144results about How to "Avoid loss of accuracy" patented technology

A CT image pulmonary nodule detection method based on a 3D residual convolutional neural network includes a training process and a testing process. The training process includes the following steps: S1, preprocessing an original image, resetting the voxel spacing to (1, 1, 1), and converting the voxel spacing into voxel coordinates; S2, capturing 3D positive and negative samples from a CT image; S3, setting a maximum and a minimum, and standardizing the sample data; S4, constructing a 3D convolutional neural network; S6, setting training hyper-parameters, and importing the training hyper-parameters to a data training model in the form of mini-batch; and S6, saving the model after the model is fully trained. The testing process includes a step S7: preprocessing test CTs, sampling the test CTs one by one in the form of sliders, importing the test CTs to the model for calculation, selecting samples with high confidence, and deleting repeated samples through a non-maximum suppression algorithm. The method is of high accuracy, and can be used to analyze whether there is a nodule in an image and the specific position of the nodule in the image.

The invention discloses a method and system for unlocking a touch screen of a mobile terminal. The method includes the steps that (A) the mobile terminal receives a user operation instruction, the screen is lightened, an unlocking interface is displayed, and a prestored image containing an unlocking pattern is called randomly and displayed on the unlocking interface; (B) a pattern drawing operation instruction of a user on the unlocking interface is received, a pattern drawn by the user is obtained, and similarity comparison is carried out on the pattern drawn by the user and the displayed unlocking pattern to obtain a shape similarity value; (C) the obtained shape similarity value is compared with a preset threshold value, and if the shape similarity value is larger than or equal to the threshold value, the touch screen is unlocked. The unlocking method is achieved in a mode that the user draws patterns displayed randomly, the unlocking processes are different, freshness and interestingness of unlocking are increased on the premise of simple unlocking operation, the problem that precision of a sensor is reduced due to the fact that an unlocking area is fixed in the existing unlocking technology is avoided at the same time, the service life of the touch screen is prolonged, and product experience of the user is improved.

The invention discloses a touch screen unlocking method and system for a mobile terminal. The method includes the steps that (A) a user instruction is received, a screen lights, a screen locking interface is displayed, it is detected that a user touches a touch screen, and a sliding track of the user on the touch screen is obtained; (B) touch points which the sliding track covers are analyzed, and the area of each triangle composed of any three touch points is calculated by means of the analyzed touching points; (C) the largest area of the triangle is obtained to be compared with a preset threshold, and when the obtained area of the triangle is larger than the threshold, the touch screen is unlocked. The unlocking method is achieved by judging the area of the touching points, so that an unlocking interface is in full use, and the problem that precision of a sensor is lowered in the existing unlocking process due to the fact that an unlocking area is fixed is solved. Meanwhile, the user can automatically choose the sliding track in each sliding unlocking process of the mobile terminal, and operation is easy, convenient and quick to achieve.

The invention belongs to the intelligent vehicle environment perception technology field and especially relates to a calibration device and a method of a multi-line laser radar. The device comprises acalibration plate, a calibration rod, a rotating disk, a spring lock mechanism, an end cap, a bearing, a ball and a pedestal. The calibration plate is clamped on an upper end of the calibration rod.A lower end of the calibration rod is connected to the rotating disk in a threaded connection mode. The rotating disk is clamped in a ring groove of the pedestal and is fixedly arranged with the pedestal through the spring lock mechanism. The bearing is arranged between the pedestal and a bearing hole of the rotating disk and forms interference fit with the rotating disk. The end cap is arranged on the rotating disk and is fixedly arranged at a center shaft of the pedestal through a bolt. A pad is arranged between the bolt and the end cap. The invention discloses the convenient and high-efficient calibration device and the method aiming at a rotation-type 64-line laser radar. Three-dimensional coordinates of scanning points can be simultaneously and accurately calibrated. By using an existing laser radar, an actual position coordinate of each laser scanning point is difficult to be determined, a scanning track shape is discrete and complex and a calibration process is tedious. In the invention, the above problems are solved.

