A Method and System for Volume and Surface Co-Segmentation Oriented to Hybrid Additive and Subtractive Manufacturing

Abstract

The present invention provides a method and system for volume and surface co-segmentation for hybrid additive and subtractive manufacturing. The present invention constructs a beam search tree, and based on the beam search tree, determines corresponding nodes and their ancestor nodes, combines the blocks represented by all the nodes, and obtains a final model segmentation scheme. The present invention can achieve segmentation into as few blocks as possible while satisfying manufacturing constraints, and each block can be manufactured through an "AM-then-SM" sequence. Moreover, this method is widely applicable to various three-dimensional models.

Description

Technical Field

[0001] The present invention belongs to the technical field of additive and subtractive hybrid manufacturing, and relates to a method and system for volume and surface co-segmentation for hybrid additive and subtractive manufacturing. Background Art

[0002] The statements in this section merely provide background technical information related to the present invention and do not necessarily constitute prior art.

[0003] Additive and subtractive hybrid manufacturing (ASHM) technology combines the capabilities of additive manufacturing and subtractive manufacturing, taking advantage of 3D printing in manufacturing complex geometries and improving material utilization, as well as the advantages of numerical control machining in high machining precision and high-quality surface treatment. Compared with 3D printing, ASHM will eliminate the stepped defects caused by layer-by-layer manufacturing and achieve higher product precision and excellent surface finish. Compared with numerical control machining, ASHM is more flexible in manufacturing complex structural parts with radical geometries. More and more people believe that in the near future, ASHM will show its great potential in the workshop and become an important part of industrial, aerospace, and consumer products.

[0004] However, hybrid manufacturing must simultaneously consider the computational geometry processes of both additive and subtractive manufacturing, which is far from fully automated, especially for generally complex geometries. Hybrid manufacturing is mainly achieved by a print head and a numerical control machining tool. The former constructs a 3D volume by depositing materials layer by layer (each layer is a planar plate of the object), and the latter engraves materials on the 3D surface along a specified spatial curve. The specific manufacturing process of ASHM consists of a set of "AM-then-SM" manufacturing stages. In each stage, first, the solid forming of the three-dimensional model is realized through additive manufacturing (AM), and then the surface of the three-dimensional model is finish-machined through subtractive manufacturing (SM). Constrained by the geometry of the three-dimensional model and the motion degrees of freedom of the print head and the numerical control machining tool, it is usually impossible to complete hybrid manufacturing in a single "AM-then-SM" stage.

[0005] Sequence planning is the task of determining an alternating sequence {A1, S1, A2, S2,..., An, Sn}, where A and S represent additive and subtractive steps respectively. The most crucial point in sequence planning is to explore the minimum number of "AM-then-SM" switches. Each Ai or Si requires recalibration and certain preprocessing of the print head or CNC machining tool, which may significantly affect the overall manufacturing efficiency. In addition, the position of the process switch always generates unwanted finishing artifacts, affecting the surface quality and manufacturing accuracy. Therefore, an ideal sequence planning needs to minimize the process switches in terms of machining efficiency and product quality. The key challenge in hybrid manufacturing sequence planning is that after each "AM-then-SM" stage, the shape to be achieved changes significantly, followed by dynamic changes in tool accessibility, which greatly increases the difficulty of sequence planning. Each A or S manufacturing process cannot violate the print head / machining tool accessibility and manufacturing dependency constraints, where the print head / machining tool is not allowed to collide with the already formed part of the model and the machine itself. In addition, the model must be a support-free structure during the additive stage. Since each "AM-then-SM" sequence manufactures a block of the model, this sequence planning problem can be equivalently transformed into a 3D model segmentation problem, that is: under the premise of meeting manufacturing constraints, the 3D model is segmented into as few blocks as possible, and each block is manufactured through an "AM-then-SM" sequence.

