Research On The Theory And Method Of Construction Project Supervision Based On Computer Vision Technology

The construction industry of China is booming. During the eleventh five-year plan, the certified general contractors and specified contracors accomplished production value of 9520.6 billion yuan in 2010. The added value of contruction industry reached 2645.1 billion yuan. During the eleventh five-year plan, the average growth rate of construction industry reached 20.6%, which exceeded the development target. During this fast developing process, China’s construction industry faces a series of problems, which enhanced the needs of government supervision.. One of the key point of the industry supervision is the supervision of construction projects, because they are the entities which form the industry, and they are the cross point of each stakeholders. In civil economy, government leading real state investment is an important support. The supervision of government invested construction projects has great significance on the benign development of civil construction industry. The Chinese central government issued the executive solution to special significant problems in construction industry in 2009 and the executive opinions for the information disclosure and credit system construction in construction industry in 2011, which provided comprehensive regulations for specific problems in the lifecycle for a construction project. The main content inludes:from decision making, land expropriation, urban and rural planning, and biding in early stage, to engineering schedule, fund utilization, engineering quality, security and environment protection in the executing stage. But there lacks supervision theory and technical framework for the basic entities, which are the construction projects themselves.To solve the problem. Chapter 2 analyzed the basic problems and the ruling elements based on the summary of former studies. Chapter 3 studied the process supervision principles for construction projects from a new perspective of stochastic process and evolutionary game theory. Chapter 4 discussed the feasibility of thechonical framework based on high resolution remote sensing images and computer vision technology. Chapter 5 proposed a supervision target recognition method based on the engineering appearance vision features. Chapter 6 proposed a real-time construction project supervision model based on engineering logic. Chapter 7 proposed the basic framework of a supervision system based on the theories and technologies provided by former chapters. Based on above, the main work of this paper is as followed:(1) To extract and study the basic problems of the supervision problem, firstly, this paper summarized the understanding and analysis of former studies in the perspectives of construction project targets. At the same time, this paper studied derived problems of supervision from the perspectives of behavior and environment, which inculdes the complexity brought by the problems in supervision, rent-seeking, agent construction system and project groups. Based on this, and concentrating on the study range of "construction project", and ignoring some unrelated problems, this paper extracted the basic contradiction and problem, which is the target excursion and out of control problem. The root of the problem lies in the contradictions in the benefit persuing behaviors of the investors, the users and the construction agents, which is entangled by outside problems of project group and publicity.(2) To find out the feasible theory for the basic problems in the former paragraph, the comparison between commonly used supervision theories should be carried out, so that references for construction project supervison can be provided. This paper analyzed the solutions provided by tranditional theories like principal agent theory, evolutionary theory and their derived thoeries, and their shortages are indicated. And the significance of process and information for supervision is pointed out.(3) Based on the thoery studies of mutiple perspectives, Markov chain is selected as the basic model to discribe the basic problem as principal-agent two population evolutionary game model. This model can perform good simulation on process. At the same time, the influences of the information broadcasting cost and information distribution on the game equilibrium is studied. As a result, two important factors of supervision optimization is extracted. Those are supervision success probablity (SSP) and supervision cost (SC), and the influence on the equilibrium is studied.(4) To increase SSP and decrease SC, the feasibility of combining high-resolution remote sensing image(HRSI) and computer vision(CV) is considered. This paper compared CV recognition technologies like frequency domain based, region growing based and feature recognition based, to find out the cross point between HRSI and construction project supervision. The importance of "object" based recognition is pointed out. The result indicated that it is feasible to build a new supervison technology, which has higher SSP and lower SC and based on feature recognition and HRSI of engineering sites.(5) To find out optimized recognition method for engineering groud objects in HRSI, this paper compare the feature diffenences caused by lightening condition, satellite inclination and the height property. It is pointed out that the classifier should be built based on the feature series frameworks. And the recognition method of construction objects is derieved based on outline features, texture features and height features.(6) The mapping method of appearance schedule plan based on engineering logic is derieved. The alignment method of the plan and collected images is proposed, which is based on the satellite inclination data and the imaging-forming principles. So that the comparison method between the plan and the collected data is derieved. The characterization methods for main security and environmental targets are derived. What’s more, the characterization methods, which are based on the real-time schdule supervison model, for main security and environmental targets are derived.(7) Based on the contents in chapter 1-6, the CV based practice framework technology of construction project supervision system is proposed. The basic use cases and worflows in different scenes are derieved. The results show that the proposed supervison framework is more real-timing, and with lower cost and highter effciency, compared with tranditional methods like sending out supervison or audit agents. It can be seen that the proposed engineering supervision technical framework is a great progress from extensive supervision towards high efficeincy intensive supervision.This paper provide four main contributions to the field. ① The supervisor-agent evolutionary game simulation based on Markov chain; ② The target recognition method based on the appreaence process featrues of engineering; ③ The real-time supervision model based on engineering logic;④The design of the construction project supervision information system.