The embodiment of the invention provides a machine learning model obtaining method and obtaining device, apparatus, and storage medium, wherein the method comprises the following steps: obtaining a first sample picture labeled with a category; using the first sample picture labeled with category as a training sample, and obtaining the machine learning model by training; inputting a second sample picture to a machine learning model to obtain a category of the second sample picture predicted by the machine learning model; determining a second sample picture having a correct category predicted bythe machine learning model; obtaining a new machine learning model by using the second sample image with correct classification as the training sample. The obtaining method of the machine learning model provided by the embodiment of the invention can improve the prediction accuracy of the machine learning model, that is, improve the performance of the machine learning model.

The invention provides a refrigerator compartment temperature control method, which comprises the following steps that: the temperature of a refrigerator compartment is acquired at a predetermined time interval for N times to obtain N temperature sample values; the N temperature sample values are summed to obtain a first temperature value, an average value of the N temperature sample values is calculated and the average value is rounded to obtain a second temperature value; the first temperature value is complemented to obtain a residue of the first temperature value, the residue is compared with N / 2, and a target temperature value is obtained according to the comparison result and the second temperature value; and the compartment temperature is controlled according to the target temperature value. According to the acquisition method provided by the invention, the acquired compartment temperature is more accurate, and further the control of the refrigerator for the compartment temperature is improved, so that the fluctuation of the compartment temperature is stabilized. The invention also provides a refrigerator compartment temperature control device, and a refrigerator with the same.

The invention relates to the technical field of packaging and particularly discloses a Luolong radix codonopsis packaging device. The Luolong radix codonopsis packaging device comprises a blanking hopper, wherein a discharging opening is formed in the bottom of the blanking hopper; a material receiving tank is arranged right under the discharging opening; an iron support plate is arranged under the material receiving tank; the material receiving tank is located on the support plate; a transverse wedge block is arranged on the right side of the material receiving tank; the material receiving tank is abutted against the wedge surface of the wedge block; the wedge block is connected with the support plate in a sliding manner; the bottom of the right side of the material receiving tank is articulated on the support plate; a blanking opening is formed in one side away from the wedge block of the material receiving tank and is articulated with a baffle; a support frame is arranged under thesupport plate; an elastic part is connected between the support plate and the support plate; a magnetic part is arranged under the support plate and is welded on the support frame; a quantification gear is arranged on the left side of the support plate and is meshed with a quantification rack; the quantification rack is welded on the left part of the support plate. Through application of the technical scheme, the Luolong radix codonopsis packaging device is capable of supplying objects with fixed weight, automatically and intermittently conveying packaging boxes, reducing the time and improving the operating efficiency.

The invention discloses a novel range-adjustable magnetic liquid acceleration sensor, and belongs to the technical field of inertial sensors. Current magnetic liquid acceleration sensors are short in range and large in size, and also have the problem that for long-term use, the accuracy of sensor is reduced due to that restoring forces are reduced because of demagnetization of permanent magnets. The sensor is composed of an air tank, an air duct, a valve, a pressure gauge, an end cap, a shell, an inductive coil, a cylindrical permanent magnet, a rubber seal ring, a bolt / nut connecting assembly, air, magnetic liquid and a signal circuit, wherein the cylindrical permanent magnet and the magnetic liquid form inertial mass. While enabling the permanent magnet to float in the inner cavity of the shell, the magnetic liquid also can seal the air in the cavities at two ends in the inner cavity of the shell. When moving axially, the inertial mass compresses the air in the cavity at one end to generate the restoring force. The range of the sensor can be adjusted by changing the restoring force through adjusting air pressure in the cavities at two ends in the inner cavity of the shell. The sensor has adjustable range, small size and long service life.