[0006] As understood by the inventors, there is little research on this problem currently, and existing methods limit this problem to specific shapes (mainly columnar structures) and cannot be applied to general models. Summary of the Invention

[0007] To solve the above problems, the present invention proposes a volume and surface co-segmentation method and system for hybrid additive and subtractive manufacturing. The present invention can be divided into as few blocks as possible under the premise of meeting manufacturing constraints, and each block can be manufactured through an "AM-then-SM" sequence. And this method is widely applicable to various 3D models.

[0008] According to some embodiments, the present invention adopts the following technical solutions:

[0009] A volume and surface co-segmentation method for hybrid additive and subtractive manufacturing, comprising:

[0010] Obtain a 3D model;

[0011] Perform subtractive collision detection on the 3D model and establish a graph to represent the reachability relationship between nodes;

[0012] Build a beam search tree. According to the beam search tree, determine the corresponding nodes and their ancestor nodes, combine the blocks represented by all nodes, and obtain the final model segmentation scheme. In each iteration of the search, perform the following steps:

[0013] Search for the blocks corresponding to the ancestor nodes, and find the nodes in the graph constructed by these blocks. Delete the nodes and the connected edges from the graph to form a new graph

[0014] Sample multiple candidate additive directions. Under each direction, horizontally slice the model, perform collision detection of the print head, and build a graph

[0015] Integrate and Build a graph

[0016] Iteratively merge the nodes in the graph to obtain the corresponding blocks; in

[0017] Select several blocks with the highest scores through the evaluation function and add them to the beam search tree.

[0018] As an alternative implementation, the specific process of performing subtractive collision detection on the three-dimensional model includes uniformly sampling on the model surface using centroid Voronoi tessellation. Each Voronoi cell is represented by its site. Perform collision detection of the tool for each site with other sites. If a site is unreachable in any direction, it is called an unreachable point.

[0019] As an alternative implementation, the specific process of building a graph includes building a graph where each node represents a sampling point, and each encoded edge represents the reachability relationship between two nodes. If there is an edge pointing from node A to B, it means that when the tool needs to process A, it is blocked by B and unreachable in a certain direction, and this direction is recorded by the encoding of the edge.

[0020] As an alternative implementation, the specific process of building a beam search tree includes calculating the segmentation result of the model using a top-down beam search algorithm, and each node in the tree represents a model block.

[0021] As an alternative implementation, the specific process of forming a new graph includes, at the initial moment of each iteration of the beam search, first finding the blocks corresponding to its ancestor nodes, and finding the graph contained in these blocks of the nodes, and building a new graph by deleting these nodes and the edges connected to them from the graph

[0022] As an optional implementation, multiple candidate additive directions are sampled, the model is sliced horizontally in each direction, collision detection of the print head is performed, and a map is established. The specific process includes each time a graph is created All are sampled uniformly on the Gaussian sphere N Dir candidate additive directions. In each direction, the model is first sliced horizontally with a constant step size, and the solid area between every two adjacent slice layers is called Slab.

[0023] Create a graph Each Slab corresponds to a node in the graph, performs collision detection on the print head, and establishes edges in the graph.

[0024] As a further example, if there is an edge from node A to node B, it means that if the slab corresponding to B has been formed, the print head will collide with the slab corresponding to B when printing the slab corresponding to A;

[0025] If a slab contains an area with a large overhang angle that requires a supporting structure, the corresponding node is set with a corresponding label.

[0026] As an optional implementation, integrating and Create a graph The specific process includes for each graph Copy it to another graph Then for the picture Each node in is assigned a label to indicate whether the node satisfies the additive and subtractive reachability constraints and the support-free constraint in the current state;

[0027] The label is obtained by and pictures It is calculated by determining the many-to-one mapping relationship between the sampling point and the Slab. If a sampling point is unreachable in the current state, then Set the node label corresponding to the mapped Slab accordingly.

[0028] As an alternative implementation, iteratively merge graphs The specific process of obtaining the corresponding blocks for the nodes in the graph includes: All leaf nodes whose labels are not the corresponding set labels are merged into new nodes with other labels, and then the remaining nodes are iteratively merged one by one into the new nodes through a greedy algorithm.

