111results about How to "Does not affect detection accuracy" patented technology

The invention relates to a wide distance beam parallelism detecting apparatus, comprising a photoelectric auto-collimator and a diameter-expanding collimating arm. The apparatus can detect the parallelism of the wide distance parallel beams and the detecting distance can reach 1.5m; and two parallel beams do not need to be contained into the caliber of the photoelectric auto-collimator so as to reduce the caliber of the photoelectric auto-collimator and largely reduce the manufacturing cost of the photoelectric auto-collimator. Simultaneously, the adoption of pentaprism and optical wedge largely reduces the processing precision requirement of the pentaprism and has no impact on the detecting precision of the whole apparatus and higher precision can be provided. The use of an optical splitting system in the photoelectric auto-collimator leads the detecting wave band of the apparatus to be expanded to the laser and infrared spectrum band from the white spectrum band so as to become a multi-spectrum optical system.

The invention discloses a luminance real-time adjusting device of a light source in pole piece surface defect detection and an adjusting method thereof. The luminance real-time adjusting device comprises an LED light source, a contrast color swatch, an image collecting device and a light source constant-current driver. The light source luminance real-time adjusting device based on close loop control enables the light source to be capable of being adjusted in luminance in real time according to the joint function of light of outside environments and the light source on the contrast color swatch so as to enable a detected pole piece to be adjusted in an optimized illumination state, and the quality of images is ensured.

A urine comprehensive test method is provided, which is to combine three urine test methods comprising the urine dipsticks analysis, the flow cytometer urine sediment analysis and the microscopic examination on urine sediment. First, test the urine by the urine dipsticks analyzer and the flow cytometer urine sediment analyzer; second, judge if the microscopic examination under the doctor's advice is needed if the urine is tested to be the patient suffered from renal diseases; third, make a smart judgment to the test result of the flow cytometer urine sediment analysis; finally, make comprehensive analysis to the test result of the urine dipsticks analysis and the flow cytometer urine sediment analysis to see if the microscopic examination is needed on he urine sediment. The method can fast and correctly determine the partial important samples needed for manual urine sediment check by the microscopic examination, which saves the human and material resources with reliable test result, and obviously improve the efficiency of the urine test without any influence on the test accuracy.

The invention discloses an angle detection device based on machine vision. The angle detection device comprises a casing and a rotary table arranged inside the casing, wherein a rotating shaft is arranged in the center of the rotary table; at least two identifications with same shapes but different sizes are arranged on the rotary table; the at least two identifications are overlapped in centers and are circles, parallelograms, regular triangles or polygons with sides of positive even numbers; at least one encoding strip identification is arranged between the identifications; center lines of encoding strip identifications penetrate through the center; codes corresponding to every two encoding strip identifications are different; and the device further comprises an image acquisition device used for acquiring an image of the rotary table and a lighting source. The invention further provides an angle detection method based on the machine vision at the same time. The angle detection device is simple in structure, high in pollution resistance and higher in impact resistance of the rotary table. According to the angle detection method, detection errors caused by pollutants can be identified and removed through an image processing technology, and the detection accuracy is high.

The invention discloses a method for detecting wireless sensor network sparse events based on compressed sensing and the game theory, which belongs to the field of wireless communication and aims to solve the problem that excessively intensive sensor nodes randomly released in the wireless sensor network probably cause communication interface but sparse sensor nodes probably cause detection leak.The method disclosed by the invention comprises the following steps: (I) dividing M sensor nodes randomly released in the network system into a layers according to the thought of the game theory; (II) determining to activate one or more layers of sensor nodes according to the number (K) of the current events; (III) receiving a data vector y by a data terminal of the network system, and obtaining a measurement matrix [phi]P*N according to the detection process; and (IV) obtaining the specific distribution vector E of K happening events in N event sources by use of a greedy sparse event detection algorithm.

The present invention is antiwear conductive polymer paint for detecting high voltage cable plastic sheath and its preparation and application, and belongs to the field of conducting paint and wire and cable producing technology. The antiwear conductive polymer paint is compounded with component A and component B in the weight ratio of 3-5 to 1. The component A consists of thermosetting polymer resin 35-54 wt%, conducting stuffing 35-62 wt%, solid lubricant 0.5-7.0 wt%, coupling agent 0.1-2.0 wt%, organic thixotropic agent 0.1-2.0 wt% and defoaming agent 0.1-2.0 wt%; and the component B is curing agent phenolic amine T31. The antiwear conductive polymer paint is coated to the outer surface of polyvinyl chloride sheath of high voltage transmission cable to form coating of 0.2-0.8 mm thickness through curing at 20-100 deg.c, and the coating is suitable for DC voltage withstand detection.

