731results about How to "Avoid false detection" patented technology

The embodiment of the invention relates to a quick barrier detection method. The method comprises the steps that a two-dimensional grid map is established on the basis of point cloud data of a three-dimensional radar, the attributes of grids are calculated to determine impassable areas, points of which the scanning line gradient values exceed predetermined threshold values in the impassable areas are determined as barrier points, and correlation clustering is conducted on the barrier points by traversing the grid map.

The invention provides a method and a system for transmitting a control format indicating value. The method comprises the following steps: configuring indication signaling in the downlink control information DCI of a physical downlink control channel PDCCH on a current component carrier, wherein the indication signaling is used for indicating component carriers and indicating control format indicating values of the component carriers or a mode of acquiring the control format indicated values, and the indicated component carriers are used for transmitting the physical downlink shared channel PDSCH corresponding to the PDCCH; and transmitting the DCI to terminal UE. Through the method, the problem that possible combinations of CFI and CI cannot be completely indicated is solved, and the technical effects of avoiding CFI false detection, relatively less overhead and the like are achieved.

The invention discloses a flexible general production line and a production method of an electronic product. The production line is characterized by comprising a main flow line and an intelligent control system of a flexible production line, wherein the main flow line comprises n independent functional module tables, at least one high-temperature aging table and a mobile conveying line; each functional module table has a same structure and comprises a scanner, a test part, a frock clamp operation part, a maintenance isolation part, a computer information system and a display; the high-temperature aging table comprises a high-temperature aging box body and a computer control system; and the computer information system of each functional module table and the computer control system of the high-temperature aging table are in two-way electric connection with the intelligent control system of the flexible production line respectively for information interaction, so intelligent control of the product in the whole production process is accomplished by verifying an identification code of each process of a produced target product and identifying whether the process is qualified. The functional module tables can be increased or decreased according to the type of the produced product, so that the formed production line is simple and convenient.

The invention discloses a multimedia video object processing method which comprises the steps as follows: (1) carrying out scene dividing on an MPEG video based on macro block information; (2) pre-reading a video needed to be jointed, obtaining various information and searching a proper joint scene; (3) searching the inlet point and the outlet point of the joint and carrying out regulation on various information of the accessed video; (4) selecting a proper audio joint point to realize audio-video seamless joint; (5) setting a video buffer area and unifying the code rate of the video to be jointed; (6) carrying out coarse extraction on a moving object in the video in a time domain; (7) carrying out watershed processing on the coarse extraction result, carrying out space region merging and leading to a accurate segmentation object. The .invention is characterized by simple and high-efficiency algorithm, low system resource consumption and fast speed and high accuracy of processing.

The invention relates to a video monitoring system with an elevator invalid request signal removing function. Elevator waiting hall video acquisition modules are mounted in elevator waiting halls in each floor, and a car video acquisition module is mounted in a car. Images acquired are respectively transmitted to the elevator waiting hall video processing modules and a car video processing module. Whether people are in elevator waiting halls or in the car or whether the car is in full load conditions or not can be detected by comprehensively utilizing various image processing technologies such as human-body judgment and 'space full-load detection' technology. Further, processing results are transmitted to a CAN (controller area network) bus via a CAN module, and received by an invalid request processing module for further processing. For invalid requests, the invalid request processing module transmits removal commands to an elevator master controller, invalid stopping of the elevator in various conditions can be avoided, running efficiency of the elevator is improved and energy loss is reduced.

The invention relates to a rapid imaging detection method for gear appearance defects. The rapid imaging detection method comprises two stages including a pre-processing stage and an online detection stage. The pre-processing stage comprises three procedures including automatically setting a target region, generating a coarse detection template and generating a fine detection template. The online detection stage comprises the steps of: J1) acquiring an image; J2) preprocessing and extracting features; J3) carrying out area difference value calculation on the image to be detected and an image of the coarse detection template for coarse detection judgment; J4) registering the image to be detected and an image of the fine detection template so that gear tooth directions of the two images are coincided; J5) and carrying out exclusive-or operation on the image to be detected and the image of the fine detection template to obtain a differential image, and acquiring defect region area in the differential image for fine detection judgment. The rapid imaging detection method for gear appearance defects acquires images by utilizing machine vision, performs coarse and fine judgment by utilizing template matching and image exclusive-or operation, and is high in detection precision and fast in speed.