[0029] A volume and surface co-segmentation system for hybrid additive and subtractive manufacturing, comprising:

[0030] An input module for obtaining a 3D model;

[0031] A subtractive collision detection module for performing subtractive collision detection on the 3D model and establishing a graph to represent the reachability relationship between nodes;

[0032] A segmentation module for establishing a beam search tree, determining corresponding nodes and their ancestor nodes according to the beam search tree, and combining the blocks represented by all nodes to obtain a final model segmentation scheme;

[0033] The segmentation module specifically includes

[0034] A module for searching for the blocks corresponding to the ancestor nodes, finding the nodes in the graph constructed by these blocks, and deleting the nodes and the connected edges from the graph to form a new graph ;

[0035] A module for sampling multiple candidate additive directions, horizontally slicing the model in each direction, performing collision detection of the print head, and establishing a graph ;

[0036] A module for integrating and establishing a graph ;

[0037] A module for iteratively merging the nodes in the graph to obtain the corresponding blocks;

[0038] A module for selecting several blocks with the highest scores through an evaluation function and adding them to the beam search tree.

[0039] Compared with the prior art, the beneficial effects of the present invention are:

[0040] The present invention encodes manufacturing constraints into multiple directed graphs, simplifies the graphs through graph search and specific decision criteria, and finally can achieve the co-segmentation of the model body and surface with the goal of minimizing the number of switching times between additive and subtractive processes, so that each block can be manufactured through a single additive and subtractive process sequence under the premise of meeting manufacturing constraints, and this method is widely applicable to various 3D models.

[0041] To make the above objects, features, and advantages of the present invention more obvious and understandable, the following specific preferred embodiments are given in conjunction with the accompanying drawings and are described in detail as follows. Description of the Drawings

[0042] The specification drawings forming a part of the present invention are used to provide a further understanding of the present invention. The schematic embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention.

[0043] Figure 1It is the overall flowchart of the present invention;

[0044] Figure 2 It is the algorithm flowchart of the present invention;

[0045] Figure 3 It is the schematic diagram of the establishment process of the figure in the present invention;

[0046] Figure 4 It is the figure in the present invention;

[0047] Figure 5 It is the figure in the present invention;

[0048] Figure 6 It is the figure in the present invention;

[0049] Figure 7 It is the comparison result graph when different weights are set for each sub - evaluation function in the beam search evaluation function;

[0050] Figure 8 It is the partial result display graph in the present invention. Specific embodiments

[0051] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0052] It should be noted that the following detailed description is illustrative and is intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by those of ordinary skill in the technical field to which the present invention belongs.

[0053] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they specify the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0054] A volume - surface co - segmentation method for hybrid additive - subtractive manufacturing enables the model to be divided into as few blocks as possible while meeting manufacturing constraints, and each block can be manufactured through an "AM - then - SM" sequence. The present invention directly faces the manufacturing process, analyzes the three - dimensional model given by the user, and generates a model segmentation scheme for hybrid additive - subtractive manufacturing.

[0055] According to some examples, such as Figure 1 、Figure 2 As shown in the figure, it includes the following steps:

[0056] Step (1): Obtain the stl or obj file corresponding to the 3D model input by the user;

[0057] Step (2): Perform subtractive collision detection and establish a graph

[0058] Step (3): Establish a beam search tree and execute steps (4)-(8) in each iteration of the search;

[0059] Step (4): Establish a graph

[0060] Step (5): Sample N Dir candidate additive directions, and establish a graph under each direction

[0061] Step (6): Integrate and to establish a graph

[0062] Step (7): Iteratively merge the nodes in the graph to obtain the corresponding blocks;

[0063] Step (8): Select several blocks with the highest scores through the evaluation function and add them to the beam search tree;

[0064] Step (9): Obtain the final model segmentation scheme according to the established beam search tree.

[0065] Specific elaboration is carried out for each step:

[0066] In this embodiment, in step (1), the 3D model is a closed model.