The invention discloses a portable type ozone concentration online monitoring device and an ozone concentration online method. The device comprises a housing, wherein a gas exhausting device, a filter system, a gas absorption tank and a data processing system are arranged in the housing; a data display module and a starting button for starting the device are arranged on a panel of the housing; the filter system, the gas exhausting system and the gas absorption tank are communicated through a sealed PTFE tube; the gas absorption tank and the data processing system are connected through an optical fiber. The portable type ozone concentration online monitoring device has the characteristics of being simple in structure, high in integration, convenient to operate, high in measuring precision, high in measuring speed and capable of performing online monitoring in real time, not only can be suitable for monitoring ozone concentration in a transformer station environment, but also can be used for indirectly judging an operation state of electrically charged equipment in a transformer station.

The disclosure provides a target tracking method and apparatus based on human-machine interaction, and a storage medium. According to the method, a to-be-tracked target can be determined through a triggering instruction generated based on a target selection operation of a user without operation learning of multiple continuous frames of images, the acquisition operation of the target is reduced, and the target acquisition efficiency and the tracking efficiency can be improved; the dimension and the scale of a target image in a next frame can be detected based on extracted image characteristicsso that the tracking effect of a rigid target, a target with scale transformation, a target with high-speed motion, and a target which is shielded for a long time can be optimized to a certain extent,and the precision of target tracking is improved; and the effect of the operation efficiency and the target tracking can be guaranteed on general portable devices.

The invention discloses a congestion pre-judging and traveling route planning method as well as a traveling route planning system. The congestion pre-judging method is based on a concept of sequencing transportation means; and the traveling route planning method and the traveling route planning system are formed in conjunction with the congestion pre-judging method on the basis of automatically dividing the traveling route. By adopting the method and the system, the problem that the precision is lowered due to error accumulation can be effectively avoided.

The embodiment of the invention discloses an image classification method and device and a style migration model training method and device, and relates to the field of deep learning, cloud computing and computer vision in artificial intelligence. The specific implementation scheme comprises the steps of inputting a first style image into a style migration model to obtain a second style image corresponding to the first style image; and inputting the second style image into the image classification model to obtain a classification result of the second style image, wherein the style migration model is obtained by training based on a sample image of a first style and a sample image of a second style, and the image classification model is obtained based on second style sample image training. According to the embodiment of the invention, the image classification accuracy can be improved. The embodiment of the invention can be applied to fundus screening.

The invention discloses a microbeam plasma arc additive manufacturing device and method for aluminum alloy wires. A forming base of the device is installed on a horizontal moving mechanism. The horizontal moving mechanism is carried by the forming base and reciprocates horizontally in the Y-axis direction. A plasma gun driving mechanism comprises Z-axis direction motion controllers, an X-axis direction motion controller, Z-axis motion guide rails and an X-axis motion guide rail. The X-axis direction motion controller is installed on the X-axis motion guide rail. The two ends of the X-axis motion guide rail are installed on the two Z-axis direction motion controllers correspondingly. A movable forming platform is adopted, the complexity of the structure of the plasma welding gun driving system is lowered, the problem of precision decline caused by long-distance and long-span motion of a welding gun is avoided, and meanwhile, the forming efficiency is improved; and the device is simple in overall structure, and microbeam plasma arc additive manufacturing for large-size aluminum alloy wire components of complex structures can be achieved.

The invention provides an automatic calibrating method for a hand-eye relationship of a robot based on incremental compensation. The automatic calibrating method comprises the steps of guiding the tail end of a mechanical arm to a proper calibrating origin, collecting an image of a calibrated object on a working table board, and extracting pixel coordinates of a reference point therein; presettingTCP offset of each data collecting point relative to the calibration origin; controlling the mechanical arm to arrive each data collecting point along a TCP coordinate system according to the presetoffset to collect the pixel coordinates of the reference point; and calculating calibrating parameters according the offset of collected pixel coordinates relative to the pixel coordinates at the calibrating origin as well as the TCP offset, relative to the calibrating origin, corresponding to the pixel coordinates. The automatic calibrating method is high in calibrating precision, is implementedautomatically, is convenient and rapid and high in calibrating efficiency, and can be applied to the mechanical arm with a fixed base or an unfixed base.