The invention discloses a detection method of electrocardiosignal R characteristic waves. A wavelet decomposition method for improving threshold value is carried out on collected electrocardiosignals to conduct filtering, then wavelet reconstruction protruding out of a QRS wave group is carried out, energy window conversion from time domain analysis to energy domain analysis is carried out, then a maximum value point is selected and optimized, and finally an wave selection is carried out according to R selection logics. The detection method uses multiresolution characteristics of wavelet analysis to decompose signals on different scales according to frequency to accordingly process the signals on the different scales in a targeted mode and increase algorithmic flexibility. The energy window conversion method is adopted to convert the signals from a time domain to an energy domain, effectively restrains interference of high-frequency noise and improves algorithmic stability. By means of wavelet reconstruction, the QRS wave groups are extracted, P waves and T waves serve as noise to be rejected, false detection caused by huge P waves and T waves in detection can be effectively avoided, and detection accuracy is improved.

The invention relates to a vehicle environment sensing system and method based on an unmanned plane, and the system and method are used for quickly acquiring the road environment information around a vehicle. The sensing system comprises the unmanned plane and an earth station, the unmanned plane id connected with the vehicle through a mooring rope, the road surface is shot by the unmanned plane, and a video image is transmitted to the earth station; the earth station is arranged in the target vehicle, the earth station can process the video image to obtain the vehicle environment information, and the earth station also can calculate the flight control parameter and send the parameter to the unmanned plane to make the unmanned plane and the vehicle move together. The sensing method is based on the sensing system, the unmanned plane takes off and hovers above the vehicle, the unmanned plane establishes a communication for sensing the environment and transmits the video image to the earth station, the earth station processes the video image, calculates the vehicle environment information and transmits the information to the vehicle, and the earth station calculates the flight control parameter and transmits the parameter to the unmanned plane. According to the invention, the system can perform the wide area observation in the direction vertical to the vehicle plane direction, the coverage is wide, the road environment information around the target vehicle can be easily sensed, and the distance between the objects in the image can be precisely calculated.

The invention discloses a remote sensing image change detection method based on multi-feature proof integration and structure similarity, comprising the following steps: A. data preparation: selecting remote sensing images to be carried out with change detection in different time phase and in the same spatial scope; B. feature extraction: respectively extracting features of textures, grads and edges of images in two time phases; C. structure similarity calculation: calculating the texture structure similarity, grads structure similarity and edge structure similarity of the remote sensing images in two time phases in a working window; D. basic probability assignment function determination: building basic probability assignment function (BPAF) of evidence theory based on texture structure similarity, the grads structure similarity and edge structure similarity; and E. forming evidence and obtaining change areas to realize change detection. The method carries out change detection based on the area statistic features of edges, texture sand grads of images and can effectively avoid detection miss and false detection caused by the change detection using single feature; and the algorithm has robustness.

The invention discloses a holographic detection system for a steel wire rope. The holographic detection system for the steel wire rope comprises a detection device, a journey metering device, a data acquisition transition work station, an alarm controller, a terminal control master station and the like. The detection device comprises a steel wire rope electromagnetic detecting device, a steel wire rope damage machine vision device, a steel wire rope diameter measurement machine vision device, a damage position marking device and the like. The holographic detection system for the steel wire rope has technical advantages in many aspects and will bring good market prospects and economic benefits.

The invention provides an eyeglass detection method and device. The method concretely comprises the following steps of: acquiring an eyeglass region image from a face image to be detected; carrying out primary discrimination on the eyeglass region image according to a single-eye region classifier, and detecting according to a primary discrimination result, wherein the detecting process comprises the following steps of: if the absolute value of the primary discrimination result is more than a preset confidence threshold, outputting a detection result indicating whether a pair of eyeglasses are worn or not according to the primary discrimination result; and if the absolute value of the primary discrimination result is less than or equal to the preset confidence threshold, carrying out secondary discrimination on the eyeglass region image according a double-eye region classifier, and outputting the detection result indicating whether the eyeglasses are worn or not according to the secondary discrimination result. By adopting the method provided by the invention, the requirement on an initial face position is low, and the accuracy of eyeglass detection can be improved.