[0067] In this embodiment, step (2) mainly includes the following steps:

[0068] Step (2-1): Perform uniform sampling on the model surface using centroid Voronoi tessellation, and each Voronoi cell is represented by its site. Then perform the reachability analysis of the machining tool: perform collision detection of the tool for each sampling point (site) with other sampling points. If a sampling point is unreachable in any direction, it is called an "unreachable point".

[0069] Step (2-2): Establish a graph Each node represents a sampling point, and each encoded edge represents the reachability relationship between two nodes. If there is an edge pointing from node A to B, it means that when the tool needs to machine A, it is blocked and unreachable by B in a certain direction, and this direction is recorded by the encoding of the edge, such as Figure 3 (a) of

[0070] In this embodiment, step (3) mainly includes the following steps:

[0071] Step (3-1): Calculate the segmentation result of the model using a top-down beam search algorithm. This algorithm is actually a process of building a beam search tree, where each node in the tree represents a model block. In each iteration of the beam search, the nodes to be added to the beam search tree are selected by the methods in steps (4)-(8).

[0072] Step (4) mainly includes the following steps:

[0073] Step (4-1): At the initial moment of each iteration of the beam search, first find the blocks corresponding to its ancestor nodes, and find the nodes contained in these blocks. A new graph is established by deleting these nodes and the edges connected to them from (such as (b) of Figure 3 ).

[0074] Step (5) mainly includes the following steps:

[0075] Step (5-1): Each time a graph is established, N Dir candidate additive manufacturing directions are uniformly sampled on the Gaussian sphere. In each direction, first, the model is horizontally sliced with a constant step size, and the solid region between every two adjacent sliced layers is called a "Slab". Then, a graph is established. Each Slab corresponds to a node in the graph. The collision detection of the print head is performed by calling the method in the article "As-continuous-as-possible Extrusion-based Fabrication of Surface Models", and the edges in the graph are established. If there is an edge pointing from node A to node B, it means that if the Slab corresponding to B has been formed, the print head will collide with the Slab corresponding to B when printing the Slab corresponding to A. In addition, if a Slab contains an area with an overhanging angle too large and requires a support structure, the corresponding node is set with a red label (such as Figure 4 (a) of ). During the collision detection process, we do not need to extract the tetrahedrons of the three-dimensional model for calculation, but calculate using the closed curves obtained by slicing, thus greatly improving the calculation speed (such asFigure 4 (b) of

[0076] In this embodiment, step (6) mainly includes the following steps:

[0077] Step (6-1): For each graph Copy it to another graph Then for the graph Assign a label to each node in it to indicate whether this node satisfies the additive and subtractive accessibility constraints and the support-free constraint in the current state. This label is calculated through the many-to-one mapping between graph and graph , that is, by determining the many-to-one spatial position mapping relationship between the sampling points and the slabs (such as Figure 5 (a) and (b) of . If a certain sampling point is unreachable in the current state, then in graph set the node label corresponding to the slab it maps to red. Similarly, if the node label corresponding to a certain slab in graph is red, then set its corresponding node label in

[0078] to red as well (as shown in Fig. 5(c)).

[0079] Step (7-1): For each graph established Merge all leaf nodes with non-red labels into a new blue node, and then iteratively merge the remaining nodes into the new blue node one by one through the greedy algorithm (breadth-first search). The merging criterion is: the node to be merged is a leaf node and its label is not red. When graph cannot be merged anymore, the combination of all slabs included in the new blue node is a candidate block of the 3D model ( Figure 6 shows an example of iterative merging of graph ).

[0080] Step (8) mainly includes the following steps:

[0081] Step (8-1): All graphs calculated in each iteration of beam search each correspond to a candidate block. Score each candidate block through an evaluation function:

[0082] F(B) = ∑w k f k (B), k ∈ [1, 2, …, 6], where f k (B) represents a sub-evaluation function, and there are six in total:

[0083] fsize (B) = ∑ Length(s B ), s B ∈ S B ;

[0084] f priority (B) = ∑ Priority(s B ), s B ∈ S B ;

[0085] f connect (B) = Component(S B );

[0086]

[0087]

[0088]