The invention discloses a method for calculating the limit speed of a double NURBS path in free-form surface processing. The method comprises the following steps: (1) calculating a control vertex anda node vector of an NURBS curve according to data points of a known discrete tool nose file, defining a fitting curve of a tool nose point track, calculating data points (cutter axis points for short)participating in cutting in the cutter axis direction through a linear relation, and defining a fitting curve of a track of the point by using the same method, so as to generate the double NURBS path; and (2) solving the curve length of an interpolation road section on a definition domain by integrating a first-order guide vector function of a tool nose point NURBS fitting curve, traversing the whole processing path, and calculating a key point of which the curvature radius reaches a local extreme value under the limitation of geometric error constraint and kinematics constraint so as to obtain the limit speed at the key point.

The invention relates to a community structure discovery method and system of a bank customer transaction network. The method comprises the following steps: S1, extracting customer transaction data toconstruct a bank customer transaction network; s2, dividing a local community for the local sub-network of each account node; s3, determining a community label propagation sequence of the account nodes; s4, calculating the community determinacy of each account node; s5, initializing community labels of all account nodes; s6, updating the community labels of all the account nodes according to thecommunity label propagation sequence; and S7, checking whether the community label of each account node is changed or not, if so, carrying out the next round of iterative updating, and if not, dividing the community label into the same community. According to the invention, the customer transaction network is constructed based on the association relationship contained in the customer transaction data, and the community structure in the customer transaction network is found through identification, so that the problem that the economic activity rule of the customer is difficult to effectively reflect due to attribute loss of the traditional transaction analysis method depending on the customer attribute is avoided.

The invention discloses an automobile clutch assembling tool which comprises a base and a rotary disc. An arc-shaped sliding groove is formed in the top of the rotary disc, a push rod is movably installed in a strip-shaped groove, a limiting block is fixedly installed at the bottom of the push rod, a clamping block is movably installed at the top of the push rod, and a telescopic rod is fixedly installed at the bottom of a lifting disc; and a first cylinder is fixedly mounted at the bottom of the telescopic rod, a fixed disc sleeves the bottom of the rotary disc in a connected manner, second cylinders are fixedly mounted at the two ends of the bottom of the base, first fixed teeth are fixedly mounted at one side of the rotary disc, and a locking device is fixedly mounted at one end of thetop of the base. A driving disc, an outer tooth friction plate, an inner tooth plate, a clutch spring and a pressing disc are sequentially placed at the top of a tray, a tool capable of abutting against the pressing disc is placed at the top of the pressing disc, a second air cylinder is used for enabling the tray to move upwards, the pressing disc abuts against the bottom of the tool, and therefore compression of the clutch spring is achieved; and the first cylinder jacks a clutch, so that the driven disc is arranged on the clutch in a sleeving manner, and assembly of the clutch is completed.

The invention provides a novel soil sampling analyzer. The novel soil sampling analyzer comprises a rotating handle, a drill pipe and a drill bit, wherein the lower end of the drill pipe is connected with the drill bit, a motor is arranged at the lower end of the rotating handle, a display is connected at the upper end of the rotating handle, a screw rod is connected to the lower end of the motor, the drill pipe is connected to the lower end of the screw rod, two baffle plates are symmetrically arranged on the inner wall of the drill pipe, a soil analyzer is arranged at the top end of the inner part of the drill pipe, two control knobs are arranged at the upper end of the drill pipe, an opening-closing device is arranged at the top end of the drill bit, an induction detection head is connected to the upper part of the opening-closing device, and the motor is connected with the display, the screw rod, the soil analyzer, the control knobs and the induction detection head through wires. The novel soil sampling analyzer disclosed by the invention has the beneficial effects that while sampling soil, the soil is analyzed preliminarily and soil related information of acidity-alkalinity, constituents, moisture and humidity and the like in the soil are detected in the first time to avoid lowering accuracy of the sampled soil conveyed to a laboratory to be detected.

