1. A scenic road landscape evaluation method based on deep learning is characterized by comprising the following steps:

a. homogenizing and sampling the landscape;

b. deep learning;

c. landscape sampling input, automatic evaluation and result output;

the method comprises the following steps that a, a scenic road landscape sampling vehicle is adopted, four front, rear, left and right sampling cameras are arranged on the roof of the vehicle, the four cameras fix landscape sampling interval time, a fixed number of scenic road landscape sampling photos are shot at fixed intervals to serve as landscape sampling results, and all the scenic road landscape sampling photos are stored in storage equipment of the scenic road landscape sampling vehicle;

b, by means of a deep learning model, performing landscape feature extraction by a large number of landscape road landscape sampling photos of different time periods, different seasons and different feature road sections in the early stage of evaluation, manually giving different weights to different landscape features, and performing weighted summation by the deep learning model according to a scoring result so as to eliminate errors of manual evaluation and improve the objectivity and accuracy of evaluation;

and c, inputting the scenic road landscape sampling picture into the scenic road landscape evaluation model system by means of the scenic road landscape evaluation model system, automatically extracting landscape characteristics of the sampling picture by the scenic road landscape evaluation model system, scoring, weighting and summing to obtain an evaluation result, wherein the scenic road landscape evaluation model system comprises a terminal, and the evaluation result is fed back through the terminal.

2. The scenic road landscape evaluation method based on deep learning of claim 1, wherein: the scenic road landscape sampling vehicle samples in different time periods and different seasons, so that accurate and complete scenic road landscape sampling is guaranteed, and the scene of the scenic road landscape is restored completely.

3. The scenic road landscape evaluation method based on deep learning of claim 1, wherein: the deep learning model adopts one or more of a circulation network, a convolution network, a common deep network, a deep production model and an auto-encoder.

4. The scenic road landscape evaluation method based on deep learning of claim 1, wherein: the landscape features include road grade, road straightness, road congestion, field of view, forest stand, color.

5. The scenic road landscape evaluation method based on deep learning of claim 1, wherein: the terminal is any one or more of a mobile phone and a computer.

Background

The existing scenic tract landscape evaluation method comprises a minimum space analysis method, a landscape comprehensive evaluation index method, an equidistant expert group visual evaluation method, an object model, an entropy weight method, a visual landscape evaluation method and the like, but the methods are ideal mathematical analysis models, the scenic tract is checked on site and photographed and sampled manually, and then the pictures are quantitatively evaluated manually through the mathematical models or quantitative analysis models.

The above method mainly has the following disadvantages: firstly, a large amount of manpower is required to be invested for on-site checking, photographing sampling and scoring evaluation, so that the efficiency is low; secondly, the scoring amount of the scenic road photos is different from person to person, and the scenic road photos have larger errors and cannot accurately reflect the quality of the scenic road photos; thirdly, the length of the landscape street is hundreds of kilometers, if the landscape street completely depends on manual shooting to mark time consumption, and the evaluation precision is greatly reduced under the influence of sampling points and sampling intervals.

Disclosure of Invention

The invention aims to overcome the defects of the technology and provide a scenic road landscape evaluation method based on deep learning.

In order to solve the technical problems, the technical scheme provided by the invention is a scenic road landscape evaluation method based on deep learning, which comprises the following steps: the method comprises the following steps: a. homogenizing and sampling the landscape; b. deep learning; c. landscape sampling input, automatic evaluation and result output;

the method comprises the following steps that a, a scenic road landscape sampling vehicle is adopted, four front, rear, left and right sampling cameras are arranged on the roof of the vehicle, the four cameras fix landscape sampling interval time, a fixed number of scenic road landscape sampling photos are shot at fixed intervals to serve as landscape sampling results, and all the scenic road landscape sampling photos are stored in storage equipment of the scenic road landscape sampling vehicle;

b, by means of a deep learning model, performing landscape feature extraction by a large number of landscape road landscape sampling photos of different time periods, different seasons and different feature road sections in the early stage of evaluation, manually giving different weights to different landscape features, and performing weighted summation by the deep learning model according to a scoring result so as to eliminate errors of manual evaluation and improve the objectivity and accuracy of evaluation;

and c, inputting the scenic road landscape sampling picture into the scenic road landscape evaluation model system by means of the scenic road landscape evaluation model system, automatically extracting landscape characteristics of the sampling picture by the scenic road landscape evaluation model system, scoring, weighting and summing to obtain an evaluation result, wherein the scenic road landscape evaluation model system comprises a terminal, and the evaluation result is fed back through the terminal.

