TW406250B - Vehicle type recognition system based on Hidden Markov Model - Google Patents

Abstract

The invention provides a vehicle type recognition system based on Hidden Markov Model (HMM), which is mainly composed of a rangefinder, a vehicle appearance data pre-processing unit, a vehicle type learning unit and a vehicle type recognition unit; said rangefinder detects the object height on the carriageway continuously, then the vehicle appearance data pre-processing unit judges the object height detected by the rangefinder and determines the appearance feature of the vehicle body, so that the vehicle type learning unit trains HMM of various vehicle types in the unit according to the appearance feature of said vehicle, after adequate learning, the vehicle type recognition unit computes the outputs of HMM of various vehicle types according to the feature of vehicle appearance, and the vehicle type corresponds to the maximum output value of HMM is the type of the vehicle appeared on the carriageway.

Description

V. 10 15 20 406 250 Description of the invention (A7

[Technical Field] The present invention relates to a type of vehicle identification system based on hidden horses, and in particular to electronic charging of vehicles and other facilities based on two types of vehicle yards, and vehicle type identification based on smart ship parking mode. system. .¾Professional Marco [Previous technology] According to, general roads, bridges, parking lots, and other traffic ^ ~ = vehicles of different types charge tolls at different rates or use% ^ = The method of charging is electronic charging, what about electronic?疋 Receive in order to be correct for all types of vehicles: take the model of the local rate and confuse the low-rate vehicle outside the 'intelligent transportation system', and separate statistics for different models; flow rate or restricted models It is not allowed to enter a specific area. Therefore, in applications such as electronic toll collection and smart wheels, automatic vehicle identification is a must-have function. —Generally automatic vehicle model identification For vehicles with two speeds, it is very easy to cause identification errors due to excessive sampling differences, or even vehicle suspension due to traffic jams, which may lead to repeated sampling. These problems need to be overcome by the current technology. < It can be seen that there are still many shortcomings in the above-mentioned conventional articles, which are not a good designer 'and need to be improved. In view of the various shortcomings of the vehicle type identification system described above, the inventor of this case, “is eager to improve and innovate,” and after years of painstaking and meticulous research, he finally successfully developed this vehicle type identification system based on the hidden Markov model. . This tape measure £ applies

(Please read the notes on the back ^ 一 ^ write this page first)-(I II. -I I I. -------- Order ---------

406250 A7 B7 V. Description of the invention (^) PA880368.TWP-4/12 10 15 Employees of the Intellectual Property Bureau of the Ministry of Economic Affairs, printed by a cooperative 20 [Objective of the invention] The purpose of the present invention is to provide an electronic device that can respect the electronics automatically. Models of vehicles in tolling places are distinguished so that they can charge different fees according to vehicles of different models in order to achieve a vehicle model identification system based on the hidden Mayu model that automatically identifies the vehicle model's effectiveness. =: The second purpose is to provide a smart vehicle transportation system that can count vehicle traffic separately for different models and can limit access to specific areas based on a hidden Markov model of vehicle models. [Technical content] The vehicle type I based on the hidden Markov model with the above advantages = identification system: mainly consists of a rangefinder, vehicle appearance data pre-processing 2: it consists of a unit and a vehicle model identification unit; its Body ": The rangefinder above Γ keeps detecting objects in the lane, and the height of the measured object is transmitted every cycle. 外 = pre-processing unit of shape data, the front of the vehicle's shape data: according to the measurement The distance from the vehicle detected by H is continuous to the right of the f ... ° beam (the rear) 'and all the inter-values between the front and the rear of the vehicle are taken as the appearance characteristics of the auxiliary vehicle (Fe age). , The vehicle shape data will be transmitted to the vehicle model in the unit of the wheel shape, and the forest model learns these vehicle characteristics to train == the model's hidden Markov. Mode (_: Difficult) to get ㈣Marco hours;

Thread paper ruler — V. Description of the invention (々) 406250 A7 B7 PA83G368.TWP-5/12 After the training, the “vehicle model” printed by the employee's consumer cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs can be used as a hidden type of training. Car model for vehicle identification _ model identification; during identification, vehicle shape data is pre-processed, Yuan will send the collected vehicle shape features to the vehicle model identification unit, and the vehicle model identification unit will 5㈣ Calculate the output value of the hidden Markov mode for each type of vehicle. The model corresponding to the hidden Markov mode with the largest wheel value is the model of the vehicle appearing in the lane. [Brief description of the drawings] Please refer to the following detailed description of a preferred embodiment of the present invention and its accompanying drawings to further understand the technical content of the present invention and its purpose and effect. The drawings related to this embodiment are: One is a block diagram of a vehicle model recognition system based on the hidden Markov mode of the present invention; and FIG. One is a vehicle model recognition unit processing mechanism diagram of the vehicle model 15 recognition system of the vehicle based on the hidden Markov mode of the present invention. [Representative symbols of main parts] Vehicle appearance data pre-processing unit 4 Vehicle model identification unit 6 Identification stage At this time, the vehicle appearance characteristics are shown in FIG. 1, which is a vehicle model identification based on the hidden Markov mode provided by the present invention. The block diagram of the system mainly includes a -5-'Not applicable to the national standard (CNS) A4 specifications of the paper size χ 297 —-(Please read the precautions on the back of Vg first page) 1 rangefinder 3 vehicle model learning Unit 5 Learning Stage 7 Hidden Markov Mode