[0089] where S B represents the set of all Slabs in block B; s B represents the closed curve at the bottom of the Slab; Length(s B ) represents the length of the curve; Priority(s B ) represents the priority value score of all sampling points in the Slab (the more other sampling points a sampling point obscures, the higher the score); Component(S B ) represents the number of connected components of block B; represents the additive direction of block x; P represents the set of all cutting planes on block x; Distance(p, v) represents the Euclidean distance between the vulnerable point v on the block and the cutting plane p; Project(p) represents the projection area of the block on the two-dimensional plane that exceeds the cutting plane p; Area(Project(p)) represents the area of the projection area.

[0090] The weight w k of each sub-evaluation function is restricted between 0 and 1 through normalization, and ∑ w k = 1. Then, the W candidate blocks with the highest scores are each represented by a node and added to the beam search tree. Figure 7 Shows the segmentation results of the model when the weights of the six sub-evaluation functions are set to 1 respectively.

[0091] In this embodiment, step (9) mainly includes the following steps:

[0092] Step (9-1): Take out the node with the highest score obtained in the last iteration of beam search, and then sequentially take out its ancestor nodes. Then combine the blocks represented by these nodes to obtain the final 3D model segmentation scheme, where each block is manufactured using an "AM-then-SM" sequence once.

[0093] Those skilled in the art should understand that the embodiments of the present invention can be provided as a method, a system, or a computer program product. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0094] The present invention is described with reference to the flowcharts and / or block diagrams of methods, apparatuses (systems), and computer program products according to the embodiments of the present invention. It should be understood that each flow and / or block in the flowchart and / or block diagram can be implemented by computer program instructions, and the combination of the flows and / or blocks in the flowchart and / or block diagram can also be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices generate means for implementing the functions specified in one Figure 1 one or more flows and / or blocks Figure 1 or multiple flows and / or blocks

[0095] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory generate a manufactured article including instruction means, and the instruction means implement the functions specified in one Figure 1 one or more flows and / or blocks Figure 1 or multiple flows and / or blocks

[0096] These computer program instructions can also be loaded onto a computer or other programmable data processing device, so that a series of operation steps are executed on the computer or other programmable device to generate a computer-implemented process, and thus the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in one Figure 1 one or more flows and / or blocks Figure 1 or multiple flows and / or blocks

[0097] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

[0098] Although the specific implementation manners of the present invention have been described above in conjunction with the accompanying drawings, it is not a limitation on the protection scope of the present invention. Those skilled in the art should understand that, based on the technical solutions of the present invention, various modifications or deformations that can be made by those skilled in the art without creative efforts are still within the protection scope of the present invention.

Claims

1. A method for volume and surface co-segmentation for hybrid additive and subtractive manufacturing, characterized in that Including: Obtain a 3D model; Perform subtractive collision detection on the 3D model and establish a graph to represent the reachability relationship between nodes; Establish a beam search tree. According to the beam search tree, determine the corresponding nodes and their ancestor nodes, combine the blocks represented by all nodes, and obtain the final model segmentation scheme. The process of establishing the beam search tree specifically includes the following steps: Search for the chunks corresponding to the ancestor nodes, and find that these chunks contain the nodes of the constructed graph. Delete the said nodes and the edges connected to them from the graph to form a new graph ; Sample multiple candidate additive directions, horizontally slice the model under each direction, perform collision detection of the print head, and establish Figure ; Integrate new figure and figure , establish figure ; Iterative merging graph The nodes in it are obtained to get the corresponding blocks; Select several blocks with the highest scores through an evaluation function and add them to the beam search tree; Integrate new figures and figure , establish figure . The specific process includes, for each figure , copying it to another figure , and then assigning a label to each node in figure to indicate whether this node satisfies the additive and subtractive reachability constraints and the support-free constraint in the current state; The said label is calculated through the many-to-one mapping between and , that is, by determining the many-to-one spatial position mapping relationship between the sampling points and the Slabs. If a certain sampling point is unreachable in the current state, then in , the node label corresponding to the Slab mapped by it is set accordingly.