The invention discloses a keyword detection method in a voice signal, a device, a terminal and a storage medium, and belongs to the terminal technical field; the method comprises the following steps:obtaining a posterior probability of each target character that serves as a keyword in the voice signal in an acquisition period; obtaining the confidences of at least two target character combinations according to the posterior probability of each target character; determining the voice signal contains the keyword when the confidences of at least two target character combinations respectively satisfy preset conditions. The method obtains the confidences of at least two target character combinations, and determines the voice signal contains the keyword when the confidence of each target character combination satisfies preset conditions; the method obtains the confidence according to the posterior probability of each target character, and detects according to the confidence obtained by theposterior probability of certain target character, thus preventing wrong detection, and providing a more accurate detection result.

The invention relates to a steel rail crack detection method based on multiple acoustic emission event probabilities. According to the steel rail crack detection method, the relative probability output by a convolutional neural network is used as the probability of an acoustic emission event, and the problem that temporal information between samples is not fully used by an existing steel rail crack detection method is solved. The steel rail crack detection method comprises the steps of (1) loading an acoustic emission time domain signal data matrix, and performing FFT (Fast Fourier Transformation) and pretreatment on acoustic emission signals, so that a spectral matrix which is folded into a three-dimensional matrix and a label vector are obtained; (2) setting structural parameters and an initial value of the convolutional network; (3) inputting the spectral matrix, calculating and iterating errors of a convolutional neural network model layer by layer, updating a weight matrix and bias, performing feature extraction, and outputting classification results and classification probabilities of a test set; (4) correcting the outputting of the convolutional neural network on the basis of the multiple acoustic emission event probabilities, and optimizing the classification results. According to the steel rail crack detection method, the classification results are improved according to the multiple acoustic emission event probabilities, so that the detection precision of steel rail crack damages is increased, and high theoretical and practical engineering significance is obtained.

The invention discloses a full-automatic intelligent sheet selection machine for neodymium-iron-boron magnetic workpieces based on machine vision. The full-automatic intelligent sheet selection machine comprises a rack, wherein an upper computer, a PLC (Programmable Logic Controller), a communication conversion card and a main load-bearing steel plate are fixedly mounted on the rack; a loading transmission belt, a servo motor, a glass turntable, an approach switch, three solenoid valves, three pneumatic nozzles, three discharging cylinders, four image collection cameras, a coaxial light source and a backlight source are mounted on the main load-bearing steel plate; the servo motor is used for driving the glass turntable to rotate at a constant speed, thereby driving the workpieces to carry out uniform circular motion and sequentially pass through the approach switch, the pneumatic nozzles and the image collection cameras; the parts respectively carry out corresponding actions under the control of the PLC. The full-automatic intelligent sheet selection machine for neodymium-iron-boron magnetic workpieces is capable of effectively integrating hardware modules from the aspects of hardware structures based on the principle of machine vision, thus solving the problems of low detection accuracy, slow speed and high costs for detecting the neodymium-iron-boron magnetic workpieces.

The invention discloses a monocular vision lane line detection method and a distance measurement method of the monocular vision lane line detection method and belongs to the field of vehicle driving safety auxiliary systems. The lane line detection method comprises the steps of shooting multiple frames of images when a vehicle is just started to obtain a reference lane line, then obtaining candidate real-time lane lines by every reading of a real-time image, performing region matching on the candidate real-time lane lines and the reference lane line, and if the region matching is successful, updating an inclination angle and outputting the reference lane line; if the region matching is failed and the failure emerges for the first time, recording the candidate real-time lane lines as temporary lane lines, if the failure emerges continuously and the reference lane line can be continuously successfully matched with the temporary lane lines, outputting the temporary lane lines. A distance between a body vehicle and a target vehicle can be calculated by setting a reference target on the basis of a projection relation between the reference target and the target vehicle in a vision collection system according to the lane line image detected by the method disclosed by the invention. The monocular vision lane line detection method and the distance measurement method of the monocular vision lane line detection method are suitable for vehicle lane line detection and distance measurement between two vehicles.

The invention discloses a vacuum leak detection method and a vacuum leak detection system for containers. The vacuum leak detection method comprises the steps of placing a container to be detected in an airtight chamber, quantitatively pumping the airtight chamber, detecting the vacuum degree in the airtight chamber and judging whether the container leaks after a preset period of time, and the like. The vacuum leak detection system comprises an air pump, an airtight chamber used for accommodating an object to be detected, and a pumping channel connecting the air pump and the airtight chamber. A detection device used for detecting the vacuum degree is connected to the pumping channel. The vacuum leak detection system further comprises a quantitative control assembly used for control. The vacuum leak detection method and the vacuum leak detection system of the invention have the advantages of accurate and reliable detection, high detection precision, and the like.