C ------- Order --------- Line — 406250 PA880368.TWP ~ 6/12

10 15 Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs 20 'Be! I distance device 丨, vehicle appearance information — ^ Vehicle model identification list Shen Shen, type learning unit distance 1§1 will continue to the lane * 4 directly above The term _ and the light or beam from the emission to turn two = 2 height ㈣,) the distance of the object, and the vehicle detected in each cycle, 2 objects = counting out = = = material_-heart car Out =; take all the height values between the front and the rear of the car to take '1 "judgment-the start of the car (head) _ = = characteristic vehicle height value from 6. Increase below the cm _ com = = So = start even; Judgment 1 The end of the car (tail) _ is = the high yield of the vehicle detected by Lai Shi seconds (miUise_d, ms): 2 at 60 cm, the shutdown-the car has ended; and 50 milliseconds can be converted into a rangefinder The detection cycle of the rangefinder, for example, the _ period of the rangefinder is 15 milliseconds' = the rule of the end of a car (the end of the car) is that the vehicle height value detected by the rangefinder for four consecutive cycles is below 60 cm ; At the learning stage 5 of the system, the vehicle shape f material pre-processing unit 2 will transmit the vehicle shape feature 8 to the vehicle model learning unit 4 for processing. The vehicle model learning sheet will cooperate with the system manager to prompt the correct vehicle type for each vehicle appearance feature 8 to train the hidden Markov model (HiddenMarkovM () del, coffee) corresponding to this type to obtain each hidden type. Parameters of the Mark ^ mode. Hidden Markov modes are widely used in computer speech recognition (# ^ Sadaoki Furui, Digital Speech Processing, Synthesis, and

This paper size applies to China National Standard (CNS) A4 (2G 297 mm)

IIIIII 406250 Α7 Β7 PA380368.TWP-7/17 Printed by the Consumer Cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs V. Invention Description (f) Recognition, Marcel Dekker, Inc., 1989. and Lawrence Rabiner and Biin-Hwang Juang, Fundamentals of Speech Recognition, Prentice-Hall International, Inc., 1993). The hidden Markov pattern is a two-layer random program. A hidden Markov pattern can be represented by the following 5 mathematical symbols: and = (Α, Β, 7Γ ), Where T_the number of height values in a certain vehicle appearance feature 8 (that is, the total number of detection cycles from the front to the rear of the vehicle); N: the number of states in the hidden Markov mode; M: Number of vehicle types; Q = {qi,% ... ",%-丨,%}: all states in hidden Markov mode; V = {V1, V2, .. ·····, &- !,%}: Various model codes; A = {aij}: a set of all state transition probabilities; airPr (qj at t + Ι | qi at t): the probability of transition from state qi to state ，, o < t < (T + i) B = {bj (k)}: Probability of appearance of all models; 1 ^) = ((\ ^ 1 吒 人): in When the state is 吒, the probability that the vehicle model is Vk, 〇 < t <(丁; π-{TTi): a set of the probability of all initial states; π i = Pr (qi at t = l): Chance of starting state is%. Therefore, the main work of the vehicle model learning unit 4 is the hidden Markov mode corresponding to the vehicle model. Turn enough of this type of car, and feature 8 is sequentially input to the hidden Markov mode to train the Markov mode. Α, Β, π and other three parameters; and the training method 10 15 20 The size of this paper @National Standard (CNS; Specifications (21〇x 297 male cage) (Please read the precautions on the back to write this page) Α Equipment ---- Order --------- line, 406250 A7 B7 PA880368.TWP-8/12 Printed jade and invention description (& invention) by the Consumer Cooperatives of Intellectual Property Bureau of the Ministry of Economic Affairs, Viterbi algorithm (Viterbi algorithm, see Sadaoki Furui, Digital Speech Processing, Synthesis, and Recognition, Marcel Dekker, Inc., 1989. and Lawrence Rabiner and Biin-Hwang Juang, Fundamentals of Speech Recognition, Prentice-Hall International, Inc., 1993 ) 5 is a commonly used method. After a sufficient learning stage, the vehicle model recognition unit 3 can use the trained hidden Markov model to do vehicle model recognition; at the recognition stage 6 The vehicle shape data pre-processing unit 2 transmits the collected vehicle shape features 8 to the vehicle model identification unit 3, and the vehicle model identification Zhuo Yuan 3 calculates 10 types of vehicles based on these vehicle shape parameters. The output value of the hidden Markov mode of the vehicle, the model corresponding to the hidden Markov mode with the largest output value is the vehicle model appearing in the lane; and the vehicle model learning unit and vehicle model identification unit are equipped for each type of vehicle It belongs to one or two or more hidden Markov modes, and can share a hidden Markov mode 15 for various types of vehicles. Please refer to FIG. 2 'It is a hidden Markov mode provided by the present invention. Based on the processing model of the vehicle model recognition unit of the vehicle model recognition system based on the vehicle, the vehicle model recognition unit 3's job is to input the vehicle appearance characteristics 8 to all the hidden Markov modes 7 and use the trained hidden Markov modes 7 A, B, ττ and other three parameters to calculate the final output value of each hidden Markov pattern 7. The output value is the probability that the vehicle's appearance characteristic 8 corresponds to various types of vehicles; the vehicle model is the model represented by the hidden Markov mode 7 with the largest output value. [Features and Effects] 20 fl ^ i > WR. · If 1 ϋ nn I · nn (Please read the precautions on the back page) Order ..