2. The volume and surface co-segmentation method for hybrid additive and subtractive manufacturing according to claim 1, characterized in that The specific process of performing subtractive collision detection on the 3D model includes implementing uniform sampling on the model surface using centroid Voronoi tessellation. Each Voronoi cell is represented by its site. Perform tool collision detection between each site and other sites. If a site is unreachable in any direction, it is called an unreachable point.

3. The volume and surface co-segmentation method for hybrid additive and subtractive manufacturing according to claim 1, characterized in that The specific process of establishing the graph includes establishing the graph , where each node represents a sampling point, and each encoded edge represents the reachability relationship between two nodes. If there is an edge pointing from node A to B, it means that when the tool needs to process A, it is blocked by B and unreachable in a certain direction, and this direction is recorded by the encoding of the edge.

4. The method for co-segmentation of solid and surface for hybrid additive and subtractive manufacturing according to claim 1, characterized in that, The specific process of establishing the beam search tree includes calculating the segmentation result of the model using a top-down beam search algorithm. Each node in the tree represents a model block.

5. A method for co-segmentation of solid and surface for hybrid additive and subtractive manufacturing according to claim 1, characterized in that Form a new graph The specific process of forming includes, at the initial moment of each iteration of beam search, first finding the blocks corresponding to its ancestor nodes and finding the nodes of the graph contained in these blocks. By deleting these nodes and the edges connected to them from the graph , a new graph is established. The nodes of the graph are found by starting from the graph . A new graph is established by deleting these nodes and the edges connected to them from the graph . .

6. A method for co-segmentation of solid and surface for hybrid additive and subtractive manufacturing according to claim 1, characterized in that Sample multiple candidate additive directions, horizontally slice the model in each direction, perform collision detection of the print head, and establish the graph The specific process of includes that every time a graph is established, uniformly sample candidate additive directions on the Gaussian sphere. In each direction, first horizontally slice the model with a constant step size, and call the solid region between every two adjacent slice layers a Slab; Build a graph , where each Slab corresponds to a node in the graph, perform collision detection on the print head, and build the edges in the graph.

7. The method for volume and surface co-segmentation for hybrid additive and subtractive manufacturing according to claim 6, characterized in that, If there is an edge pointing from node A to node B, it means that if the slab corresponding to B has been formed, the print head will collide with the slab corresponding to B when printing the slab corresponding to A; If a slab contains an area with an overhanging angle that is too large and requires a support structure, set the corresponding label for its corresponding node.

8. A method for co-segmentation of solid and surface for hybrid additive and subtractive manufacturing according to claim 1, characterized in that Iterative Merging Graph The specific process of obtaining the corresponding blocks by merging the nodes in includes, for each graph established, merging all leaf nodes with labels that are not the corresponding set labels into new nodes with other labels, and then iteratively merging the remaining nodes into the new nodes one by one through a greedy algorithm.

9. A volume and surface co-segmentation system for hybrid additive and subtractive manufacturing, which is used to execute a volume and surface co-segmentation method for hybrid additive and subtractive manufacturing according to any one of claims 1-8, characterized in that, Including: An input module for obtaining a 3D model; A subtractive collision detection module for performing subtractive collision detection on the 3D model and establishing a graph to represent the reachability relationship between nodes; A segmentation module for establishing a beam search tree, determining the corresponding nodes and their ancestor nodes according to the beam search tree, and combining the blocks represented by all nodes to obtain the final model segmentation scheme; The segmentation module specifically includes A module for searching for blocks corresponding to ancestor nodes, finding nodes in these blocks that contain nodes of the constructed graph, deleting the nodes and the edges connected to them from the graph to form a new graph ; A module for sampling multiple candidate additive directions, horizontally slicing the model under each direction, performing collision detection of the print head, and establishing the figure ; Module for integrating new figure and figure to establish figure ; Module for iteratively merging nodes in a graph to obtain corresponding segmented modules; A module for selecting several blocks with the highest scores through an evaluation function and adding them to the beam search tree.

Images