detection method based on a YOLOv5 network. The method comprises the following steps: step 1, preprocessing an original picture by using an image enhancement algorithm and dividing a data set; step 2, sending the training set pictures into a YOLOv5 network into which an attention mechanism is introduced for iterative training, thereby effectively enhancing extraction of key point information such as a human face and a mask; 3, in order to reduce overall errors, CIOU Loss is adopted as a loss function of target frame regression; 4, after training is completed, the optimal weight model is stored and tested on a test set. The result proves that under the clamping of image enhancement and an attention mechanism, the improved YOLOv5 model realizes the efficient detection of the mask wearing, not only can the face information be successfully detected, but also the mask wearing state is correctly detected, and the corresponding confidence is given. The mask wearing detection accuracy of the model under the conditions of low visibility and weak illumination intensity can reach 92%, and the model has important practical significance for epidemic situation prevention and control and public health safety maintenance.

The invention relates to a small sample semi-supervised learning method and device based on pseudo-label noise filtering, and the method comprises the steps: carrying out the alternative optimization of a main model and a pseudo-label filtering model, training the main model through a semi-supervised learning method, and training the pseudo-label filtering model through a noise label learning method. And the optimization of the main model and the pseudo label filtering model is alternately executed for a plurality of rounds, after each round, part of non-label data and the current pseudo label are classified into the data set with the label until the optimization is stagnated, and the optimized model is output. Compared with the prior art, the model obtained by the method has the advantages of better accuracy in the aspect of image recognition and the like.

The invention discloses a tea garden inspection system based on a visual technology and an inspection robot. The tea garden inspection system comprises the inspection robot, an acquisition device, a control device and a visual system, the inspection robot is controlled by the control device to move; the acquisition device is mounted on the inspection robot, and the image acquisition module is used for acquiring tea pictures and air pictures; the visual system comprises a data set and a model recognition module, the data set comprises a tea disease and insect pest image data set and an air quality data set, and the model recognition module recognizes current air quality and tea disease and insect pests by comparing real-time images acquired by the image acquisition module with the data set. A supervised lightweight MobileNetV2 model is established, the recognition precision is high, the calculation amount is small, a series of preprocessing is carried out on images collected by a high-definition camera, and then accurate classification recognition is carried out; the device is low in cost, small in size and low in power consumption; and a mobile terminal App is adopted, so that a control instruction can be conveniently sent.

The invention provides a large-scale network burst traffic identification model, a large-scale network burst traffic identification method and a training method of the model. The model is built on Spark through a TensorFlowOnSpark framework; the model comprises an input layer, a first convolution layer, a first maximum pooling layer, a second convolution layer, a second maximum pooling layer, a third convolution layer, a fourth convolution layer, a fifth convolution layer, a third maximum pooling layer, a full connection layer and an output layer which are connected in sequence, the input layer receives a 32 * 32 data input form; the first convolution layer adopts 96 pieces of 5 * 5 convolution kernels, and the step length is set to be 1; the second convolution layer adopts 192 pieces of 5 * 5 convolution kernels, and the step length is set to be 1; 384 pieces of 3 * 3 convolution kernels are adopted in the third convolution layer and the fourth convolution layer, and the step length is set to be 1; 256 pieces of 3 * 3 convolution kernels are adopted in the fifth convolution layer, and the step length is set to be 1; the pooling window of each maximum pooling layer is 2 * 2, and the step length is 2; 1024 nodes are adopted in the full connection layer; the output layer comprises two nodes.

The invention discloses a primer pair for constructing a molecular marker map of a cured tobacco variety of Hunan main tobacco varieties, and the sequence of the primer pair is as shown by SEQ ID No.1-46. The primer pair can be used for identifying cured tobacco varieties of Hunan main tobacco varieties, such as Xiangyan No.3, Xiangyan No.5, Xiangyan No.6, Xiangyan No.7, Yunyan 87, Yunyan 97, Yunyan 99, Yunyan 100, Yunyan 203, Zhongyan 100 yuan, Zhongyan 100, Guangdong tobacco 9, Gui tobacco No.1, Honghua Dajinyuan, Longjiang 911, Nanjiang No.3, NC729, G80 or K326. The primer pair is of great significance in helping tobacco enterprises to quickly, accurately and efficiently identify and grade the cured tobacco.

The invention provides a text multi-feature ambiguity resolution method and system, and belongs to the technical field of text recognition processing, and the method comprises the steps: inputting a to-be-resolved text into a trained resolution model, and carrying out the recognition and extraction of a combined ambiguity field in the text, and segmenting the extracted text according to context relevance and part-of-speech features of words in the text to obtain a text after ambiguity resolution, wherein the trained resolution model is obtained through training of a training set, and the training set comprises feature vectors composed of text weight features, context association features and part-of-speech features of the text where the ambiguous fields are located. The weight feature, the context word feature and the part-of-speech feature in the text are combined to generate the feature vector, a linear kernel function is adopted for classification by using a nonlinear SVM model, and finally a correct combined ambiguous segmentation mode is obtained, so that the classification and recognition of the text with multi-feature ambiguous fields are more accurate.