The invention provides a grain storage pest detection and identification method based on video analytics and a corresponding detection and identification device. The method of the invention comprises the following steps: S1. obtaining the multiple frames of continuous images of a moving grain storage sample to be detected; S2. cutting each frame of image into a grain pest area and a background area; S3. in each frame of cut image, defining a grain pest searching area by taking the grain pest as the center, wherein the grain pest searching area is composed of M*M grain pest areas; S4. based on a preset match criteria, searching an area matched with the pest area of the Nth frame of image in the grain pest searching area of the (N+1)th frame of image, and recording the matching degree value between areas; and S5. identifying the grain pest area of which the accumulated matching degree value exceeds a preset threshold value as the area with grain pests. The method can overcome the defects of false detection and missed detection existing in the existing detection and identification method and can judge grain pest survival and types.

The invention discloses a vehicle safety belt detection method and device based on image analysis. The vehicle safety belt detection method and device based on image analysis is applied to the image analysis field of intelligent transportation. Firstly, a vehicle widow area is positioned; the position of a safety belt is detected primarily through a line segment detection method; the safety belt is accurately positioned according to the angle and the length of the line segment. The vehicle safety belt detection method and device based on image analysis can store the information of vehicles with passengers not belted up, and provide sufficient, reliable and strict evidences for detection of the traffic offence of not fastening the safety belt.

The invention discloses a target recognition and tracking system based on multi-source video integration and belongs to the technical field of image processing and automatic control. A first video source acquisition module and a second video source acquisition module carry out video acquisition on the same scene respectively by using a digital thermal-imaging camera and a digital visible-light camera and transmit digital infrared video signals and digital visible-light video signals to a target recognition and tracking module; the target recognition and tracking module carries out integration on infrared video frames and visible-light video frames by using a multi-resolution-based multi-source video integration method and then carries out target recognition and target movement trajectory tracking on integrated video frames; a signal display module is used for displaying the movement trajectory of a target in real time. The target recognition and tracking system based on the multi-source video integration has the advantages that the system is simple in design, so that the realization is convenient; the size is smaller, so that the carrying is convenient; a DSP (Digital Signal Processor) and an ARM (Asynchronous Response Mode) serve as a processing chip, so that the operating speed is high, and the requirements on real time can be reached.

The invention provides an evaluation method for evaluation of a contact hole. The failure evaluation method comprises the following steps of: providing an evaluation wafer and forming an evaluation device graph on the evaluation wafer, wherein the evaluation device graph comprises a first device region and a second device region which has the same structure and size as a PMOS (P-type Metal Oxide Semiconductor) transistor; the difference between the first device region and the NMOS transistor lies in that a P trap corresponding to an NMOS (N-type Metal Oxide Semiconductor) transistor has different doping types, so that the conductivity of an N-type doped active region in the first device region is stronger than a source / drain electrode of the NMOS transistor when the electron beam is scanned; carrying out a forming process of the contact hole which is the same as the product of wafer on the evaluation wafer; and scanning the surface of the evaluation wafer by adopting the electron beamand acquiring the scanning brightness at each part of the wafer surface. According to the evaluation method disclosed by the invention, a brighter light spot is formed when the contact hole is scanned by using the electron beam, thereby the brighter light spot is easily distinguished from a surrounding region and the leakage detection or mistaken detection of the failure contact hole caused by a machine board is effectively avoided.

The invention provides a straight segment detecting and extracting method. The method comprises the following steps: S1, scanning a binary image, calculating a runlength histogram on the peripheries of pixels of the binary image, and extracting the corresponding runlength direction and length characteristics; S2, constructing an arc-shaped neighborhood structure for each foreground pixel point onthe binary image, and detecting local straight segments by utilizing the runlength characteristics on the peripheries of the pixels; and S3, clustering and combining the detected local straight segments to obtain a global long straight segment. Compared with the traditional method, the technology ensures that false detection and missed detection in the traditional straight segment detection method can be effectively avoided only by setting the minimum length of the straight segment to be detected. The method is applied to automatic detection and extraction of graphic targets in a bill image, and has a wide application prospect in the fields of graphic vectorization, form analysis, bank bill automatic processing and the like.