The invention discloses a method for online building of a state variable model of an aircraft engine based on a precise partial derivative. The method comprises the following steps of step A, determining an expression form of the state variable model of the aircraft engine; step B, acquiring a uniform discrete state variable form capable of expressing steady-state points and dynamic points based on Taylor expansion according to the expression form of the state variable model; step C, determining intermediate variables in a partial derivative computation process according to variables in the discrete state model needing to be established, and adopting an analysis method to establish a corresponding partial derivative model on the basis of aircraft engine component-level models and a chain type derivative rule; and step D, combining the component-level models and the partial derivative model to carry out joint computation through a multiple-pass algorithm, acquiring a state variable model coefficient matrix and an initial value of a current working point, and acquiring a state variable model of the corresponding working point. According to the method, the accurate state variable model can be computed on line at any working point.

The invention relates to the machining field of spare parts, in particular to a method for improving nitriding bevel gear wheel part machining precision. The method is characterized by comprising the following steps that 1, rough turning is conducted on a workblank; 2, heat treatment hardening and tempering are conducted; 3, semi-finish turning is conducted on gear blanks; 4, auxiliary procedures are conducted; 5, stress relieving is conducted; 6, gear blank fine machining is conducted; 7, a thickness side face of a gear is subjected to plane grinding; 8, a spherical surface is turned; 9, fine grinding is conducted on an addendum circle; 10, a top circular cone is turned; 11, a back circular cone is turned; 12, fine machining is conducted on the gear; 13, examination is conducted; 14, gear surface nitrogen treatment is conducted; 15, examination is conducted. The method for improving the nitriding bevel gear wheel part machining precision can improve the nitriding effect under the premise that the bevel gear wheel precision is guaranteed.

The invention discloses a step-by-step punching system, comprising an upper die, an upper die driving device, a lower die and a lower die driving device, wherein the lower die and the lower die driving device are installed on a guide rail, the upper die driving device and the lower die driving device are both connected with an electronic control module; and corresponding step pitch and step number are set for the electronic control module according to the punching requirement of a workpiece to be punched. According to the invention, an electronic control manner is adopted to substitute a common pure mechanical design mode, and step-by-step punching effect is realized, thus die structure is simple, the various problem of precision reduction caused by the complicated die structure can be effectively avoided, and most of high die development and manufacturing costs are reduced.

The invention discloses a double-laser head dust detection method. Two laser heads are employed alternately to emit laser to one detection area, the laser scattered after being emitted to dust is captured through a photoelectric receiver, and the laser passing through the detection area is reflected to other directions except for the detection area through reflecting surfaces. The invention also discloses a double-laser head dust detection sensor, the sensor includes a box internally equipped with the corresponding laser heads and the photoelectric receiver, the number of the laser heads is two, and the number of the photoelectric receiver is one, the box is internally provided with a detection area, and the photoelectric receiver is located on one side of the detection area. The box is internally provided with a circuit board connected to the laser heads and the photoelectric receiver, and is also equipped with the reflecting surfaces. According to the double-laser head dust detection method and sensor provided by the invention, the two laser heads are used alternately, the service life of the laser head combination can reach 30 thousand to 40 thousand hours, and the lifetime is greatly improved.