As an improvement, the scenic road landscape sampling vehicle samples in different time periods and different seasons, so that accurate and complete scenic road landscape sampling is guaranteed, and the scene of the scenic road landscape is restored completely.

As an improvement, the deep learning model adopts one or more of a cyclic network, a convolutional network, a common deep network, a deep production model and an auto-encoder.

As an improvement, the landscape features include road grade, road flatness, road congestion, field of view, forest stand, color.

As an improvement, the terminal is any one or more of a mobile phone and a computer.

Compared with the prior art, the invention has the advantages that: by means of a standardized automatic evaluation model, the evaluation efficiency of landscape quality of scenic roads is improved; by means of a deep learning model, automation of scenic road landscape feature sampling, quality evaluation, result feedback and model optimization is achieved, subjective interference factors in the artificial evaluation process are eliminated, and evaluation precision is greatly improved; the influence of landscape channel scale factors is eliminated, and the homogeneity of landscape sampling is ensured.

Drawings

FIG. 1 is a flow chart of a scenic road landscape evaluation method based on deep learning according to the present invention.

Detailed Description

The following describes the scenic road landscape evaluation method based on deep learning in further detail with reference to the accompanying drawings.

With reference to fig. 1, a scenic road landscape evaluation method based on deep learning includes the following steps: a. homogenizing and sampling the landscape; b. deep learning; c. landscape sampling input, automatic evaluation and result output;

the method comprises the following steps that a, a landscape road landscape sampling vehicle is adopted, four front, rear, left and right sampling cameras are arranged on the roof of the vehicle, the four cameras fix landscape sampling interval time, a fixed number of landscape road landscape sampling photos are shot at fixed intervals to serve as landscape sampling results, all the landscape road landscape sampling photos are stored in storage equipment of the landscape road landscape sampling vehicle, and automatic equal-time and equal-distance sampling is carried out, so that the sampling efficiency and precision are greatly improved;

b, by means of a deep learning model, performing landscape feature extraction by a large number of landscape road landscape sampling photos of different time periods, different seasons and different feature road sections in the early stage of evaluation, manually giving different weights to different landscape features, and performing weighted summation by the deep learning model according to a scoring result so as to eliminate errors of manual evaluation and improve the objectivity and accuracy of evaluation;

and c, inputting the scenic road landscape sampling picture into the scenic road landscape evaluation model system by means of the scenic road landscape evaluation model system, automatically extracting landscape characteristics of the sampling picture by the scenic road landscape evaluation model system, scoring, weighting and summing to obtain an evaluation result, realizing automation of scenic road landscape evaluation, and improving the scenic road landscape evaluation efficiency in multiples, wherein the scenic road landscape evaluation model system comprises a terminal, and the evaluation result is fed back through the terminal.

The scenic road landscape sampling vehicle samples in different time periods and different seasons, so that accurate and complete scenic road landscape sampling is guaranteed, and the scene of the scenic road landscape is restored completely.

The deep learning model adopts one or more of a circulation network, a convolution network, a common deep network, a deep production model and an auto-encoder.

The landscape features include road grade, road straightness, road congestion, field of view, forest stand, color.

The terminal is any one or more of a mobile phone and a computer.

By means of a standardized automatic evaluation model, the evaluation efficiency of landscape quality is improved; by means of a deep learning model, automation of scenic road landscape feature sampling, quality evaluation, result feedback and model optimization is achieved, subjective interference factors in the artificial evaluation process are eliminated, and evaluation precision is greatly improved; the influence of landscape channel scale factors is eliminated, and the homogeneity of landscape sampling is ensured.

The present invention and the embodiments thereof have been described above without limitation, and those skilled in the art will be able to devise embodiments similar to the above embodiments without departing from the spirit and scope of the invention.