The invention discloses a paper feed machine for a product check machine and a paper feed method of the paper feed machine. The paper feed machine comprises an uninterruptible paper feed mechanism. An induced draft paper supply mechanism is correspondingly arranged above the upper paper outlet end of the uninterruptible paper feed mechanism. A paper aligning plate is vertically arranged on one side of the uninterruptible paper feed mechanism and corresponding to the output end of the induced draft paper supply mechanism. The output end of the induced draft paper supply mechanism and the input end of a blowing paper distributing mechanism are correspondingly arranged. The paper feed method includes the steps that a main paper feed table and an auxiliary paper feed table alternatively support and lift a printed matter stack with printing surfaces placed upwards; printed matter is uninterruptedly supplied to the induced draft paper supply mechanism; the induced draft paper supply mechanism continuously adsorbs a single sheet of printed matter from the position above the printed matter stack and transfers the single sheet of printed matter to the blowing paper distributing mechanism, and then the blowing paper distributing mechanism sends out the single sheet of printed matter after separating the single sheet of printed matter from the induced draft paper supply mechanism. By the adoption of the mode of adsorbing paper feed with the printing surfaces placed upwards, the single sheet of printed matter is uninterruptedly conveyed to a surface quality detector for surface quality detection, and therefore scratches of the printing surfaces of the printed matter are reduced or eliminated.

The invention provides an adaptive collection empirical mode decomposition based electrocardiosignal characteristic point identification method. The method comprises the steps that two parameters of Ratio and NEEMD combining empirical mode decomposition are calculated adaptively according to the self characteristics of to-be-identified electrocardiosignal, AEEMD noise-reduction processing is conducted on the to-be-identified electrocardiosignal according to the determined parameters, an electrocardiogram noise-reduction signal is obtained, AEEMD noise-reduction processing and characteristic point identification of the electrocardiosignal are combined, hierarchies large in QRS waveform energy ratio are extracted from intrinsic mode function components of all hierarchies obtained in the AEEMD noise-reduction process are overlapped to serve as a detection layer, after difference operation is conducted, an adaptive section division mode is adopted for conducting section processing, electrocardiogram difference section signals are obtained, and finally, by combining the electrocardiogram noise-reduction signal with the electrocardiogram difference section signals, positioning identification of QRS waves in the to-be-identified electrocardiosignal is achieved according to distances between all the characteristic points in the electrocardiosignal and a waveform slope relation; the amount of computation can be subjected to optimal identification, the identification efficiency is improved, and the identification accuracy is improved.

The invention discloses an integrated glue coating machine and an operation method thereof. The integrated glue coating machine comprises a screwing device, a glue coating device, a detection device and a robot, wherein the screwing device comprises a screwing shaft capable of moving in the vertical direction, and the end portion of the screwing shaft is detachably provided with a sleeve matched with a bolt; the glue coating device comprises a quantitative glue gun, the glue inlet end of the quantitative glue gun is connected with a glue storing tank, and the glue outlet end of the quantitative glue gun is connected with a glue needle; the detection device comprises a detection part used for workpiece positioning, machine type recognition and glue coating qualification degree detection; and the robot is provided with an operation end, and the screwing device, the glue coating device and the detection device are all connected with the operation end. The screwing device, the glue coating device and the detection device are all integrated on the robot, the function that one machine is multipurpose is achieved, the floor area of equipment is reduced, the production efficiency is improved, meanwhile, the integrated glue coating machine can be better compatible with different machine types, flexible production and equipment maintenance are facilitated, and the machine replaces the manual work to meet the real demands of clients for reducing the production cost and improving the production efficiency.

The invention discloses a device for scanning vehicles in three-dimensional mode through laser. The device comprises a laser emitting module for emitting a laser beam; a vertical movement control module for controlling the laser beam emitted by the laser emitting module to swing along a direction perpendicular to a vehicle moving direction to generate a laser scanning face; a parallel movement control module for controlling the laser scanning face to swing along the vehicle moving direction to generate a three-dimensional scanning area; a laser receiving module for receiving a reflecting light signal reflected when the vehicles pass through the three-dimensional scanning area; a data conversion module for converting the reflecting light signal into a distance signal; and a data analyzing module for generating vehicle information according to the distance signal. The method and device can determine vehicle shape correctly, improves traffic condition survey quality and avoids phenomena of detection negligence and mistaken detection and the like.