The invention provides a method for positioning and orientation of a tunneller. The method comprises a navigation position calculation process for acquiring real-time speed and position information ofa tunneller and an inertia calculation process for acquiring real-time attitude information of the tunneller. During the inertia calculation process, an attitude quaternion is introduced to calculatereal-time attitude information of the tunneller and is obtained by calculating corresponding speeds and intermediate speeds on all direction axes of a navigation coordinate system, wherein the navigation coordinate system includes an x axis, a y axis, and a z axis. With the provided method, the influence of the working environment on the positioning and orientation of a tunneller canb e eliminated; full autonomous navigation can be realized; the trouble of setting the rear tunnel reference manually can be saved; and the construction efficiency can be improved substantially. In addition, the invention also provides an apparatus for the method for positioning and orientation of a tunneller. The apparatus has a simple structure; and a problem of low positioning accuracy of the tunneller dueto the working environment in the prior art is solved.

The invention provides an energy-saving and environmentally friendly high-accuracy pile driver for building construction. The energy-saving and environmentally friendly high-accuracy pile driver for building construction comprises a ground pile, a hammer dropping hammer head, a connecting hinge, a surrounding cylinder, an accuracy control mechanism and a noise reduction and dust removal mechanism,wherein the ground pile is used for being driven underground; the hammer dropping hammer head is used for beating the ground pile to enable the ground pile to penetrate underground; the connecting hinge is used for connecting the hammer dropping hammer head with a hoisting device; the surrounding cylinder is used for surrounding the pile driving area; the accuracy control mechanism is used for limiting the ground pile and is convenient for high-accuracy driving of the ground pile; and the noise reduction and dust removal mechanism is used for reducing noises and removing dust. The invention relates to the field of buildings. According to the energy-saving and environmentally friendly high-accuracy pile driver for building construction, by analyzing the pile driving technology, the designis simple, the force of the hammer head can be more sufficiently developed to the ground pile by means of outside force, and therefore, the efficiency is greatly enhanced, more energy is saved, the noise can be reduced, the dust can also be removed, and the energy-saving and environmentally friendly high-accuracy pile driver for building construction is more environmentally friendly; and therefore, the problems that a common pile driver is hard to get outside force to realize a supporting effect during using due to large impact force of the hammer head and the ground pile is easy to be inclined to cause pile driving failure due to different underground soil force bearing capabilities are effectively solved.

The invention discloses a steel rail detection method for railway construction. The method comprises the steps that S1, detection preparation is conducted, specifically, a detection vehicle is placed on a steel rail, a detection block is connected with the steel rail in a clamped mode, a contact block is connected to the top face of the steel rail in an overlapped mode, and a first rotating rod and a second rotating rod are located on a same straight line, so that follow-up detection is facilitated; and S2, primiary detection is conducted, specifically, balls slide along the top surface of the steel rail, when the balls meet a deformation position of the steel rail in the vertical direction, the position of a liquid level plug in a liquid level pipe is changed, and the deformation degree and position of the steel rail are known. According to the invention, the deformation degree of the steel rail is determined by a deformation detection assembly, the internal damage condition is detected by a flaw detection assembly, the detection range is narrowed, the detection precision is improved, deformation and internal flaw detection are carried out in a centralized manner, the operation steps are simplified, and a pretreatment auxiliary assembly can play an auxiliary role in the detection of the deformation detection assembly and before the detection of the flaw detection assembly, so that the detection precision and efficiency are further improved.

The invention provides a method for identifying insulator defects by pruning YOLOv3 with a small on-site sample size. The method comprises the steps: 1, transmitting picture data in real time according to the inspection of an unmanned plane, and obtaining an insulator image and corresponding XML label data; 2, processing the insulator image; 3, performing pruning modification operation by taking YOLOv3 as a basic framework, and then establishing a deep learning model; 4, introducing an SPPnet module composed of four parallel maxpool layers into the pruned YOLOv3, and enabling the SPPnet moduleand the pruned YOLOv3 to be subjected to pruning; 5, converting the processed insulator image and the modified TXT label data into training data; 6, obtaining an insulator recognition model; 7, obtaining insulator picture frame information detected in each picture; and 8, judging the fault of the insulator. The method has the advantages that under the condition that the average precision is not reduced, the training time is shortened, and the requirement for the picture quality is reduced, meanwhile, the requirement for the computing power of a vehicle-mounted server is reduced, and the intelligent level of insulator inspection is improved.