The invention relates to an ultrasonic phased array imaging detection method for steam turbine blade root bolt holes. The detection method comprises the following steps that a first phased array probe is arranged on an inner arc side shoulder on the air inlet side of a blade root, and the air outlet side of the front fork tooth bolt hole and the air outlet side of the rear fork tooth bolt hole are detected; a second phased array probe is arranged on an outer arc front side shoulder on the air outlet side of the blade root, and the air inlet side of the front fork tooth bolt hole is detected; a third phased array probe is arranged on the outer arc rear side shoulder of the air outlet side of the blade root, and the air inlet side of the rear fork tooth bolt hole is detected. The ultrasonic phased array imaging detection method for the steam turbine blade root bolt holes has the advantages that the blade root of a steam turbine does not need to be demounted and mounted, and thus the assembling cost and the time cost of halt operation are saved while the detection precision is not affected. After parameters of a phased array detector are adjusted and stored, a detection result can be obtained without moving of the probes or with little moving of the probes after the phased array probes are attached to a workpiece, the detection efficiency is high, and time cost is saved.

The invention relates to a pedestrian detection method based on GMM background difference and combined features, and belongs to the technical field of video monitoring and recognition. The method comprises the steps: modeling the background of a video image through a hybrid Gaussian model, obtaining a background image, carrying out the difference calculation with a current frame of a video, obtaining the position of a moving foreground object, determining a to-be-detected region of the image, carrying out the detection of the to-be-detected region through a classifier trained through the combined features, and finally obtaining a block diagram of a pedestrian. The method can effectively shorten the search region of a sliding window of the classifier and reduces the number of search times of the classifier. The method quickly obtains the specific position of the pedestrian under the condition of determining the moving object, and also has a better detection effect for a group of pedestrians. The experiment proves that the method, compared with a conventional method, can achieve a better effect.

The invention relates to an image processing technology, is applied to the field of smart medical treatment, and discloses an abnormal cell automatic labeling method. The method comprises: performingGaussian convolution smoothing processing on a cell pathology picture to obtain an image pyramid; performing detection fitting on the image pyramid through a Hough transform circle detection method toobtain a low-power fitting region of interest; mapping the low-power fitting region of interest to the acquired cell pathology picture to generate a fitting high-power image; segmenting a region of interest in the fitting high-power image to generate a segmented high-power image; labeling and cutting the segmented high-power image through an adaptive threshold segmentation algorithm to obtain anabnormal cell labeling set; and mapping the abnormal cell labeling set to the cell pathology picture to obtain an abnormal cell labeling picture. The accuracy of abnormal cell labeling is improved, and the calculation and storage pressure is reduced. In addition, the invention also relates to a blockchain technology, and the abnormal cell label set can be stored in a blockchain.

The invention discloses a pair of shoes for tumble detection, and tumble detection method and device thereof, wherein the device comprises a pressure sensor, a first photosensitive sensor, an alarm and a controller. The pressure sensor is arranged at the bottom of a shoe and used for detecting pressure generated by a person to the shoe; the first photosensitive sensor is arranged at the sole of the shoe and used for detecting light projected to the sole; the alarm is used for giving out an alarm prompt under the condition that a person wearing the shoe tumbles; the controller is connected with the pressure sensor, the first photosensitive sensor and the alarm respectively and used for controlling the alarm to give an alarm according to signals detected by the pressure sensor and the first photosensitive sensor. By adopting the first photosensitive sensor to detect the light rays of the sole and the pressure sensor to detect the pressure of the sole, the device can judge whether a person tumbles or not by combining the sole light ray change and the sole pressure change, accordingly, on the basis of not influencing the detective accuracy, the device reduces the erroneous judgment probability, and is suitable for any shoes.

The invention provides a multi-station detection device based on automatic control, belongs to the technical field of detection devices, and aims to solve the problem that errors are easy to generate after a servo motor or other angle motors rotate in one direction for a long time when a multi-station turntable in the existing multi-station detection device rotates. The device comprises a workbench. The upper part of the left side of the workbench is provided with a detection member through a pillar, and the middle side of the upper part of the workbench is provided with a multi-station turntable. The multi-station rotating disc comprises a supporting shaft sleeve and a worm wheel, and a rotating shaft is arranged in the center of the bottom end face of the multi-station rotating disc. After the worm rotates three circles anticlockwise, because the transmission ratio between the worm and the worm gear is 18: 1, the worm gear drives the multi-station turntable to rotate one sixth circle, namely 60 degrees, so that the rotation angle of the multi-station turntable does not generate errors, the detection precision of a detection piece on the multi-station detection device is not influenced, and the reliability of the multi-station detection device in use is greatly improved.