C-line 406250 A7 B7 PA880368.TWP-9/12 V. Description of the invention (γ) The vehicle type identification system based on the hidden Markov mode provided by the present invention, when compared with other conventional technologies, has the following Advantages: 1. The invention is difficult to identify errors for vehicles traveling at variable speeds. 5 2. The present invention makes it difficult to identify errors due to suspension of repeated sampling of vehicles due to traffic jams. . The above detailed description is a specific description of a feasible embodiment of the present invention, but this embodiment is not intended to limit the scope of the patent of the present invention. Any equivalent implementation or change that does not depart from this Maoyue technology sage should be It is included in the scope of patent of this 10 case. To sum up, this case is not only technically innovative, but also enhances the above-mentioned multiple effects over conventional items. It should have fully met the new and progressive statutory invention patent requirements. You are encouraged to apply in accordance with the law, and we ask your office for approval. This invention patent application is designed to encourage inventions, and it is a matter of virtue. (Please read the precautions on the back before writing this page) -------- Order --------- Line C. Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs This paper is suitable for China National Standard (CNS) A4 g (210 ^ 7 ^^ -----

Claims (1)

A8 Βδ C8 PA88036S.TWP-10/12 Patent scope: A vehicle identification system based on the hidden Markov model: continuous _ vehicle continuous height value, and the value of the sum of the sum of ② to the vehicle ㈣Data pre-processing unit, the vehicle's exterior shell material processing unit is then based on the vehicle's continuous f a vehicle's start (front) and end (back), and ^ all height values between the front and rear of the vehicle are taken as The appearance of the vehicle ^ During the system learning phase, the vehicle ’s exterior processing unit and the vehicle ’s appearance characteristics are sent to the vehicle ’s vehicle model. Unit 10 uses the vehicle ’s appearance characteristics to train the hidden type: After getting KK Zhou Suiyi Jing_, the vehicle-shaped parameter 1 15 will be sent to the fish. The vehicle unit will output the Tibetan Markov model based on the vehicle 2 ::: Yuanyuan model, and the output value will be Corresponding models of various types of hidden types, that is, vehicles appearing on the lanes 4. In the Ministry of Economic Affairs-quasi-stationary workers-printed by the Zuosha 20 Identification system, The scale model and vehicle model identification sheet of the car are first equipped for each type of vehicle: Early-Hidden Markov Model.-Belongs to the 4th patent scope! The hidden-Markov model identification system is described in the item, Among them, the car ^ and vehicle model identification sheet 4 are for each type of hidden vehicle model (inclusive) or more. Hidden Markov model. S 10 15 The Ministry of Economic Affairs Central Standards Bureau Offset Consumer Cooperative Cooperative India 20 8. The patent application model is based on the hidden vehicle model identification system as described in item # 1 of the patent application scope. The ~ Kuff mode is used as a system, in which the vehicle model learning unit and the vehicle identification unit are targeted at various types of vehicles. : White early Wuzang Markov pattern. Sharing a hidden secret = patent application Fan Ye! The item mentioned above is based on the hidden vehicle model recognition system, which is straight and straight, and f1 is used as the vehicle shape feature in front of the Zhongli vehicle shape information, which is then used for vehicle type learning sheet 7L and vehicle model identification. Unit processing. According to the vehicle model identification system based on the hidden Markov model described in item 1 of the scope of the patent application, the iJ management unit converts the vehicle height value into a frequency r (Spectrum coefficient before the profile data as the vehicle) The appearance characteristics are then processed by the vehicle model learning unit and the vehicle model recognition unit with the spectral coefficients. The province ’s Hou Hou patent scope item #! Is the vehicle model recognition system based on the hidden Markov model. The vehicle shape data pre-processing is performed as early as 4 vehicles. The light height value is converted into a linear estimation (Predlctive Coding, Lpc) coefficient as the vehicle appearance feature, and the linear prediction coding coefficient is used for the vehicle model learning unit and vehicle model identification. Unit processing. As described in item 1 of the scope of the patent application, the vehicle based on the hidden Markov model & identification system, system 'where the vehicle shape data pre-processing unit converts the vehicle height value into a cepstmm coefficient f It is a vehicle appearance characteristic, and then the cepstrum coefficient is provided for the vehicle model learning unit and the vehicle model identification unit to process. Please read the caution page on the back first) • < 1 II II-V— * ^ n li n I. -Li H ---------- U-(210X297, Ii