The invention relates to a high-precision sea wave characteristic analysis method; a 9-SPS sensor is adopted for data acquisition, and the sensor is in contact with the sea waves, so that the acceleration of the sensor can represent the acceleration of a tested object. Nine piezoelectric ceramics are connected to the input ends of all channels of a charge amplifier through wires respectively, andsignals output by the charge amplifier are collected through a data acquisition card after impedance conversion and signal conditioning and amplification are carried out. In an industrial personal computer, matlab development software and a decoupling algorithm are utilized to design software, the deformation amount of the piezoelectric ceramics is calculated according to the amount of electric charges generated by the piezoelectric ceramics in the industrial personal computer, and finally the six-dimensional acceleration information of the tested object is obtained. After the acceleration information is obtained, further integral processing is carried out on the industrial personal computer, and the wave characteristics such as wave height, wave period and the like are calculated througha cross-zero method. A user can implement remote operation on the industrial personal computer through TeamViewer software on a PC.

The invention provides a pickup device comprising a proximity pickup module, a microphone array pickup module and an output control module, wherein the proximity pickup module at least comprises a proximity microphone, and the microphone array pickup module at least comprises a microphone array. The output control module is used for controlling to output a sound signal picked up by the pickup module and / or the sound signal picked up by the microphone array pickup module according to the distance between a sound source and the pickup device. Based on the function of the output control module, the pickup device can output the sound signal output from the microphone array to a voice transfer device in the case of a midfield sound source, thereby avoiding the problem that the voice transfer accuracy is reduced due to the far distance between a speaker and the proximity microphone. Moreover, the pickup module can also output the sound information output by the proximity pickup module to a sound amplifying device, thereby avoiding the echo interference problem caused by using the sound signal output by the microphone array as a sound amplifying source while ensuring the voice transfer accuracy.

The invention provides a transporting system and a transporting method, and belongs to the technical field of Internet. The system comprises a transporting robot and a scheduling server, the transporting robot comprises a transporting main body, a control module, a positioning sensor and a temperature regulation module, the positioning sensor is used for determining the current first position of the transporting robot and sending the first position to the scheduling server, the control module is used for receiving a transporting instruction sent by the scheduling server and controlling the transporting robot to transport goods transporting, the temperature regulation module comprises a temperature sensor and a temperature rising device, the temperature sensor is used for measuring the temperature of the positioning sensor, and the temperature rising device is used for improving the temperature of the positioning sensor. According to the transporting system, by adding the temperature regulation module onto the transporting robot, in the scene of cross temperature area transportation, the temperature of the positioning sensor is raised through the temperature rising device, water droplets of the positioning sensor due to the influence of temperature difference can be avoided, the accuracy of the positioning sensor is ensured, and the efficiency and safety of goods transporting are improved.

The invention discloses a double-accuracy weighted random forest algorithm based on particle swarm optimization, and the algorithm comprises the following steps: S1, determining an original data set D, randomly determining the number K of decision trees and the number m of features of the decision trees, and determining a pre-test sample rate X when m is smaller than or equal to n; s2, sampling the training data set Sk by using a Bootstrap sampling method to obtain an out-of-bag data Ok training sub Tk; s3, generating a kth decision tree according to a C4.5 algorithm, and calculating the finalweight of the decision tree; s4, repeating S2 and S3 until the number of decision trees is K; s5, using the decision tree set to test and divide the test data; and S6, taking the accuracy obtained inthe S5 as a fitness value of the particle swarm, performing iterative optimization by adopting a particle swarm optimization algorithm, comparing with historical classification accuracy, and finallyselecting an optimal model parameter. According to the invention, the accuracy of the random forest model is improved; Meanwhile, the problem that the algorithm accuracy is reduced due to parameter selection based on experience is avoided, and the performance of the algorithm is further improved.