The invention discloses a pipeline turning angle detection method, which comprises the following steps of: respectively acquiring three-axis acceleration, a three-axis angular rate and an axial speedby utilizing a three-axis accelerometer, a three-axis gyroscope and a multi-mileage instrument through the movement of a detection device in a pipeline; meanwhile, processing the acquired measurementvalues by means of a data processing unit and storing the measurement values in a data storage unit; after the detection is completed, combining output information of the three-axis accelerometer, thethree-axis gyroscope and the multi-mileage instrument under the offline condition, and calculating posture, speed and position information of a small-diameter pipeline detection robot moving in the pipeline by adopting a strapdown inertial navigation algorithm under the condition that the initial posture, speed and position of the detection device before detection are known; and correcting the speed information of the detection device moving in the pipeline according to the measurement information of the multi-mileage instrument, so that the positioning precision is improved. The detection result of the pipeline turning angle detection method can be suitable for improving the pipeline detection and positioning precision in environments such as urban tall buildings, viaducts, tunnels and indoor environments which cannot be positioned by GPS.

The invention relates to a cigarette automatic detection method based on deep learning in a monitoring scene, and the method comprises the steps: firstly carrying out the overturning, zooming and smoothing of a cigarette data set photographed and downloaded from a network, obtaining a bigger data set, training the data set based on a YOLOv3 deep learning network, and forming a template library; performing corresponding image enhancement processing on the to-be-detected image or video frame by using an image enhancement method; carrying out image segmentation on the large-size image, cigarettesor people with cigarettes in the image are separated out, and the time needed by detection is shortened; then pre-generating prediction boxes on the image to be detected, and comparing each prediction box with a pre-trained template library; and finally, selecting a prediction box higher than a preset threshold value from the detection confidence coefficients of all the prediction boxes, and determining the prediction box as a target object. After the whole image of the current frame is scanned, all detected targets are marked and displayed on the image, and cigarette detection is completed.The method can effectively improve the detection accuracy and shorten the detection time.

Embodiments of the invention provide a living body detection method and device and electronic equipment. The method comprises the steps of extracting spatial features and time features of multiple frames of to-be-detected face images; inputting the spatial features and the time features into a first convolutional neural network model, and obtaining low-dimensional spatial features and low-dimensional time features output by a global pooling layer in the first convolutional neural network; inputting the multiple frames of face images into a second convolutional neural network model, and obtaining biological motion features output by a full connection layer in a second convolutional neural network; and calculating a living body detection result of the multiple frames of face images accordingto the low-dimensional spatial features, the low-dimensional time features and the biological motion features.

The invention discloses a high-sensitivity solid-state color image sensor. Particularly, a large-area MOS (Metal Oxide Semiconductor) capacitor is used as a photoelectric conversion region, the generated and collected photoelectrons are transferred to a channel region arranged below the grid electrode of an adjacent small-sized MOSFET (Metal-Oxide-Semiconductor Field Effect Transistor) storage, furthermore, the transferred photoelectrons are injected into a storage layer of the storage to enable the threshold voltage of the storage to change, and finally, the intensity of the detected light is determined through reading the drain current of the storage. Pixel units of the color image sensor are as follows: each pixel unit of the color image sensor is composed of three pixel subunits which have the same structures and are respectively red, green and blue, and each of the red, green and blue pixel units is respectively composed of one MOS capacitor and one MOSFET structural storage tube. A manufacturing process of the color image sensor is completely compatible with a CMOS (Complementary Metal-Oxide-Semiconductor) process; and the color image sensor is matured in process, low in cost and easy to achieve.