The invention relates to an ultrasonic, vortex and EMAT integrated lossless thickness tester and a method thereof. The ultrasonic, vortex and EMAT integrated lossless thickness tester comprises an excitation module, a receiving pre-processing module, an analog switch, an A / D converter, a self-checking module, an MCU, a CPU module, a power supply module and a CPU periphery display storage part, wherein the excitation module is connected with the receiving pre-processing module, the receiving pre-processing module is connected with the analog switch, the analog switch is connected with the A / D converter, the A / D converter is connected with the MCU, the self-checking module is connected with the MCU, and the excitation module comprises an ultrasonic excitation module, a vortex excitation module and an EMAT excitation module; the receiving pre-processing module comprises an ultrasonic receiving pre-processing module, a vortex receiving pre-processing module and an EMAT receiving pre-processing module. The comparability of a measurement result is increased through the integration of the ultrasonic, the vertex and the EMAT, the misjudgment rate caused by measurement through a single method is reduced, and meanwhile the limitation of measurement through the single method is overcome.

The invention discloses an earphone, which comprises an earphone cable and receivers, wherein each receiver is provided with a distance sensor for detecting a distance between the ear and the receiver; and the distance sensors are connected with the earphone cable. The invention also discloses terminal equipment. The terminal equipment is connected with the earphone cable of the earphone through an earphone jack, and is provided with a circuit for receiving and processing signals of the distance sensors. The invention also discloses a method for controlling audio presentation through the earphone. By detecting the earphone, the audio representation is automatically controlled, and the earphone can provide convenient and intelligent experience for a user.

The invention discloses a target detection and motion state estimation method based on vision and laser radar, and the method comprises the following steps: 1, installing the positions of a camera andthe laser radar, carrying out the calibration between every two binocular cameras, and carrying out the combined calibration between a left camera in front of a vehicle and the laser radar; fusing and supplementing the sparse point clouds respectively generated by the stereoscopic vision system and the laser radar to form a global three-dimensional point cloud graph; 2, detecting the two-dimensional image and the three-dimensional point cloud by using a deep learning method, and fusing an image target three-dimensional motion state generated by stereoscopic vision with a laser radar three-dimensional point cloud target to obtain comprehensive target three-dimensional feature points; and 3, tracking the target by utilizing Kalman filtering according to the three-dimensional feature pointsof the target, and analyzing the motion state of the target. The key of the method lies in the fusion of vision and laser radar sensors, and improves the sensing capability of an intelligent automobile for the surrounding environment.

The invention discloses a honeycomb paper core defect detection method based on machine vision. Aims at various defect problems generated during a honeycomb paper core production process, the method comprises the following steps of acquiring a honeycomb paper core picture on a production site; detecting the defects in the honeycomb paper core by adopting an SSD deep neural network; judging the defect category of the honeycomb paper core and outputting the specific position, then using a machine vision algorithm for rapid reinspection to prevent the false inspection, finally transmitting an obtained result to the honeycomb paper core defect repairing system, and providing a correct feedback signal to realize the automatic repairing of the honeycomb paper core defects. The method detects thehoneycomb paper core defects in real time through the deep learning model and the machine vision algorithm, can provide the feedback information for an automatic defect repairing system during the honeycomb paper core production process, has the advantages of being accurate in recognition, accurate in positioning and fast in recognition speed, and can meet the requirement for honeycomb paperboardproduction automation.

The present invention discloses a pedestrian detection method based on semantic segmentation information, and relates to the field of a pedestrian detection method based on a neural network. The pedestrian detection method based on semantic segmentation information comprises the steps of: 1: inputting original RGB images in training set samples into a backbone network, inputting corresponding semantic segmentation images into branch networks, and setting a loss function of the whole network to complete training; 2: inputting the original RGB images in THE training set samples into the backbonenetwork which completes training to perform convolutional feature extraction and generate a multi-layer feature map; 3: inputting the multi-layer feature map into an area generation network which completes training to perform pedestrian candidate frame extraction and generate a pedestrian candidate area; and 4: performing classification and location of the pedestrian candidate area by a classification regression network which completes training to output detection result images including a pedestrian position surrounding frame. The problem is solved that pedestrians and backgrounds are difficult to be distinguished in a low-resolution condition to cause low detection precision in the current pedestrian detection, and the precision of pedestrian detection is improved in the low-resolutioncondition.