The invention relates to the technical field of character recognition, and discloses a handwritten text recognition method and device and electronic equipment. The handwritten text recognition methodcomprises the steps: obtaining a to-be-recognized handwritten text image; performing high-order feature extraction on the handwritten text image by using the convolutional layer of the convolutional neural network model obtained by training to obtain feature data; performing serialization processing on the feature data through a recursion block of the convolutional neural network model, and performing serial processing in the depth direction on the serialized data obtained by the serialization processing to obtain serial serialized data; and mapping the series serialized data to an output label through a linear layer of the convolutional neural network model to obtain an output value. According to the handwritten text recognition method, the target text corresponding to the handwritten text image is acquired according to the output value and the preset semantic library, and the target text corresponding to the handwritten text image is recognized through the convolutional neural network model, and the calculation cost can be greatly reduced on the premise that the detection precision is not affected, and the training speed and the recognition speed of the model are increased.

The invention discloses a vehicle hub comprehensive measuring instrument, and belongs to the technical field of vehicle hub metering detection. When the hub is installed on the vehicle, the end face,close to the vehicle, of the hub is a bottom surface, and the end face, facing the exterior of the vehicle, of the hub is a top surface. The vehicle hub comprehensive measuring instrument comprises arack, an electric cabinet is arranged at the bottom in the rack, a base is arranged at the top of the electric cabinet, a precise rotary table is rotationally connected to the base. A hub positioningtable is installed on the precise rotary table, a positioning pin is arranged on the hub positioning table, and a hub standard piece is arranged on the hub positioning table. A lifting mechanism is arranged at the inner top portion of the rack, and a detection mechanism is installed on the lifting mechanism. Practical use proves that the comprehensive measuring instrument can operate independentlyby one operator. The operation is simple and convenient, and the performance is reliable. Detection of one hub can be completed within 25s, and the requirement for manufacturing the detection beat ismet.

The invention discloses an insufficient-driving detection method for a screw, and belongs to the technical field of screw driving. The insufficient-driving detection method for the screw comprises thesteps of screw obtaining and screw driving, wherein during the screw obtaining, the tail end of a location component is abutted against a location plane, a screw bit is controlled to move in a firstdirection, the tail end of the screw bit is abutted against a screw head, and a first distance between the screw bit and the location component in the first direction at this moment is obtained; during the screw driving, the tail end of the location component is abutted against a driving plane, the screw bit is controlled to move in the first direction to complete the screw driving, and a second distance between the screw bit and the location component in the first direction when the driving is completed is obtained; and if the second distance is greater than the first distance, insufficient-driving of the screw is determined. Insufficient-driving detection can be carried out through the change of relative positions between the screw bit and the location component, the detection complexityis lowered, the detection time is shorter, and the production efficiency is increased; and the error of a product does not influence the detection accuracy, and the detection accuracy is higher.

The invention discloses a tensile sample for a clad sheet steel substrate. The tensile sample for the clad sheet steel substrate has a length of 400mm; the length of parallel segment is 80mm; the circular arc radius is 25mm; the gauge length of sample is 50mm. Compared with the prior art, the tensile sample for the clad sheet steel substrate fully meets national and international standards; no additional clamp is required when the tensile sample for the clad sheet steel substrate is detected; the tensile sample for the clad sheet steel substrate can be directly put into a conventional tensiletesting machine for detection; the operation is convenient and quick and the detection precision is not influenced; the tensile test detection can be quickly and accurately performed; the tensile sample for the clad sheet steel substrate is an ideal tensile sample for detecting the clad sheet steel substrate of pipeline steel and stainless steel.

The invention relates to an environment-friendly chemical waste gas detection device. The device comprises a temporary storage bottom box and a detection camera obscura; the temporary storage bottom box and the detection camera obscura are provided with a gas inlet device, a repeated conveying device, a monitoring device, a displacement device, a rotating device, a filtering device, a gas pushingdevice and an opening and closing device; one side of the temporary storage bottom box is fixedly connected with the detection camera obscura; the detection camera obscura is of a structure with openings in the two sides; the other side of the detection camera obscura is fixedly connected with supporting frames; a temporary storage top box is fixedly connected between the supporting frames; and the temporary storage top box is of a structure with openings in the two sides. Through cooperation of the gas inlet device, the repeated conveying device, the monitoring device, the displacement device, the rotating device, the filtering device, the gas pushing device and the opening and closing device, waste gas of all volumes can be detected in the detection and discharging process, the detectionprecision is improved, and the surrounding environment cannot be damaged.

A detection apparatus of long stroke leading screw installation precision and an application method thereof relate to the detection apparatus of a driving leading screw of a numerical control turret punch and a numerical control laser cutting machine workbench and the application method thereof. The detection apparatus of the long stroke leading screw installation precision and the application method thereof are provided. By using the apparatus and the method of the invention, under the condition of conforming to a natural law, detection data can be accurately obtained and the product installation precision can be increased. The leading screw is flexibly connected on a rack. A pair of Y direction guide rails which is parallel to the leading screw is arranged on the rack. A leading screw nut is connected on the leading screw. The detection apparatus comprises a portal frame, an X axis guide rail, an X axis slide block, a table 1 and a table 2. The application method of the invention comprises the following steps: 1) Installing the detection apparatus on the pair of the Y direction guide rails; 2) Moving the leading screw nut to a middle position of the leading screw; 3) Overcoming deflection; 4) Detecting one end; 5) Detecting the other end; 6) Finishing.

The invention discloses a software interface element detection and positioning method and system based on multi-granularity learning, computer equipment and a storage medium. The method comprises thesteps of: obtaining an operation interface image; inputting the image into a deep convolutional neural network to extract features to obtain the intermediate result of the same size; performing meshing division on the intermediate result; inputting the intermediate result into a multi-granularity cascade background discriminator, and carrying out background grid detection on the intermediate imageby a detector by taking a grid as a unit; after background grids are removed, partially inputting the remaining grids into a multi-granularity software interface element detection model for softwareinterface element detection, and outputting the type and the area size of a control to be detected. According to the multi-granularity cascaded background grid detection device provided by the invention, the background grid can be quickly and effectively found out, and the efficiency and accuracy of software interface element detection and positioning are improved.

The invention relates to detection equipment, in particular to ceramic tile quality automatic detection equipment. The technical problem to be solved is how to design the ceramic tile quality automatic detection equipment which can replace manpower to detect the quality of ceramic tiles, is labor-saving and does not influence the detection accuracy. The tile quality automatic detection equipment comprises a base, wherein four vertical plates are fixedly connected to one side of the base at intervals; and first small supports which are symmetrically and fixedly connected to the middle of the side, away from the vertical plates, of the base. Ceramic tiles are placed on a left belt, a gear motor is started to rotate forwards, the left belt can drive the ceramic tiles to move rightwards, a first sliding rod drives the ceramic tiles to move upwards to make contact with a suction cup to be sucked and fixed, and when the suction cup loosens the ceramic tiles, the ceramic tiles fall onto the left belt to be subjected to quality detection. Thus, a rubber hammer does not need to be continuously waved by people to beat the ceramic tiles, the labor is saved, and the detection accuracy is not influenced.

The invention discloses a bridge bolt detection method based on a self-attention and central point regression model. The method comprises the steps: acquiring a to-be-detected bridge bolt image; detecting the bridge bolt image through a pre-trained model based on self-attention and central point regression; and according to a detection result of the self-attention and central point regression model, determining whether the bridge bolt image bolt has a disease or not, wherein the model based on self-attention and central point regression comprises a semantic segmentation network and a detection network at the same time, the two networks share middle-layer features extracted by the convolutional neural network, the semantic segmentation network aims to perform semantic segmentation on a rectangular area where the bolt is located, and features for semantic segmentation and features of the detection module are connected to jointly position the bolt. The method is applied to the field of bolt disease detection and recognition, can overcome the defects of a traditional bridge disease image detection and recognition technology, and can well solve the problems of efficiency, cost, safety and the like in bolt disease detection and recognition.

A B-mode ultrasound probe with a puncture function comprises a supporting frame. The upper end of the supporting frame is connected with a handle, and the lower end of the supporting frame is connected with an acoustic lens with the center being provided with a through hole. A guiding groove penetrating through the handle and the supporting frame is arranged in the middle of the handle. Containingcavities are formed in the positions, corresponding to the two sides of the guiding groove, in the handle, and a one-way separation part is arranged on the position, corresponding to the upper portion of the through hole, of the lower end of the guiding groove. A guiding pipe is arranged in the guiding groove, and the two sides of the guiding pipe are connected with puncture frames arranged in the containing cavities. Adjusting parts used for pushing the puncture frames to move are arranged on the upper sides of the puncture frames. The function that puncture can be conducted without manual adjusting of the angles of the puncture frames while an infection focus is detected is achieved, positioning is accurate and rapid, the puncture function is achieved while the detecting precision is ensured, the structure is simple, and using is convenient.

The invention relates to detection devices and particularly relates to a composite insulator hydrophobicity detection device. The detection device includes a detection device frame, an air ventilator, an insulator, and an insulator base. A moving door, a water pump, a water tank and the air ventilator are arranged on the detection device frame. The insulator base is arranged inside the detection device frame and is connected to the insulator. The detection device is convenient to install and has low cost, which is much lower than that of similar products on market, due to the mechanical structures. For different types of the insulators, corresponding structural heights of the device are different, wherein the optimized structural height of the device can reach 100 cm, so that the detection device can adapt a wide range of insulators and can detect composite insulators in various electric systems. The detection device can standardize detection environment and greatly improves detection accuracy. In the image processing field, non-uniformity of illumination is a common difficult issue; the detection device can improve illumination environment for image processing to certain degree by using an annular light source and controlling illumination distance, thus reducing influence due to the non-uniformity of illumination.

The invention discloses a phased array detection method for a fillet weld of a pipe base. The method comprises the following steps: drawing weld graphic expressions corresponding to different angles; connecting a probe with a phased array ultrasonic test instrument; converting an ultrasonic beam into a scanning image by the phased array ultrasonic test instrument; introducing the weld graphic expressions corresponding to different angles into the phased array ultrasonic test instrument, converting the weld graphic expressions to a corresponding angle following the probe once moving the probe for a certain angle while detecting and then correspondingly matching the introduced graph with a practical S scanned image, thereby guaranteeing one-to-one correspondence of the detected signal wave and the weld graphic expressions; if the position of a generated echo is overlapped with a workpiece loading graph, confirming the echo as the echo generated at a workpiece boundary; if not, confirming the echo as the echo generated by defect. The detection data can be acquired after the probe is adhered to the workpiece and is used for scanning from a starting point to a destination, the detection efficiency is high and the time cost is saved. The defects of the fillet weld of the pipe base can be detected and the omission ratio is lower.

The invention provides a high-side current sampling voltage expansion circuit which comprises a sampling resistor, a high-side current detection chip, a voltage dropping resistor, an MOS tube, a voltage stabilizing diode and a current limiting resistor, the two ends of the sampling resistor are connected with the input end of the high-side current detection chip, the sampling resistor is arranged on the high side, namely the positive electrode, of an output loop, the current is output to a load through the sampling resistor; the output end of the high-side current detection chip is connected to one end of the dropping resistor, and the other end of the dropping resistor is connected with the source electrode of the MOS tube; the grid electrode of the MOS tube is connected with one end of the current-limiting resistor, one end of the current-limiting resistor is further connected with the anode of the voltage-stabilizing diode, and the cathode of the voltage-stabilizing diode is connected with the anode of the circuit; the other end of the current-limiting resistor is grounded; and the drain electrode is used as a current output end.

The invention provides a small-diameter pipeline robot high-precision positioning method based on a turning angle. The small-diameter pipeline robot high-precision positioning method based on the turning angle mainly comprises the following steps of S11, detecting a pipeline by using a trolley with an inertial measurement unit and a plurality of odometers; S12, calculating the turning angle of the pipeline by using inertial measurement unit data and odometer data respectively; S13, performing position, speed and attitude angle error calculation on the turning angle detection information in combination with odometer detection information and non-integrity constraints in the pipeline; S14, calculating three-dimensional position information according to an initial position; correcting the inertial navigation calculation position by means of the length information of a straight pipeline section with the known length; S15, performing error compensation by adopting a Kalman filtering estimation method; S16, smoothing the data; and S17, determining the defect position of the pipeline. According to the small-diameter pipeline robot high-precision positioning method based on the turning angle provided by the invention, the problem that the overall positioning precision is influenced because the error is not considered from the overall perspective is solved.

The invention relates to an online fruit detection device and a use method thereof. The online fruit detection device comprises a spectrograph mechanism, an automatic control mechanism and a driving mechanism, wherein the automatic control mechanism comprises a controller, an image acquisition unit, an image processing unit, an image storage unit and a driving mechanism control unit, the driving mechanism comprises a fruit cup, a transmission mechanism and a rotary driving piece and / or a fruit unloading driving piece, the controller controls the driving mechanism control unit to make an instruction of opening and closing the driving mechanism according to a comparison result and / or a judgment result, and the driving mechanism control unit controls the driving mechanism to rotate or not rotate fruits on the fruit cup according to the received instruction, and to unload or not unload the fruits. The device has the characteristics of reasonable structural design, convenience in operationand use and high detection accuracy, and can ensure that the spectrograph can accurately acquire fruit spectral data in the fruit detection process by combining image detection and mechanical motion control, so that the detection accuracy is greatly improved.